/*
* (Un)mount the device (for tape devices)
*/
static bool do_mount(DCR *dcr, int mount, int dotimeout)
{
DEVRES *device = dcr->dev->device;
POOL_MEM ocmd(PM_FNAME);
POOLMEM *results;
char *icmd;
int status, tries;
berrno be;
Dsm_check(200);
if (mount) {
icmd = device->mount_command;
} else {
icmd = device->unmount_command;
}
dcr->dev->edit_mount_codes(ocmd, icmd);
Dmsg2(100, "do_mount: cmd=%s mounted=%d\n", ocmd.c_str(), dcr->dev->is_mounted());
if (dotimeout) {
/* Try at most 10 times to (un)mount the device. This should perhaps be configurable. */
tries = 10;
} else {
tries = 1;
}
results = get_memory(4000);
/* If busy retry each second */
Dmsg1(100, "do_mount run_prog=%s\n", ocmd.c_str());
while ((status = run_program_full_output(ocmd.c_str(), dcr->dev->max_open_wait / 2, results)) != 0) {
if (tries-- > 0) {
continue;
}
Dmsg5(100, "Device %s cannot be %smounted. stat=%d result=%s ERR=%s\n", dcr->dev->print_name(),
(mount ? "" : "un"), status, results, be.bstrerror(status));
Mmsg(dcr->dev->errmsg, _("Device %s cannot be %smounted. ERR=%s\n"),
dcr->dev->print_name(), (mount ? "" : "un"), be.bstrerror(status));
free_pool_memory(results);
Dmsg0(200, "============ mount=0\n");
Dsm_check(200);
return false;
}
free_pool_memory(results);
Dmsg1(200, "============ mount=%d\n", mount);
return true;
}
bool run_cmd(JCR *jcr)
{
struct timeval tv;
struct timezone tz;
struct timespec timeout;
int errstat = 0;
Dsm_check(200);
Dmsg1(200, "Run_cmd: %s\n", jcr->dir_bsock->msg);
/* If we do not need the FD, we are doing a migrate, copy, or virtual
* backup.
*/
if (jcr->no_client_used()) {
do_mac(jcr);
return false;
}
jcr->sendJobStatus(JS_WaitFD); /* wait for FD to connect */
gettimeofday(&tv, &tz);
timeout.tv_nsec = tv.tv_usec * 1000;
timeout.tv_sec = tv.tv_sec + me->client_wait;
Dmsg3(50, "%s waiting %d sec for FD to contact SD key=%s\n",
jcr->Job, (int)(timeout.tv_sec-time(NULL)), jcr->sd_auth_key);
Dmsg2(800, "Wait FD for jid=%d %p\n", jcr->JobId, jcr);
/*
* Wait for the File daemon to contact us to start the Job,
* when he does, we will be released, unless the 30 minutes
* expires.
*/
P(mutex);
while ( !jcr->authenticated && !job_canceled(jcr) ) {
errstat = pthread_cond_timedwait(&jcr->job_start_wait, &mutex, &timeout);
if (errstat == ETIMEDOUT || errstat == EINVAL || errstat == EPERM) {
break;
}
Dmsg1(800, "=== Auth cond errstat=%d\n", errstat);
}
Dmsg3(50, "Auth=%d canceled=%d errstat=%d\n", jcr->authenticated,
job_canceled(jcr), errstat);
V(mutex);
Dmsg2(800, "Auth fail or cancel for jid=%d %p\n", jcr->JobId, jcr);
memset(jcr->sd_auth_key, 0, strlen(jcr->sd_auth_key));
if (jcr->authenticated && !job_canceled(jcr)) {
Dmsg2(800, "Running jid=%d %p\n", jcr->JobId, jcr);
run_job(jcr); /* Run the job */
}
Dmsg2(800, "Done jid=%d %p\n", jcr->JobId, jcr);
return false;
}
void unstrip_path(FF_PKT *ff_pkt)
{
if (!bit_is_set(FO_STRIPPATH, ff_pkt->flags) || ff_pkt->strip_path <= 0) {
return;
}
strcpy(ff_pkt->fname, ff_pkt->fname_save);
if (ff_pkt->type != FT_LNK && ff_pkt->fname != ff_pkt->link) {
Dmsg2(500, "strcpy link=%s link_save=%s\n", ff_pkt->link,
ff_pkt->link_save);
strcpy(ff_pkt->link, ff_pkt->link_save);
Dmsg2(500, "strcpy link=%d link_save=%d\n", strlen(ff_pkt->link),
strlen(ff_pkt->link_save));
Dsm_check(200);
}
}
static void job_monitor_watchdog(watchdog_t *self)
{
JCR *control_jcr, *jcr;
control_jcr = (JCR *)self->data;
Dsm_check(100);
Dmsg1(800, "job_monitor_watchdog %p called\n", self);
foreach_jcr(jcr) {
bool cancel = false;
if (jcr->JobId == 0 || job_canceled(jcr) || jcr->no_maxtime) {
Dmsg2(800, "Skipping JCR=%p Job=%s\n", jcr, jcr->Job);
continue;
}
/* check MaxWaitTime */
if (job_check_maxwaittime(jcr)) {
jcr->setJobStatus(JS_Canceled);
Qmsg(jcr, M_FATAL, 0, _("Max wait time exceeded. Job canceled.\n"));
cancel = true;
/* check MaxRunTime */
} else if (job_check_maxruntime(jcr)) {
jcr->setJobStatus(JS_Canceled);
Qmsg(jcr, M_FATAL, 0, _("Max run time exceeded. Job canceled.\n"));
cancel = true;
/* check MaxRunSchedTime */
} else if (job_check_maxrunschedtime(jcr)) {
jcr->setJobStatus(JS_Canceled);
Qmsg(jcr, M_FATAL, 0, _("Max run sched time exceeded. Job canceled.\n"));
cancel = true;
}
if (cancel) {
Dmsg3(800, "Cancelling JCR %p jobid %d (%s)\n", jcr, jcr->JobId, jcr->Job);
UAContext *ua = new_ua_context(jcr);
ua->jcr = control_jcr;
cancel_job(ua, jcr);
free_ua_context(ua);
Dmsg2(800, "Have cancelled JCR %p Job=%d\n", jcr, jcr->JobId);
}
}
/* Keep reference counts correct */
endeach_jcr(jcr);
}
/**
* If requested strip leading components of the path so that we can
* save file as if it came from a subdirectory. This is most useful
* for dealing with snapshots, by removing the snapshot directory, or
* in handling vendor migrations where files have been restored with
* a vendor product into a subdirectory.
