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_tape_position_ok(JCR *jcr, DEVICE *dev);
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 */
197 autoload_device(dcr, 0, NULL);
200 * This code ensures that the device is ready for
201 * reading. If it is a file, it opens it.
202 * If it is a tape, it checks the volume name
204 Dmsg1(100, "bstored: open vol=%s\n", dcr->VolumeName);
205 if (dev->open(dcr, OPEN_READ_ONLY) < 0) {
206 Jmsg3(jcr, M_WARNING, 0, _("Read open device %s Volume \"%s\" failed: ERR=%s\n"),
207 dev->print_name(), dcr->VolumeName, dev->bstrerror());
210 Dmsg1(50, "opened dev %s OK\n", dev->print_name());
212 /* Read Volume Label */
213 Dmsg0(50, "calling read-vol-label\n");
214 vol_label_status = read_dev_volume_label(dcr);
215 switch (vol_label_status) {
218 memcpy(&dev->VolCatInfo, &dcr->VolCatInfo, sizeof(dev->VolCatInfo));
222 * Send error message generated by read_dev_volume_label()
223 * only we really had a tape mounted. This supresses superfluous
224 * error messages when nothing is mounted.
226 if (tape_previously_mounted) {
227 Jmsg(jcr, M_WARNING, 0, "%s", jcr->errmsg);
231 if (tape_initially_mounted) {
232 tape_initially_mounted = false;
235 /* If polling and got a previous bad name, ignore it */
236 if (dev->poll && strcmp(dev->BadVolName, dev->VolHdr.VolumeName) == 0) {
239 bstrncpy(dev->BadVolName, dev->VolHdr.VolumeName, sizeof(dev->BadVolName));
241 if (!unload_autochanger(dcr, -1)) {
242 /* at least free the device so we can re-open with correct volume */
247 Jmsg1(jcr, M_WARNING, 0, "%s", jcr->errmsg);
249 tape_previously_mounted = true;
252 * If the device requires mount, close it, so the device can be ejected.
254 if (dev->requires_mount()) {
258 /* Call autochanger only once unless ask_sysop called */
259 if (try_autochanger) {
261 Dmsg2(200, "calling autoload Vol=%s Slot=%d\n",
262 dcr->VolumeName, dcr->VolCatInfo.Slot);
263 stat = autoload_device(dcr, 0, NULL);
265 try_autochanger = false;
266 continue; /* try reading volume mounted */
268 /* Try closing and re-opening */
271 if (dev->open(dcr, OPEN_READ_ONLY) >= 0) {
276 /* Mount a specific volume and no other */
277 Dmsg0(200, "calling dir_ask_sysop\n");
278 if (!dir_ask_sysop_to_mount_volume(dcr, ST_READ)) {
279 goto get_out; /* error return */
281 try_autochanger = true; /* permit using autochanger again */
282 continue; /* try reading again */
287 Jmsg1(jcr, M_FATAL, 0, _("Too many errors trying to mount device %s for reading.\n"),
294 set_jcr_job_status(jcr, JS_Running);
295 dir_send_job_status(jcr);
296 Jmsg(jcr, M_INFO, 0, _("Ready to read from volume \"%s\" on device %s.\n"),
297 dcr->VolumeName, dev->print_name());
301 dcr->clear_reserved();
303 * Normally we are blocked, but in at least one error case above
304 * we are not blocked because we unsuccessfully tried changing
307 if (dev->is_blocked()) {
308 dev->dunblock(DEV_LOCKED);
310 Dmsg1(950, "jcr->dcr=%p\n", jcr->dcr);
316 * Acquire device for writing. We permit multiple writers.
317 * If this is the first one, we read the label.
319 * Returns: NULL if failed for any reason
321 * Note, normally reserve_device_for_append() is called
322 * before this routine.
324 DCR *acquire_device_for_append(DCR *dcr)
326 DEVICE *dev = dcr->dev;
330 init_device_wait_timers(dcr);
332 dev->dblock(BST_DOING_ACQUIRE);
333 Dmsg1(100, "acquire_append device is %s\n", dev->is_tape()?"tape":
334 (dev->is_dvd()?"DVD":"disk"));
337 * With the reservation system, this should not happen
339 if (dev->can_read()) {
340 Jmsg1(jcr, M_FATAL, 0, _("Want to append, but device %s is busy reading.\n"), dev->print_name());
341 Dmsg1(200, "Want to append but device %s is busy reading.\n", dev->print_name());
346 * have_vol defines whether or not mount_next_write_volume should
347 * ask the Director again about what Volume to use.
349 if (dev->can_append() && dcr->is_suitable_volume_mounted() &&
350 strcmp(dcr->VolCatInfo.VolCatStatus, "Recycle") != 0) {
351 Dmsg0(190, "device already in append.\n");
353 * At this point, the correct tape is already mounted, so
354 * we do not need to do mount_next_write_volume(), unless
355 * we need to recycle the tape.