*/
void strip_path(FF_PKT *ff_pkt)
{
if (!bit_is_set(FO_STRIPPATH, ff_pkt->flags) || ff_pkt->strip_path <= 0) {
Dmsg1(200, "No strip for %s\n", ff_pkt->fname);
return;
}
if (!ff_pkt->fname_save) {
ff_pkt->fname_save = get_pool_memory(PM_FNAME);
ff_pkt->link_save = get_pool_memory(PM_FNAME);
}
pm_strcpy(ff_pkt->fname_save, ff_pkt->fname);
if (ff_pkt->type != FT_LNK && ff_pkt->fname != ff_pkt->link) {
pm_strcpy(ff_pkt->link_save, ff_pkt->link);
Dmsg2(500, "strcpy link_save=%d link=%d\n", strlen(ff_pkt->link_save),
strlen(ff_pkt->link));
Dsm_check(200);
}
/**
* Strip path. If it doesn't succeed put it back. If it does, and there
* is a different link string, attempt to strip the link. If it fails,
* back them both back. Do not strip symlinks. I.e. if either stripping
* fails don't strip anything.
*/
if (!do_strip(ff_pkt->strip_path, ff_pkt->fname)) {
unstrip_path(ff_pkt);
goto rtn;
}
/**
* Strip links but not symlinks
*/
if (ff_pkt->type != FT_LNK && ff_pkt->fname != ff_pkt->link) {
if (!do_strip(ff_pkt->strip_path, ff_pkt->link)) {
unstrip_path(ff_pkt);
}
}
rtn:
Dmsg3(100, "fname=%s stripped=%s link=%s\n", ff_pkt->fname_save, ff_pkt->fname, ff_pkt->link);
}
void catalog_request(JCR *jcr, BSOCK *bs)
{
MEDIA_DBR mr, sdmr;
JOBMEDIA_DBR jm;
char Job[MAX_NAME_LENGTH];
char pool_name[MAX_NAME_LENGTH];
int index, ok, label, writing;
POOLMEM *omsg;
POOL_DBR pr;
uint32_t Stripe, Copy;
uint64_t MediaId;
utime_t VolFirstWritten;
utime_t VolLastWritten;
memset(&sdmr, 0, sizeof(sdmr));
memset(&jm, 0, sizeof(jm));
Dsm_check(100);
/*
* Request to find next appendable Volume for this Job
*/
Dmsg1(100, "catreq %s", bs->msg);
if (!jcr->db) {
omsg = get_memory(bs->msglen+1);
pm_strcpy(omsg, bs->msg);
bs->fsend(_("1990 Invalid Catalog Request: %s"), omsg);
Jmsg1(jcr, M_FATAL, 0, _("Invalid Catalog request; DB not open: %s"), omsg);
free_memory(omsg);
return;
}
/*
* Find next appendable medium for SD
*/
if (sscanf(bs->msg, Find_media, &Job, &index, &pool_name, &mr.MediaType) == 4) {
memset(&pr, 0, sizeof(pr));
bstrncpy(pr.Name, pool_name, sizeof(pr.Name));
unbash_spaces(pr.Name);
ok = db_get_pool_record(jcr, jcr->db, &pr);
if (ok) {
mr.PoolId = pr.PoolId;
set_storageid_in_mr(jcr->res.wstore, &mr);
mr.ScratchPoolId = pr.ScratchPoolId;
ok = find_next_volume_for_append(jcr, &mr, index, fnv_create_vol, fnv_prune);
Dmsg3(050, "find_media ok=%d idx=%d vol=%s\n", ok, index, mr.VolumeName);
}
/*
* Send Find Media response to Storage daemon
*/
if (ok) {
send_volume_info_to_storage_daemon(jcr, bs, &mr);
} else {
bs->fsend(_("1901 No Media.\n"));
Dmsg0(500, "1901 No Media.\n");
}
/*
* Request to find specific Volume information
*/
} else if (sscanf(bs->msg, Get_Vol_Info, &Job, &mr.VolumeName, &writing) == 3) {
Dmsg1(100, "CatReq GetVolInfo Vol=%s\n", mr.VolumeName);
/*
* Find the Volume
*/
unbash_spaces(mr.VolumeName);
if (db_get_media_record(jcr, jcr->db, &mr)) {
const char *reason = NULL; /* detailed reason for rejection */
/*
* If we are reading, accept any volume (reason == NULL)
* If we are writing, check if the Volume is valid
* for this job, and do a recycle if necessary
*/
if (writing) {
/*
* SD wants to write this Volume, so make
* sure it is suitable for this job, i.e.
* Pool matches, and it is either Append or Recycle
* and Media Type matches and Pool allows any volume.
*/
if (mr.PoolId != jcr->jr.PoolId) {
reason = _("not in Pool");
} else if (!bstrcmp(mr.MediaType, jcr->res.wstore->media_type)) {
reason = _("not correct MediaType");
} else {
/*
* Now try recycling if necessary
* reason set non-NULL if we cannot use it
*/
check_if_volume_valid_or_recyclable(jcr, &mr, &reason);
}
}
if (!reason && mr.Enabled != 1) {
reason = _("is not Enabled");
}
if (reason == NULL) {
/*
* Send Find Media response to Storage daemon
*/
send_volume_info_to_storage_daemon(jcr, bs, &mr);
} else {
/* Not suitable volume */
bs->fsend(_("1998 Volume \"%s\" catalog status is %s, %s.\n"), mr.VolumeName,
mr.VolStatus, reason);
}
} else {
bs->fsend(_("1997 Volume \"%s\" not in catalog.\n"), mr.VolumeName);
Dmsg1(100, "1997 Volume \"%s\" not in catalog.\n", mr.VolumeName);
}
/*
* Request to update Media record. Comes typically at the end
* of a Storage daemon Job Session, when labeling/relabeling a
* Volume, or when an EOF mark is written.
*/
} else if (sscanf(bs->msg, Update_media, &Job, &sdmr.VolumeName,
&sdmr.VolJobs, &sdmr.VolFiles, &sdmr.VolBlocks, &sdmr.VolBytes,
&sdmr.VolMounts, &sdmr.VolErrors, &sdmr.VolWrites, &sdmr.MaxVolBytes,
&VolLastWritten, &sdmr.VolStatus, &sdmr.Slot, &label, &sdmr.InChanger,
&sdmr.VolReadTime, &sdmr.VolWriteTime, &VolFirstWritten) == 18) {
db_lock(jcr->db);
Dmsg3(400, "Update media %s oldStat=%s newStat=%s\n", sdmr.VolumeName,
mr.VolStatus, sdmr.VolStatus);
bstrncpy(mr.VolumeName, sdmr.VolumeName, sizeof(mr.VolumeName)); /* copy Volume name */
unbash_spaces(mr.VolumeName);
if (!db_get_media_record(jcr, jcr->db, &mr)) {
Jmsg(jcr, M_ERROR, 0, _("Unable to get Media record for Volume %s: ERR=%s\n"),
mr.VolumeName, db_strerror(jcr->db));
bs->fsend(_("1991 Catalog Request for vol=%s failed: %s"),
mr.VolumeName, db_strerror(jcr->db));
goto bail_out;
}
/* Set first written time if this is first job */
if (mr.FirstWritten == 0) {
if (VolFirstWritten == 0) {
mr.FirstWritten = jcr->start_time; /* use Job start time as first write */
} else {
mr.FirstWritten = VolFirstWritten;
}
mr.set_first_written = true;
}
/* If we just labeled the tape set time */
if (label || mr.LabelDate == 0) {
mr.LabelDate = jcr->start_time;
mr.set_label_date = true;
if (mr.InitialWrite == 0) {
mr.InitialWrite = jcr->start_time;
}
Dmsg2(400, "label=%d labeldate=%d\n", label, mr.LabelDate);
} else {
/*
* Insanity check for VolFiles get set to a smaller value
*/
if (sdmr.VolFiles < mr.VolFiles) {
Jmsg(jcr, M_FATAL, 0, _("Volume Files at %u being set to %u"
" for Volume \"%s\". This is incorrect.\n"),
mr.VolFiles, sdmr.VolFiles, mr.VolumeName);
bs->fsend(_("1992 Update Media error. VolFiles=%u, CatFiles=%u\n"),
sdmr.VolFiles, mr.VolFiles);
goto bail_out;
}
}
Dmsg2(400, "Update media: BefVolJobs=%u After=%u\n", mr.VolJobs, sdmr.VolJobs);
/*
* Check if the volume has been written by the job,
* and update the LastWritten field if needed.
*/
if (mr.VolBlocks != sdmr.VolBlocks && VolLastWritten != 0) {
mr.LastWritten = VolLastWritten;
}
/*
* Update to point to the last device used to write the Volume.
* However, do so only if we are writing the tape, i.e.
* the number of VolWrites has increased.
*/
if (jcr->res.wstore && sdmr.VolWrites > mr.VolWrites) {
Dmsg2(050, "Update StorageId old=%d new=%d\n",
mr.StorageId, jcr->res.wstore->StorageId);
/* Update StorageId after write */
set_storageid_in_mr(jcr->res.wstore, &mr);
} else {
/* Nothing written, reset same StorageId */
set_storageid_in_mr(NULL, &mr);
}
/* Copy updated values to original media record */
mr.VolJobs = sdmr.VolJobs;
mr.VolFiles = sdmr.VolFiles;
mr.VolBlocks = sdmr.VolBlocks;
mr.VolBytes = sdmr.VolBytes;
mr.VolMounts = sdmr.VolMounts;
mr.VolErrors = sdmr.VolErrors;
mr.VolWrites = sdmr.VolWrites;
mr.Slot = sdmr.Slot;
mr.InChanger = sdmr.InChanger;
bstrncpy(mr.VolStatus, sdmr.VolStatus, sizeof(mr.VolStatus));
mr.VolReadTime = sdmr.VolReadTime;
mr.VolWriteTime = sdmr.VolWriteTime;
Dmsg2(400, "db_update_media_record. Stat=%s Vol=%s\n", mr.VolStatus, mr.VolumeName);
/*
* Update the database, then before sending the response to the
* SD, check if the Volume has expired.
*/
if (!db_update_media_record(jcr, jcr->db, &mr)) {
Jmsg(jcr, M_FATAL, 0, _("Catalog error updating Media record. %s"),
db_strerror(jcr->db));
bs->fsend(_("1993 Update Media error\n"));
Dmsg0(400, "send error\n");
} else {
(void)has_volume_expired(jcr, &mr);
send_volume_info_to_storage_daemon(jcr, bs, &mr);
}
bail_out:
db_unlock(jcr->db);
Dmsg1(400, ">CatReq response: %s", bs->msg);
Dmsg1(400, "Leave catreq jcr 0x%x\n", jcr);
return;
/*
* Request to create a JobMedia record
*/
} else if (sscanf(bs->msg, Create_job_media, &Job,
&jm.FirstIndex, &jm.LastIndex, &jm.StartFile, &jm.EndFile,
&jm.StartBlock, &jm.EndBlock, &Copy, &Stripe, &MediaId) == 10) {
if (jcr->mig_jcr) {
jm.JobId = jcr->mig_jcr->JobId;
} else {
jm.JobId = jcr->JobId;
}
jm.MediaId = MediaId;
Dmsg6(400, "create_jobmedia JobId=%d MediaId=%d SF=%d EF=%d FI=%d LI=%d\n",
jm.JobId, jm.MediaId, jm.StartFile, jm.EndFile, jm.FirstIndex, jm.LastIndex);
if (!db_create_jobmedia_record(jcr, jcr->db, &jm)) {
Jmsg(jcr, M_FATAL, 0, _("Catalog error creating JobMedia record. %s"),
db_strerror(jcr->db));
bs->fsend(_("1992 Create JobMedia error\n"));
} else {
Dmsg0(400, "JobMedia record created\n");
bs->fsend(OK_create);
}
} else {
omsg = get_memory(bs->msglen+1);
pm_strcpy(omsg, bs->msg);
bs->fsend(_("1990 Invalid Catalog Request: %s"), omsg);
Jmsg1(jcr, M_FATAL, 0, _("Invalid Catalog request: %s"), omsg);
free_memory(omsg);
}
Dmsg1(400, ">CatReq response: %s", bs->msg);
Dmsg1(400, "Leave catreq jcr 0x%x\n", jcr);
return;
}
bool DEVICE::scan_dir_for_volume(DCR *dcr)
{
DIR* dp;
struct dirent *entry, *result;
int name_max;
char *mount_point;
VOLUME_CAT_INFO dcrVolCatInfo, devVolCatInfo;
char VolumeName[MAX_NAME_LENGTH];
struct stat statp;
bool found = false;
POOL_MEM fname(PM_FNAME);
bool need_slash = false;
int len;
dcrVolCatInfo = dcr->VolCatInfo; /* structure assignment */
devVolCatInfo = VolCatInfo; /* structure assignment */
bstrncpy(VolumeName, dcr->VolumeName, sizeof(VolumeName));
name_max = pathconf(".", _PC_NAME_MAX);
if (name_max < 1024) {
name_max = 1024;
}
if (device->mount_point) {
mount_point = device->mount_point;
} else {
mount_point = device->device_name;
}
if (!(dp = opendir(mount_point))) {
berrno be;
dev_errno = errno;
Dmsg3(29, "scan_dir_for_vol: failed to open dir %s (dev=%s), ERR=%s\n",
mount_point, print_name(), be.bstrerror());
goto get_out;
}
len = strlen(mount_point);
if (len > 0) {
need_slash = !IsPathSeparator(mount_point[len - 1]);
}
entry = (struct dirent *)malloc(sizeof(struct dirent) + name_max + 1000);
for ( ;; ) {
if ((readdir_r(dp, entry, &result) != 0) || (result == NULL)) {
dev_errno = EIO;
Dmsg2(129, "scan_dir_for_vol: failed to find suitable file in dir %s (dev=%s)\n",
mount_point, print_name());
break;
}
if (strcmp(result->d_name, ".") == 0 ||
strcmp(result->d_name, "..") == 0) {
continue;
}
if (!is_volume_name_legal(result->d_name)) {
continue;
}
pm_strcpy(fname, mount_point);
if (need_slash) {
pm_strcat(fname, "/");
}
pm_strcat(fname, result->d_name);
if (lstat(fname.c_str(), &statp) != 0 ||
!S_ISREG(statp.st_mode)) {
continue; /* ignore directories & special files */
}
/*
* OK, we got a different volume mounted. First save the
* requested Volume info (dcr) structure, then query if
* this volume is really OK. If not, put back the desired
* volume name, mark it not in changer and continue.