357 if (dev->num_writers == 0) {
358 memcpy(&dev->VolCatInfo, &dcr->VolCatInfo, sizeof(dev->VolCatInfo));
360 if (!is_tape_position_ok(jcr, dev)) {
364 Dmsg1(190, "jid=%u Do mount_next_write_vol\n", (uint32_t)jcr->JobId);
365 if (!dcr->mount_next_write_volume()) {
366 if (!job_canceled(jcr)) {
367 /* Reduce "noise" -- don't print if job canceled */
368 Jmsg(jcr, M_FATAL, 0, _("Could not ready device %s for append.\n"),
370 Dmsg1(200, "Could not ready device %s for append.\n",
375 Dmsg2(190, "Output pos=%u:%u\n", dcr->dev->file, dcr->dev->block_num);
378 dev->num_writers++; /* we are now a writer */
379 if (jcr->NumWriteVolumes == 0) {
380 jcr->NumWriteVolumes = 1;
382 dev->VolCatInfo.VolCatJobs++; /* increment number of jobs on vol */
383 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
388 dcr->clear_reserved();
389 dev->dunblock(DEV_LOCKED);
390 return ok ? dcr : NULL;
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 static bool is_tape_position_ok(JCR *jcr, DEVICE *dev)
408 if (dev->is_tape() && dev->num_writers == 0) {
409 int32_t file = dev->get_os_tape_file();
410 if (file >= 0 && file != (int32_t)dev->get_file()) {
411 Jmsg(jcr, M_FATAL, 0, _("Invalid tape position on volume \"%s\""
412 " on device %s. Expected %d, got %d\n"),
413 dev->VolHdr.VolumeName, dev->print_name(), dev->get_file(), file);
422 * This job is done, so release the device. From a Unix standpoint,
423 * the device remains open.
425 * Note, if we are spooling, we may enter with the device locked.
426 * However, in all cases, unlock the device when leaving.
429 bool release_device(DCR *dcr)
432 DEVICE *dev = dcr->dev;
435 /* lock only if not already locked by this thread */
436 if (!dcr->is_dev_locked()) {
440 Dmsg2(100, "release_device device %s is %s\n", dev->print_name(), dev->is_tape()?"tape":"disk");
442 /* if device is reserved, job never started, so release the reserve here */
443 dcr->clear_reserved();
445 if (dev->can_read()) {
446 dev->clear_read(); /* clear read bit */
447 Dmsg0(100, "dir_update_vol_info. Release0\n");
448 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
451 } else if (dev->num_writers > 0) {
453 * Note if WEOT is set, we are at the end of the tape
454 * and may not be positioned correctly, so the
455 * job_media_record and update_vol_info have already been
456 * done, which means we skip them here.
459 Dmsg1(100, "There are %d writers in release_device\n", dev->num_writers);
460 if (dev->is_labeled()) {
461 Dmsg2(200, "dir_create_jobmedia. Release vol=%s dev=%s\n",
462 dev->VolCatInfo.VolCatName, dev->print_name());
463 if (!dev->at_weot() && !dir_create_jobmedia_record(dcr)) {
464 Jmsg(jcr, M_FATAL, 0, _("Could not create JobMedia record for Volume=\"%s\" Job=%s\n"),
465 dcr->VolCatInfo.VolCatName, jcr->Job);
467 /* If no more writers, and no errors, and wrote something, write an EOF */
468 if (!dev->num_writers && dev->can_write() && dev->block_num > 0) {
470 write_ansi_ibm_labels(dcr, ANSI_EOF_LABEL, dev->VolHdr.VolumeName);
472 if (!dev->at_weot()) {
473 dev->VolCatInfo.VolCatFiles = dev->file; /* set number of files */
474 /* Note! do volume update before close, which zaps VolCatInfo */
475 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
476 Dmsg2(200, "dir_update_vol_info. Release vol=%s dev=%s\n",
477 dev->VolCatInfo.VolCatName, dev->print_name());
479 if (!dev->is_busy()) { /* if not being used */
480 volume_unused(dcr); /* we obviously are not using the volume */
486 * If we reach here, it is most likely because the job
487 * has failed, since the device is not in read mode and
488 * there are no writers. It was probably reserved.