*/
/* Check if this is a valid Volume in the pool */
bstrncpy(dcr->VolumeName, result->d_name, sizeof(dcr->VolumeName));
if (!dir_get_volume_info(dcr, GET_VOL_INFO_FOR_WRITE)) {
continue;
}
/* This was not the volume we expected, but it is OK with
* the Director, so use it.
*/
VolCatInfo = dcr->VolCatInfo; /* structure assignment */
found = true;
break; /* got a Volume */
}
free(entry);
closedir(dp);
get_out:
if (!found) {
/* Restore VolumeName we really wanted */
bstrncpy(dcr->VolumeName, VolumeName, sizeof(dcr->VolumeName));
dcr->VolCatInfo = dcrVolCatInfo; /* structure assignment */
VolCatInfo = devVolCatInfo; /* structure assignment */
}
Dsm_check(100);
return found;
}
/* Update the free space on the device */
bool DEVICE::update_freespace()
{
POOL_MEM ocmd(PM_FNAME);
POOLMEM* results;
char* icmd;
int timeout;
uint64_t free;
char ed1[50];
bool ok = false;
int status;
if (!is_dvd() || is_freespace_ok()) {
return true;
}
/* The device must be mounted in order to dvd-freespace to work */
mount(1);
Dsm_check(400);
icmd = device->free_space_command;
if (!icmd) {
free_space = 0;
free_space_errno = 0;
clear_freespace_ok(); /* No valid freespace */
clear_media();
Dmsg2(29, "ERROR: update_free_space_dev: free_space=%s, free_space_errno=%d (!icmd)\n",
edit_uint64(free_space, ed1), free_space_errno);
Mmsg(errmsg, _("No FreeSpace command defined.\n"));
return false;
}
edit_mount_codes(ocmd, icmd);
Dmsg1(29, "update_freespace: cmd=%s\n", ocmd.c_str());
results = get_pool_memory(PM_MESSAGE);
/* Try at most 3 times to get the free space on the device. This should perhaps be configurable. */
timeout = 3;
while (1) {
berrno be;
Dmsg1(20, "Run freespace prog=%s\n", ocmd.c_str());
status = run_program_full_output(ocmd.c_str(), max_open_wait/2, results);
Dmsg2(500, "Freespace status=%d result=%s\n", status, results);
if (status == 0) {
free = str_to_int64(results);
Dmsg1(400, "Free space program run: Freespace=%s\n", results);
if (free >= 0) {
free_space = free;
free_space_errno = 0;
set_freespace_ok(); /* have valid freespace */
set_media();
Mmsg(errmsg, "");
ok = true;
break;
}
}
free_space = 0;
free_space_errno = EPIPE;
clear_freespace_ok(); /* no valid freespace */
Mmsg2(errmsg, _("Cannot run free space command. Results=%s ERR=%s\n"),
results, be.bstrerror(status));
if (--timeout > 0) {
Dmsg4(40, "Cannot get free space on device %s. free_space=%s, "
"free_space_errno=%d ERR=%s\n", print_name(),
edit_uint64(free_space, ed1), free_space_errno,
errmsg);
bmicrosleep(1, 0);
continue;
}
dev_errno = free_space_errno;
Dmsg4(40, "Cannot get free space on device %s. free_space=%s, "
"free_space_errno=%d ERR=%s\n",
print_name(), edit_uint64(free_space, ed1),
free_space_errno, errmsg);
break;
}
free_pool_memory(results);
Dmsg4(29, "leave update_freespace: free_space=%s freespace_ok=%d free_space_errno=%d have_media=%d\n",
edit_uint64(free_space, ed1), !!is_freespace_ok(), free_space_errno, !!have_media());
Dsm_check(400);
return ok;
}
/*
* Note!!!! Part numbers now begin at 1. The part number is
* suppressed from the first part, which is just the Volume
* name. Each subsequent part is the Volumename.partnumber.
*
* Write a part (Vol, Vol.2, ...) from the spool to the DVD
* This routine does not update the part number, so normally, you
* should call open_next_part()
*
* It is also called from truncate_dvd to "blank" the medium, as
* well as from block.c when the DVD is full to write the last part.
*/
bool dvd_write_part(DCR *dcr)
{
DEVICE *dev = dcr->dev;
POOL_MEM archive_name(PM_FNAME);
/*
* Don't write empty part files.
* This is only useful when growisofs does not support write beyond
* the 4GB boundary.
* Example :
* - 3.9 GB on the volume, dvd-freespace reports 0.4 GB free
* - Write 0.2 GB on the volume, Bacula thinks it could still
* append data, it creates a new empty part.
* - dvd-freespace reports 0 GB free, as the 4GB boundary has
* been crossed
* - Bacula thinks he must finish to write to the device, so it
* tries to write the last part (0-byte), but dvd-writepart fails...
*
* There is one exception: when recycling a volume, we write a blank part
* file, so, then, we need to accept to write it.
*/
if (dev->part_size == 0 && !dev->truncating) {
Dmsg2(29, "dvd_write_part: device is %s, won't write blank part %d\n", dev->print_name(), dev->part);
/* Delete spool file */
make_spooled_dvd_filename(dev, archive_name);
unlink(archive_name.c_str());
dev->set_part_spooled(false);
Dmsg1(29, "========= unlink(%s)\n", archive_name.c_str());
Dsm_check(400);
return true;
}
POOL_MEM ocmd(PM_FNAME);
POOL_MEM results(PM_MESSAGE);
char* icmd;
int status;
int timeout;
char ed1[50];
dev->clear_freespace_ok(); /* need to update freespace */
Dsm_check(400);
Dmsg3(29, "dvd_write_part: device is %s, part is %d, is_mounted=%d\n", dev->print_name(), dev->part, dev->is_mounted());
icmd = dev->device->write_part_command;
dev->edit_mount_codes(ocmd, icmd);
/*
* original line follows
* timeout = dev->max_open_wait + (dev->max_part_size/(1350*1024/2));
* I modified this for a longer timeout; pre-formatting, blanking and
* writing can take quite a while
*/
/* Explanation of the timeout value, when writing the first part,
* by Arno Lehmann :
* 9 GB, write speed 1x: 6990 seconds (almost 2 hours...)
* Overhead: 900 seconds (starting, initializing, finalizing,probably
* reloading 15 minutes)
* Sum: 15780.
* A reasonable last-exit timeout would be 16000 seconds. Quite long -
* almost 4.5 hours, but hopefully, that timeout will only ever be needed
* in case of a serious emergency.
*/
if (dev->part == 1) {
timeout = 16000;
} else {
timeout = dev->max_open_wait + (dev->part_size/(1350*1024/4));
}
Dmsg2(20, "Write part: cmd=%s timeout=%d\n", ocmd.c_str(), timeout);
status = run_program_full_output(ocmd.c_str(), timeout, results.addr());
Dmsg2(20, "Write part status=%d result=%s\n", status, results.c_str());
dev->blank_dvd = false;
if (status != 0) {
Jmsg2(dcr->jcr, M_FATAL, 0, _("Error writing part %d to the DVD: ERR=%s\n"),
dev->part, results.c_str());
Mmsg1(dev->errmsg, _("Error while writing current part to the DVD: %s"),
results.c_str());
Dmsg1(100, "%s\n", dev->errmsg);
dev->dev_errno = EIO;
if (!dev->truncating) {
dcr->mark_volume_in_error();
}
Dsm_check(400);
return false;
}
Jmsg(dcr->jcr, M_INFO, 0, _("Part %d (%lld bytes) written to DVD.\n"), dev->part, dev->part_size);
Dmsg3(400, "dvd_write_part: Part %d (%lld bytes) written to DVD\nResults: %s\n",
dev->part, dev->part_size, results.c_str());
dev->num_dvd_parts++; /* there is now one more part on DVD */
dev->VolCatInfo.VolCatParts = dev->num_dvd_parts;
dcr->VolCatInfo.VolCatParts = dev->num_dvd_parts;
Dmsg1(100, "Update num_parts=%d\n", dev->num_dvd_parts);
/* Delete spool file */
make_spooled_dvd_filename(dev, archive_name);
unlink(archive_name.c_str());
dev->set_part_spooled(false);
Dmsg1(29, "========= unlink(%s)\n", archive_name.c_str());
Dsm_check(400);
/* growisofs umounted the device, so remount it (it will update the free space) */
dev->clear_mounted();
dev->mount(1);
Jmsg(dcr->jcr, M_INFO, 0, _("Remaining free space %s on %s\n"),
edit_uint64_with_commas(dev->free_space, ed1), dev->print_name());
Dsm_check(400);
return true;
}
/*
* (Un)mount the device (For a FILE device)
*/
static bool do_mount(DCR *dcr, bool mount, int dotimeout)
{
DEVRES *device = dcr->dev->device;
POOL_MEM ocmd(PM_FNAME);
POOLMEM *results;
DIR* dp;
char *icmd;
struct dirent *entry, *result;
int status, tries, name_max, count;
berrno be;
Dsm_check(200);
if (mount) {
icmd = device->mount_command;
} else {
icmd = device->unmount_command;
}
dcr->dev->edit_mount_codes(ocmd, icmd);
Dmsg2(100, "do_mount: cmd=%s mounted=%d\n", ocmd.c_str(), dcr->dev->is_mounted());
if (dotimeout) {
/* Try at most 10 times to (un)mount the device. This should perhaps be configurable. */
tries = 10;
} else {
tries = 1;
}
results = get_memory(4000);
/* If busy retry each second */
Dmsg1(100, "do_mount run_prog=%s\n", ocmd.c_str());
while ((status = run_program_full_output(ocmd.c_str(), dcr->dev->max_open_wait / 2, results)) != 0) {
/* Doesn't work with internationalization (This is not a problem) */
if (mount && fnmatch("*is already mounted on*", results, 0) == 0) {
break;
}
if (!mount && fnmatch("* not mounted*", results, 0) == 0) {
break;
}
if (tries-- > 0) {
/* Sometimes the device cannot be mounted because it is already mounted.
* Try to unmount it, then remount it */
if (mount) {
Dmsg1(400, "Trying to unmount the device %s...\n", dcr->dev->print_name());
do_mount(dcr, 0, 0);
}
bmicrosleep(1, 0);
continue;
}
Dmsg5(100, "Device %s cannot be %smounted. status=%d result=%s ERR=%s\n", dcr->dev->print_name(),
(mount ? "" : "un"), status, results, be.bstrerror(status));
Mmsg(dcr->dev->errmsg, _("Device %s cannot be %smounted. ERR=%s\n"),
dcr->dev->print_name(), (mount ? "" : "un"), be.bstrerror(status));
/*
* Now, just to be sure it is not mounted, try to read the filesystem.