494 /* If no writers, close if file or !CAP_ALWAYS_OPEN */
495 if (dev->num_writers == 0 && (!dev->is_tape() || !dev->has_cap(CAP_ALWAYSOPEN))) {
496 dvd_remove_empty_part(dcr); /* get rid of any empty spool part */
500 /* Fire off Alert command and include any output */
501 if (!job_canceled(jcr) && dcr->device->alert_command) {
505 char line[MAXSTRING];
506 alert = get_pool_memory(PM_FNAME);
507 alert = edit_device_codes(dcr, alert, dcr->device->alert_command, "");
508 bpipe = open_bpipe(alert, 0, "r");
510 while (fgets(line, sizeof(line), bpipe->rfd)) {
511 Jmsg(jcr, M_ALERT, 0, _("Alert: %s"), line);
513 status = close_bpipe(bpipe);
519 Jmsg(jcr, M_ALERT, 0, _("3997 Bad alert command: %s: ERR=%s.\n"),
520 alert, be.bstrerror(status));
523 Dmsg1(400, "alert status=%d\n", status);
524 free_pool_memory(alert);
526 pthread_cond_broadcast(&dev->wait_next_vol);
527 Dmsg1(100, "JobId=%u broadcast wait_device_release\n", (uint32_t)jcr->JobId);
528 pthread_cond_broadcast(&wait_device_release);
531 detach_dcr_from_dev(dcr);
533 if (jcr->read_dcr == dcr) {
534 jcr->read_dcr = NULL;
536 if (jcr->dcr == dcr) {
541 Dmsg2(100, "===== Device %s released by JobId=%u\n", dev->print_name(),
542 (uint32_t)jcr->JobId);
547 * Clean up the device for reuse without freeing the memory
549 bool clean_device(DCR *dcr)
552 dcr->keep_dcr = true; /* do not free the dcr */
553 ok = release_device(dcr);
554 dcr->keep_dcr = false;
559 * Create a new Device Control Record and attach
560 * it to the device (if this is a real job).
561 * Note, this has been updated so that it can be called first
562 * without a DEVICE, then a second or third time with a DEVICE,
563 * and each time, it should cleanup and point to the new device.
564 * This should facilitate switching devices.
565 * Note, each dcr must point to the controlling job (jcr). However,
566 * a job can have multiple dcrs, so we must not store in the jcr's
567 * structure as previously. The higher level routine must store
568 * this dcr in the right place
571 DCR *new_dcr(JCR *jcr, DCR *dcr, DEVICE *dev)
574 dcr = (DCR *)malloc(sizeof(DCR));
575 memset(dcr, 0, sizeof(DCR));
576 dcr->tid = pthread_self();
579 dcr->jcr = jcr; /* point back to jcr */
580 /* Set device information, possibly change device */
583 free_block(dcr->block);
585 dcr->block = new_block(dev);
587 free_record(dcr->rec);
589 dcr->rec = new_record();
590 if (dcr->attached_to_dev) {
591 detach_dcr_from_dev(dcr);
593 /* Use job spoolsize prior to device spoolsize */
594 if (jcr->spool_size) {
595 dcr->max_job_spool_size = jcr->spool_size;
597 dcr->max_job_spool_size = dev->device->max_job_spool_size;
599 dcr->device = dev->device;
601 attach_dcr_to_dev(dcr);
607 * Search the dcrs list for the given dcr. If it is found,
608 * as it should be, then remove it. Also zap the jcr pointer
609 * to the dcr if it is the same one.
611 * Note, this code will be turned on when we can write to multiple
612 * dcrs at the same time.
615 static void remove_dcr_from_dcrs(DCR *dcr)
621 int num = jcr->dcrs->size();
622 for (i=0; i < num; i++) {
623 ldcr = (DCR *)jcr->dcrs->get(i);
625 jcr->dcrs->remove(i);
626 if (jcr->dcr == dcr) {
635 static void attach_dcr_to_dev(DCR *dcr)
637 DEVICE *dev = dcr->dev;
640 if (jcr) Dmsg1(500, "JobId=%u enter attach_dcr_to_dev\n", (uint32_t)jcr->JobId);
641 if (!dcr->attached_to_dev && dev->initiated && jcr && jcr->JobType != JT_SYSTEM) {
642 dev->attached_dcrs->append(dcr); /* attach dcr to device */
643 dcr->attached_to_dev = true;
644 Dmsg1(500, "JobId=%u attach_dcr_to_dev\n", (uint32_t)jcr->JobId);
648 void detach_dcr_from_dev(DCR *dcr)
650 DEVICE *dev = dcr->dev;
651 Dmsg0(500, "Enter detach_dcr_from_dev\n"); /* jcr is NULL in some cases */
653 /* Detach this dcr only if attached */
654 if (dcr->attached_to_dev && dev) {
656 dcr->unreserve_device();
657 dcr->dev->attached_dcrs->remove(dcr); /* detach dcr from device */
658 dcr->attached_to_dev = false;
659 // remove_dcr_from_dcrs(dcr); /* remove dcr from jcr list */
665 * Free up all aspects of the given dcr -- i.e. dechain it,
666 * release allocated memory, zap pointers, ...
668 void free_dcr(DCR *dcr)
672 detach_dcr_from_dev(dcr);
675 free_block(dcr->block);
678 free_record(dcr->rec);
680 if (jcr && jcr->dcr == dcr) {