*/
name_max = pathconf(".", _PC_NAME_MAX);
if (name_max < 1024) {
name_max = 1024;
}
if (!(dp = opendir(device->mount_point))) {
berrno be;
dcr->dev->dev_errno = errno;
Dmsg3(100, "do_mount: failed to open dir %s (dev=%s), ERR=%s\n",
device->mount_point, dcr->dev->print_name(), be.bstrerror());
goto get_out;
}
entry = (struct dirent *)malloc(sizeof(struct dirent) + name_max + 1000);
count = 0;
while (1) {
if ((readdir_r(dp, entry, &result) != 0) || (result == NULL)) {
dcr->dev->dev_errno = EIO;
Dmsg2(129, "do_mount: failed to find suitable file in dir %s (dev=%s)\n",
device->mount_point, dcr->dev->print_name());
break;
}
if (!bstrcmp(result->d_name, ".") && !bstrcmp(result->d_name, "..") && !bstrcmp(result->d_name, ".keep")) {
count++; /* result->d_name != ., .. or .keep (Gentoo-specific) */
break;
} else {
Dmsg2(129, "do_mount: ignoring %s in %s\n", result->d_name, device->mount_point);
}
}
free(entry);
closedir(dp);
Dmsg1(100, "do_mount: got %d files in the mount point (not counting ., .. and .keep)\n", count);
if (count > 0) {
/* If we got more than ., .. and .keep */
/* there must be something mounted */
if (mount) {
Dmsg1(100, "Did Mount by count=%d\n", count);
break;
} else {
/* An unmount request. We failed to unmount - report an error */
free_pool_memory(results);
Dmsg0(200, "== error mount=1 wanted unmount\n");
return false;
}
}
get_out:
free_pool_memory(results);
Dmsg0(200, "============ mount=0\n");
Dsm_check(200);
return false;
}
free_pool_memory(results);
Dmsg1(200, "============ mount=%d\n", mount);
return true;
}
bool setup_job(JCR *jcr)
{
int errstat;
jcr->lock();
Dsm_check(100);
init_msg(jcr, jcr->res.messages, job_code_callback_director);
/*
* Initialize termination condition variable
*/
if ((errstat = pthread_cond_init(&jcr->term_wait, NULL)) != 0) {
berrno be;
Jmsg1(jcr, M_FATAL, 0, _("Unable to init job cond variable: ERR=%s\n"), be.bstrerror(errstat));
jcr->unlock();
goto bail_out;
}
jcr->term_wait_inited = true;
/*
* Initialize nextrun ready condition variable
*/
if ((errstat = pthread_cond_init(&jcr->nextrun_ready, NULL)) != 0) {
berrno be;
Jmsg1(jcr, M_FATAL, 0, _("Unable to init job nextrun cond variable: ERR=%s\n"), be.bstrerror(errstat));
jcr->unlock();
goto bail_out;
}
jcr->nextrun_ready_inited = true;
create_unique_job_name(jcr, jcr->res.job->name());
jcr->setJobStatus(JS_Created);
jcr->unlock();
/*
* Open database
*/
Dmsg0(100, "Open database\n");
jcr->db = db_sql_get_pooled_connection(jcr,
jcr->res.catalog->db_driver,
jcr->res.catalog->db_name,
jcr->res.catalog->db_user,
jcr->res.catalog->db_password,
jcr->res.catalog->db_address,
jcr->res.catalog->db_port,
jcr->res.catalog->db_socket,
jcr->res.catalog->mult_db_connections,
jcr->res.catalog->disable_batch_insert);
if (jcr->db == NULL) {
Jmsg(jcr, M_FATAL, 0, _("Could not open database \"%s\".\n"),
jcr->res.catalog->db_name);
goto bail_out;
}
Dmsg0(150, "DB opened\n");
if (!jcr->fname) {
jcr->fname = get_pool_memory(PM_FNAME);
}
if (!jcr->res.pool_source) {
jcr->res.pool_source = get_pool_memory(PM_MESSAGE);
pm_strcpy(jcr->res.pool_source, _("unknown source"));
}
if (!jcr->res.npool_source) {
jcr->res.npool_source = get_pool_memory(PM_MESSAGE);
pm_strcpy(jcr->res.npool_source, _("unknown source"));
}
if (jcr->JobReads()) {
if (!jcr->res.rpool_source) {
jcr->res.rpool_source = get_pool_memory(PM_MESSAGE);
pm_strcpy(jcr->res.rpool_source, _("unknown source"));
}
}
/*
* Create Job record
*/
init_jcr_job_record(jcr);
if (!get_or_create_client_record(jcr)) {
goto bail_out;
}
if (!db_create_job_record(jcr, jcr->db, &jcr->jr)) {
Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
goto bail_out;
}
jcr->JobId = jcr->jr.JobId;
Dmsg4(100, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
jcr->JobId, jcr->Job, jcr->jr.JobType, jcr->jr.JobLevel);
new_plugins(jcr); /* instantiate plugins for this jcr */
generate_plugin_event(jcr, bDirEventJobStart);
if (job_canceled(jcr)) {
goto bail_out;
}
if (jcr->JobReads() && !jcr->rstorage) {
if (jcr->res.job->storage) {
copy_rwstorage(jcr, jcr->res.job->storage, _("Job resource"));
} else {
copy_rwstorage(jcr, jcr->res.job->pool->storage, _("Pool resource"));
}
}
if (!jcr->JobReads()) {
free_rstorage(jcr);
}
/*
* Now, do pre-run stuff, like setting job level (Inc/diff, ...)
* this allows us to setup a proper job start record for restarting
* in case of later errors.
*/
switch (jcr->getJobType()) {
case JT_BACKUP:
switch (jcr->getJobProtocol()) {
case PT_NDMP:
if (!do_ndmp_backup_init(jcr)) {
ndmp_backup_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
break;
default:
if (jcr->is_JobLevel(L_VIRTUAL_FULL)) {
if (!do_native_vbackup_init(jcr)) {
native_vbackup_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
} else {
if (!do_native_backup_init(jcr)) {
native_backup_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
}
break;
}
break;
case JT_VERIFY:
if (!do_verify_init(jcr)) {
verify_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
break;
case JT_RESTORE:
switch (jcr->getJobProtocol()) {
case PT_NDMP:
if (!do_ndmp_restore_init(jcr)) {
ndmp_restore_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
break;
default:
/*
* Any non NDMP restore is not interested at the items
* that were selected for restore so drop them now.
*/
if (jcr->restore_tree_root) {
free_tree(jcr->restore_tree_root);
jcr->restore_tree_root = NULL;
}
if (!do_native_restore_init(jcr)) {
native_restore_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
break;
}
break;
case JT_ADMIN:
if (!do_admin_init(jcr)) {
admin_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
break;
case JT_COPY:
case JT_MIGRATE:
if (!do_migration_init(jcr)) {
migration_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
break;
default:
Pmsg1(0, _("Unimplemented job type: %d\n"), jcr->getJobType());
jcr->setJobStatus(JS_ErrorTerminated);
goto bail_out;
}
generate_plugin_event(jcr, bDirEventJobInit);
Dsm_check(100);
return true;
bail_out:
return false;
}
/*
* This is the engine called by jobq.c:jobq_add() when we were pulled
* from the work queue.
* At this point, we are running in our own thread and all
* necessary resources are allocated -- see jobq.c
*/
static void *job_thread(void *arg)
{
JCR *jcr = (JCR *)arg;
pthread_detach(pthread_self());
Dsm_check(100);
Dmsg0(200, "=====Start Job=========\n");
jcr->setJobStatus(JS_Running); /* this will be set only if no error */
jcr->start_time = time(NULL); /* set the real start time */
jcr->jr.StartTime = jcr->start_time;
if (jcr->res.job->MaxStartDelay != 0 && jcr->res.job->MaxStartDelay <
(utime_t)(jcr->start_time - jcr->sched_time)) {
jcr->setJobStatus(JS_Canceled);
Jmsg(jcr, M_FATAL, 0, _("Job canceled because max start delay time exceeded.\n"));
}
if (job_check_maxrunschedtime(jcr)) {
jcr->setJobStatus(JS_Canceled);
Jmsg(jcr, M_FATAL, 0, _("Job canceled because max run sched time exceeded.\n"));
}
/* TODO : check if it is used somewhere */
if (jcr->res.job->RunScripts == NULL) {
Dmsg0(200, "Warning, job->RunScripts is empty\n");
jcr->res.job->RunScripts = New(alist(10, not_owned_by_alist));
}
if (!db_update_job_start_record(jcr, jcr->db, &jcr->jr)) {
Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
}
/* Run any script BeforeJob on dird */
run_scripts(jcr, jcr->res.job->RunScripts, "BeforeJob");
/*
* We re-update the job start record so that the start
* time is set after the run before job. This avoids
* that any files created by the run before job will
* be saved twice. They will be backed up in the current
* job, but not in the next one unless they are changed.
* Without this, they will be backed up in this job and
* in the next job run because in that case, their date
* is after the start of this run.
*/
jcr->start_time = time(NULL);
jcr->jr.StartTime = jcr->start_time;
if (!db_update_job_start_record(jcr, jcr->db, &jcr->jr)) {
Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
}
generate_plugin_event(jcr, bDirEventJobRun);
switch (jcr->getJobType()) {
case JT_BACKUP:
switch (jcr->getJobProtocol()) {
case PT_NDMP:
if (!job_canceled(jcr)) {
if (do_ndmp_backup(jcr)) {
do_autoprune(jcr);
} else {
ndmp_backup_cleanup(jcr, JS_ErrorTerminated);
}
} else {
ndmp_backup_cleanup(jcr, JS_Canceled);
}
break;
default:
if (!job_canceled(jcr)) {
if (jcr->is_JobLevel(L_VIRTUAL_FULL)) {
if (do_native_vbackup(jcr)) {
do_autoprune(jcr);
} else {
native_vbackup_cleanup(jcr, JS_ErrorTerminated);
}
} else {
if (do_native_backup(jcr)) {
do_autoprune(jcr);
} else {
native_backup_cleanup(jcr, JS_ErrorTerminated);
}
}
} else {
if (jcr->is_JobLevel(L_VIRTUAL_FULL)) {
native_vbackup_cleanup(jcr, JS_Canceled);
} else {
native_backup_cleanup(jcr, JS_Canceled);
}
}
break;
}
break;
case JT_VERIFY:
if (!job_canceled(jcr)) {
if (do_verify(jcr)) {
do_autoprune(jcr);
} else {
verify_cleanup(jcr, JS_ErrorTerminated);
}
} else {
verify_cleanup(jcr, JS_Canceled);
}
break;
case JT_RESTORE:
switch (jcr->getJobProtocol()) {
case PT_NDMP:
if (!job_canceled(jcr)) {
if (do_ndmp_restore(jcr)) {
do_autoprune(jcr);
} else {
ndmp_restore_cleanup(jcr, JS_ErrorTerminated);
}
} else {
ndmp_restore_cleanup(jcr, JS_Canceled);
}
break;
default:
if (!job_canceled(jcr)) {
if (do_native_restore(jcr)) {
do_autoprune(jcr);
} else {
native_restore_cleanup(jcr, JS_ErrorTerminated);
}
} else {
native_restore_cleanup(jcr, JS_Canceled);
}
break;
}
break;
case JT_ADMIN:
if (!job_canceled(jcr)) {
if (do_admin(jcr)) {
do_autoprune(jcr);
} else {
admin_cleanup(jcr, JS_ErrorTerminated);
}
} else {
admin_cleanup(jcr, JS_Canceled);
}
break;
case JT_COPY:
case JT_MIGRATE:
if (!job_canceled(jcr)) {
if (do_migration(jcr)) {
do_autoprune(jcr);
} else {
migration_cleanup(jcr, JS_ErrorTerminated);
}
} else {
migration_cleanup(jcr, JS_Canceled);
}
break;
default:
Pmsg1(0, _("Unimplemented job type: %d\n"), jcr->getJobType());
break;
}
run_scripts(jcr, jcr->res.job->RunScripts, "AfterJob");
/* Send off any queued messages */
if (jcr->msg_queue && jcr->msg_queue->size() > 0) {
dequeue_messages(jcr);
}
generate_plugin_event(jcr, bDirEventJobEnd);
Dmsg1(50, "======== End Job stat=%c ==========\n", jcr->JobStatus);
Dsm_check(100);
return NULL;
}
/*
* Run an external program. Optionally wait a specified number
* of seconds. Program killed if wait exceeded (it is done by the
* watchdog, as fgets is a blocking function).
*
* If the watchdog kills the program, fgets returns, and ferror is set
* to 1 (=>SUCCESS), so we check if the watchdog killed the program.
*
* Return the full output from the program (not only the first line).
*
* Contrary to my normal calling conventions, this program
*
* Returns: 0 on success
* non-zero on error == berrno status
*
*/
int run_program_full_output(char *prog, int wait, POOLMEM *&results)
{
BPIPE *bpipe;
int stat1, stat2;
char *mode;
POOLMEM* tmp;
char *buf;
const int bufsize = 32000;
Dsm_check(200);
tmp = get_pool_memory(PM_MESSAGE);
buf = (char *)malloc(bufsize+1);
results[0] = 0;
mode = (char *)"r";
bpipe = open_bpipe(prog, wait, mode);
if (!bpipe) {
stat1 = ENOENT;
goto bail_out;
}
Dsm_check(200);
tmp[0] = 0;
while (1) {
buf[0] = 0;
fgets(buf, bufsize, bpipe->rfd);
buf[bufsize] = 0;
pm_strcat(tmp, buf);
if (feof(bpipe->rfd)) {
stat1 = 0;
Dmsg1(900, "Run program fgets stat=%d\n", stat1);
break;
} else {
stat1 = ferror(bpipe->rfd);
}
if (stat1 < 0) {
berrno be;
Dmsg2(200, "Run program fgets stat=%d ERR=%s\n", stat1, be.bstrerror());
break;
} else if (stat1 != 0) {
Dmsg1(900, "Run program fgets stat=%d\n", stat1);
if (bpipe->timer_id && bpipe->timer_id->killed) {
Dmsg1(250, "Run program saw fgets killed=%d\n", bpipe->timer_id->killed);
break;
}
}
}
/*
* We always check whether the timer killed the program. We would see
* an eof even when it does so we just have to trust the killed flag
* and set the timer values to avoid edge cases where the program ends
* just as the timer kills it.
*/
if (bpipe->timer_id && bpipe->timer_id->killed) {
Dmsg1(150, "Run program fgets killed=%d\n", bpipe->timer_id->killed);
pm_strcpy(tmp, _("Program killed by Bacula (timeout)\n"));
stat1 = ETIME;
}
pm_strcpy(results, tmp);
Dmsg3(1900, "resadr=0x%x reslen=%d res=%s\n", results, strlen(results), results);
stat2 = close_bpipe(bpipe);
stat1 = stat2 != 0 ? stat2 : stat1;
Dmsg1(900, "Run program returning %d\n", stat1);
bail_out:
free_pool_memory(tmp);
free(buf);
return stat1;
}
bool run_cmd(JCR *jcr)
{
struct timeval tv;
struct timezone tz;
struct timespec timeout;
int errstat = 0;
Dsm_check(200);
Dmsg1(200, "Run_cmd: %s\n", jcr->dir_bsock->msg);
/* If we do not need the FD, we are doing a virtual backup. */
if (jcr->no_client_used()) {
do_vbackup(jcr);
return false;
}
jcr->sendJobStatus(JS_WaitFD); /* wait for FD to connect */
Dmsg2(050, "sd_calls_client=%d sd_client=%d\n", jcr->sd_calls_client, jcr->sd_client);
if (jcr->sd_calls_client) {
if (!read_client_hello(jcr)) {
return false;
}
/*
* Authenticate the File daemon
*/
Dmsg0(050, "=== Authenticate FD\n");
if (jcr->authenticated || !authenticate_filed(jcr, jcr->file_bsock, jcr->FDVersion)) {
Dmsg1(050, "Authentication failed Job %s\n", jcr->Job);
Jmsg(jcr, M_FATAL, 0, _("Unable to authenticate File daemon\n"));
} else {
jcr->authenticated = true;
}
} else if (!jcr->sd_client) {
/* We wait to receive connection from Client */
gettimeofday(&tv, &tz);
timeout.tv_nsec = tv.tv_usec * 1000;
timeout.tv_sec = tv.tv_sec + me->client_wait;
Dmsg3(050, "%s waiting %d sec for FD to contact SD key=%s\n",
jcr->Job, (int)(timeout.tv_sec-time(NULL)), jcr->sd_auth_key);
Dmsg3(800, "=== Block Job=%s jid=%d %p\n", jcr->Job, jcr->JobId, jcr);
/*
* Wait for the File daemon to contact us to start the Job,
* when he does, we will be released, unless the 30 minutes
* expires.
*/
P(mutex);
while ( !jcr->authenticated && !job_canceled(jcr) ) {
errstat = pthread_cond_timedwait(&jcr->job_start_wait, &mutex, &timeout);
if (errstat == ETIMEDOUT || errstat == EINVAL || errstat == EPERM) {
break;
}
Dmsg1(800, "=== Auth cond errstat=%d\n", errstat);
}
Dmsg4(050, "=== Auth=%d jid=%d canceled=%d errstat=%d\n",
jcr->JobId, jcr->authenticated, job_canceled(jcr), errstat);
V(mutex);
Dmsg2(800, "Auth fail or cancel for jid=%d %p\n", jcr->JobId, jcr);
}
memset(jcr->sd_auth_key, 0, strlen(jcr->sd_auth_key));
if (jcr->authenticated && !job_canceled(jcr)) {
Dmsg2(800, "Running jid=%d %p\n", jcr->JobId, jcr);
run_job(jcr); /* Run the job */
}
Dmsg2(800, "Done jid=%d %p\n", jcr->JobId, jcr);
return false;
}
bool run_cmd(JCR *jcr)
{
struct timeval tv;
struct timezone tz;
struct timespec timeout;
int errstat = 0;
BSOCK *cl;
int fd_version = 0;
int sd_version = 0;
char job_name[500];
int i;
int stat;
Dsm_check(200);
Dmsg1(200, "Run_cmd: %s\n", jcr->dir_bsock->msg);
/* If we do not need the FD, we are doing a virtual backup. */
if (jcr->no_client_used()) {
do_vbackup(jcr);
return false;
}
jcr->sendJobStatus(JS_WaitFD); /* wait for FD to connect */
Dmsg2(050, "sd_calls_client=%d sd_client=%d\n", jcr->sd_calls_client, jcr->sd_client);
if (jcr->sd_calls_client) {
/* We connected to Client, so finish work */
cl = jcr->file_bsock;
if (!cl) {
Jmsg0(jcr, M_FATAL, 0, _("Client socket not open. Could not connect to Client.\n"));
Dmsg0(050, "Client socket not open. Could not connect to Client.\n");
return false;
}
/* Get response to Hello command sent earlier */
Dmsg0(050, "Read Hello command from Client\n");
for (i=0; i<60; i++) {
stat = cl->recv();
if (stat <= 0) {
bmicrosleep(1, 0);
} else {
break;
}
}
if (stat <= 0) {
berrno be;
Jmsg1(jcr, M_FATAL, 0, _("Recv request to Client failed. ERR=%s\n"),
be.bstrerror());
Dmsg1(050, _("Recv request to Client failed. ERR=%s\n"), be.bstrerror());
return false;
}
Dmsg1(050, "Got from FD: %s\n", cl->msg);
if (sscanf(cl->msg, "Hello Bacula SD: Start Job %127s %d %d", job_name, &fd_version, &sd_version) != 3) {
Jmsg1(jcr, M_FATAL, 0, _("Bad Hello from Client: %s.\n"), cl->msg);
Dmsg1(050, _("Bad Hello from Client: %s.\n"), cl->msg);
return false;
}
unbash_spaces(job_name);
jcr->FDVersion = fd_version;
jcr->SDVersion = sd_version;
Dmsg1(050, "FDVersion=%d\n", fd_version);
/*
* Authenticate the File daemon
*/
Dmsg0(050, "=== Authenticate FD\n");
if (jcr->authenticated || !authenticate_filed(jcr)) {
Dmsg1(050, "Authentication failed Job %s\n", jcr->Job);
Jmsg(jcr, M_FATAL, 0, _("Unable to authenticate File daemon\n"));
} else {
jcr->authenticated = true;
}
} else if (!jcr->sd_client) {
/* We wait to receive connection from Client */
gettimeofday(&tv, &tz);
timeout.tv_nsec = tv.tv_usec * 1000;
timeout.tv_sec = tv.tv_sec + me->client_wait;
Dmsg3(050, "%s waiting %d sec for FD to contact SD key=%s\n",
jcr->Job, (int)(timeout.tv_sec-time(NULL)), jcr->sd_auth_key);
Dmsg3(800, "=== Block Job=%s jid=%d %p\n", jcr->Job, jcr->JobId, jcr);
/*
* Wait for the File daemon to contact us to start the Job,
* when he does, we will be released, unless the 30 minutes
* expires.
*/
P(mutex);
while ( !jcr->authenticated && !job_canceled(jcr) ) {
errstat = pthread_cond_timedwait(&jcr->job_start_wait, &mutex, &timeout);
if (errstat == ETIMEDOUT || errstat == EINVAL || errstat == EPERM) {
break;
}
Dmsg1(800, "=== Auth cond errstat=%d\n", errstat);
}
Dmsg4(050, "=== Auth=%d jid=%d canceled=%d errstat=%d\n",
jcr->JobId, jcr->authenticated, job_canceled(jcr), errstat);
V(mutex);
Dmsg2(800, "Auth fail or cancel for jid=%d %p\n", jcr->JobId, jcr);
}
memset(jcr->sd_auth_key, 0, strlen(jcr->sd_auth_key));
if (jcr->authenticated && !job_canceled(jcr)) {
Dmsg2(800, "Running jid=%d %p\n", jcr->JobId, jcr);
run_job(jcr); /* Run the job */
}
Dmsg2(800, "Done jid=%d %p\n", jcr->JobId, jcr);
return false;
}
