/*
Bacula® - The Network Backup Solution
- Copyright (C) 2000-2008 Free Software Foundation Europe e.V.
+ Copyright (C) 2000-2010 Free Software Foundation Europe e.V.
The main author of Bacula is Kern Sibbald, with contributions from
many others, a complete list can be found in the file AUTHORS.
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
02110-1301, USA.
- Bacula® is a registered trademark of John Walker.
+ Bacula® is a registered trademark of Kern Sibbald.
The licensor of Bacula is the Free Software Foundation Europe
(FSFE), Fiduciary Program, Sumatrastrasse 25, 8006 Zürich,
Switzerland, email:ftf@fsfeurope.org.
*
* Kern E. Sibbald, December 2000
*
- * Version $Id$
- *
* These routines are thread safe.
*
* The job list routines were re-written in May 2005 to
const int dbglvl = 3400;
-/*
- * Setting a NULL in tsd doesn't clear the tsd but instead tells
- * pthreads not to call the tsd destructor. Consequently, we
- * define this *invalid* jcr address and stuff it in the tsd
- * when the jcr is no longer valid.
- */
-#define INVALID_JCR ((JCR *)(-1))
-
/* External variables we reference */
-extern time_t watchdog_time;
/* External referenced functions */
void free_bregexps(alist *bregexps);
void term_last_jobs_list()
{
if (last_jobs) {
+ lock_last_jobs_list();
while (!last_jobs->empty()) {
void *je = last_jobs->first();
last_jobs->remove(je);
}
delete last_jobs;
last_jobs = NULL;
+ unlock_last_jobs_list();
}
if (jcrs) {
delete jcrs;
{
struct s_last_job *je, job;
uint32_t num;
+ bool ok = true;
Dmsg1(100, "read_last_jobs seek to %d\n", (int)addr);
- if (addr == 0 || lseek(fd, (off_t)addr, SEEK_SET) < 0) {
+ if (addr == 0 || lseek(fd, (boffset_t)addr, SEEK_SET) < 0) {
return false;
}
if (read(fd, &num, sizeof(num)) != sizeof(num)) {
if (num > 4 * max_last_jobs) { /* sanity check */
return false;
}
+ lock_last_jobs_list();
for ( ; num; num--) {
if (read(fd, &job, sizeof(job)) != sizeof(job)) {
berrno be;
Pmsg1(000, "Read job entry. ERR=%s\n", be.bstrerror());
- return false;
+ ok = false;
+ break;
}
if (job.JobId > 0) {
je = (struct s_last_job *)malloc(sizeof(struct s_last_job));
}
}
}
- return true;
+ unlock_last_jobs_list();
+ return ok;
}
uint64_t write_last_jobs_list(int fd, uint64_t addr)
{
struct s_last_job *je;
uint32_t num;
+ ssize_t stat;
Dmsg1(100, "write_last_jobs seek to %d\n", (int)addr);
- if (lseek(fd, (off_t)addr, SEEK_SET) < 0) {
+ if (lseek(fd, (boffset_t)addr, SEEK_SET) < 0) {
return 0;
}
if (last_jobs) {
+ lock_last_jobs_list();
/* First record is number of entires */
num = last_jobs->size();
if (write(fd, &num, sizeof(num)) != sizeof(num)) {
berrno be;
Pmsg1(000, "Error writing num_items: ERR=%s\n", be.bstrerror());
- return 0;
+ goto bail_out;
}
foreach_dlist(je, last_jobs) {
if (write(fd, je, sizeof(struct s_last_job)) != sizeof(struct s_last_job)) {
berrno be;
Pmsg1(000, "Error writing job: ERR=%s\n", be.bstrerror());
- return 0;
+ goto bail_out;
}
}
+ unlock_last_jobs_list();
}
/* Return current address */
- ssize_t stat = lseek(fd, 0, SEEK_CUR);
+ stat = lseek(fd, 0, SEEK_CUR);
if (stat < 0) {
stat = 0;
}
return stat;
+bail_out:
+ unlock_last_jobs_list();
+ return 0;
}
void lock_last_jobs_list()
V(last_jobs_mutex);
}
+/* Get an ASCII representation of the Operation being performed as an english Noun */
+const char *JCR::get_OperationName()
+{
+ switch(m_JobType) {
+ case JT_BACKUP:
+ return _("Backup");
+ case JT_VERIFY:
+ return _("Verifying");
+ case JT_RESTORE:
+ return _("Restoring");
+ case JT_ARCHIVE:
+ return _("Archiving");
+ case JT_COPY:
+ return _("Copying");
+ case JT_MIGRATE:
+ return _("Migration");
+ case JT_SCAN:
+ return _("Scanning");
+ default:
+ return _("Unknown operation");
+ }
+}
+
+/* Get an ASCII representation of the Action being performed either an english Verb or Adjective */
+const char *JCR::get_ActionName(bool past)
+{
+ switch(m_JobType) {
+ case JT_BACKUP:
+ return _("backup");
+ case JT_VERIFY:
+ return (past == true) ? _("verified") : _("verify");
+ case JT_RESTORE:
+ return (past == true) ? _("restored") : _("restore");
+ case JT_ARCHIVE:
+ return (past == true) ? _("archived") : _("archive");
+ case JT_COPY:
+ return (past == true) ? _("copied") : _("copy");
+ case JT_MIGRATE:
+ return (past == true) ? _("migrated") : _("migrate");
+ case JT_SCAN:
+ return (past == true) ? _("scanned") : _("scan");
+ default:
+ return _("unknown action");
+ }
+}
+
+bool JCR::JobReads()
+{
+ switch (m_JobType) {
+ case JT_VERIFY:
+ case JT_RESTORE:
+ case JT_COPY:
+ case JT_MIGRATE:
+ return true;
+ case JT_BACKUP:
+ if (m_JobLevel == L_VIRTUAL_FULL) {
+ return true;
+ }
+ break;
+ default:
+ break;
+ }
+ return false;
+}
+
/*
* Push a subroutine address into the job end callback stack
*/
}
}
+/*
+ * Create thread key for thread specific data
+ */
void create_jcr_key()
{
int status = pthread_key_create(&jcr_key, NULL);
memset(jcr, 0, size);
jcr->my_thread_id = pthread_self();
jcr->msg_queue = New(dlist(item, &item->link));
+ if ((status = pthread_mutex_init(&jcr->msg_queue_mutex, NULL)) != 0) {
+ berrno be;
+ Jmsg(NULL, M_ABORT, 0, _("Could not init msg_queue mutex. ERR=%s\n"),
+ be.bstrerror(status));
+ }
jcr->job_end_push.init(1, false);
jcr->sched_time = time(NULL);
jcr->daemon_free_jcr = daemon_free_jcr; /* plug daemon free routine */
/* Setup some dummy values */
bstrncpy(jcr->Job, "*System*", sizeof(jcr->Job));
jcr->JobId = 0;
- jcr->JobType = JT_SYSTEM; /* internal job until defined */
- jcr->JobLevel = L_NONE;
- set_jcr_job_status(jcr, JS_Created); /* ready to run */
+ jcr->setJobType(JT_SYSTEM); /* internal job until defined */
+ jcr->setJobLevel(L_NONE);
+ jcr->setJobStatus(JS_Created); /* ready to run */
set_jcr_in_tsd(jcr);
sigtimer.sa_flags = 0;
sigtimer.sa_handler = timeout_handler;
if (jcr->msg_queue) {
delete jcr->msg_queue;
jcr->msg_queue = NULL;
+ pthread_mutex_destroy(&jcr->msg_queue_mutex);
}
close_msg(jcr); /* close messages for this job */
free_guid_list(jcr->id_list);
jcr->id_list = NULL;
}
- /* Invalidate the tsd jcr data */
- set_jcr_in_tsd(INVALID_JCR);
+ remove_jcr_from_tsd(jcr);
free(jcr);
}
#ifdef DEBUG
void b_free_jcr(const char *file, int line, JCR *jcr)
{
- struct s_last_job *je, last_job;
+ struct s_last_job *je;
Dmsg3(dbglvl, "Enter free_jcr jid=%u from %s:%d\n", jcr->JobId, file, line);
void free_jcr(JCR *jcr)
{
- struct s_last_job *je, last_job;
+ struct s_last_job *je;
Dmsg3(dbglvl, "Enter free_jcr jid=%u use_count=%d Job=%s\n",
jcr->JobId, jcr->use_count(), jcr->Job);
#endif
- dequeue_messages(jcr);
lock_jcr_chain();
jcr->dec_use_count(); /* decrement use count */
if (jcr->use_count() < 0) {
- Emsg2(M_ERROR, 0, _("JCR use_count=%d JobId=%d\n"),
+ Jmsg2(jcr, M_ERROR, 0, _("JCR use_count=%d JobId=%d\n"),
jcr->use_count(), jcr->JobId);
}
if (jcr->JobId > 0) {
jcr->JobId, jcr->use_count(), jcr->Job);
}
remove_jcr(jcr); /* remove Jcr from chain */
+ unlock_jcr_chain();
+ dequeue_messages(jcr);
job_end_pop(jcr); /* pop and call hooked routines */
Dmsg1(dbglvl, "End job=%d\n", jcr->JobId);
/* Keep some statistics */
- switch (jcr->JobType) {
+ switch (jcr->getJobType()) {
case JT_BACKUP:
case JT_VERIFY:
case JT_RESTORE:
case JT_MIGRATE:
case JT_COPY:
case JT_ADMIN:
- num_jobs_run++;
- last_job.Errors = jcr->Errors;
- last_job.JobType = jcr->JobType;
- last_job.JobId = jcr->JobId;
- last_job.VolSessionId = jcr->VolSessionId;
- last_job.VolSessionTime = jcr->VolSessionTime;
- bstrncpy(last_job.Job, jcr->Job, sizeof(last_job.Job));
- last_job.JobFiles = jcr->JobFiles;
- last_job.JobBytes = jcr->JobBytes;
- last_job.JobStatus = jcr->JobStatus;
- last_job.JobLevel = jcr->JobLevel;
- last_job.start_time = jcr->start_time;
- last_job.end_time = time(NULL);
/* Keep list of last jobs, but not Console where JobId==0 */
- if (last_job.JobId > 0) {
+ if (jcr->JobId > 0) {
+ lock_last_jobs_list();
+ num_jobs_run++;
je = (struct s_last_job *)malloc(sizeof(struct s_last_job));
- memcpy((char *)je, (char *)&last_job, sizeof(last_job));
+ memset(je, 0, sizeof(struct s_last_job)); /* zero in case unset fields */
+ je->Errors = jcr->JobErrors;
+ je->JobType = jcr->getJobType();
+ je->JobId = jcr->JobId;
+ je->VolSessionId = jcr->VolSessionId;
+ je->VolSessionTime = jcr->VolSessionTime;
+ bstrncpy(je->Job, jcr->Job, sizeof(je->Job));
+ je->JobFiles = jcr->JobFiles;
+ je->JobBytes = jcr->JobBytes;
+ je->JobStatus = jcr->JobStatus;
+ je->JobLevel = jcr->getJobLevel();
+ je->start_time = jcr->start_time;
+ je->end_time = time(NULL);
+
if (!last_jobs) {
init_last_jobs_list();
}
last_jobs->remove(je);
free(je);
}
+ unlock_last_jobs_list();
}
break;
default:
break;
-
}
+
if (jcr->daemon_free_jcr) {
jcr->daemon_free_jcr(jcr); /* call daemon free routine */
}
- unlock_jcr_chain();
free_common_jcr(jcr);
close_msg(NULL); /* flush any daemon messages */
garbage_collect_memory_pool();
Dmsg0(dbglvl, "Exit free_jcr\n");
}
+/*
+ * Remove jcr from thread specific data, but
+ * but make sure it is us who are attached.
+ */
+void remove_jcr_from_tsd(JCR *jcr)
+{
+ JCR *tjcr = get_jcr_from_tsd();
+ if (tjcr == jcr) {
+ set_jcr_in_tsd(INVALID_JCR);
+ }
+}
+
+/*
+ * Put this jcr in the thread specifc data
+ */
void set_jcr_in_tsd(JCR *jcr)
{
int status = pthread_setspecific(jcr_key, (void *)jcr);
}
}
+/*
+ * Give me the jcr that is attached to this thread
+ */
JCR *get_jcr_from_tsd()
{
JCR *jcr = (JCR *)pthread_getspecific(jcr_key);
return jcr;
}
+/*
+ * Given a thread id, find the JobId
+ * Returns: JobId on success
+ * 0 on failure
+ */
+uint32_t get_jobid_from_tid(pthread_t tid)
+{
+ JCR *jcr = NULL;
+ bool found = false;
+
+ foreach_jcr(jcr) {
+ if (pthread_equal(jcr->my_thread_id, tid)) {
+ found = true;
+ break;
+ }
+ }
+ endeach_jcr(jcr);
+ if (found) {
+ return jcr->JobId;
+ }
+ return 0;
+}
+
+
/*
* Given a SessionId and SessionTime, find the JCR
* Returns: jcr on success
return jcr;
}
-void set_jcr_job_status(JCR *jcr, int JobStatus)
+static void update_wait_time(JCR *jcr, int newJobStatus)
{
+ bool enter_in_waittime;
+ int oldJobStatus = jcr->JobStatus;
+
+ switch (newJobStatus) {
+ case JS_WaitFD:
+ case JS_WaitSD:
+ case JS_WaitMedia:
+ case JS_WaitMount:
+ case JS_WaitStoreRes:
+ case JS_WaitJobRes:
+ case JS_WaitClientRes:
+ case JS_WaitMaxJobs:
+ case JS_WaitPriority:
+ enter_in_waittime = true;
+ break;
+ default:
+ enter_in_waittime = false; /* not a Wait situation */
+ break;
+ }
+
/*
- * For a set of errors, ... keep the current status
- * so it isn't lost. For all others, set it.
+ * If we were previously waiting and are not any more
+ * we want to update the wait_time variable, which is
+ * the start of waiting.
*/
- Dmsg3(300, "jid=%u OnEntry JobStatus=%c set=%c\n", (uint32_t)jcr->JobId,
- jcr->JobStatus, JobStatus);
- switch (jcr->JobStatus) {
+ switch (oldJobStatus) {
+ case JS_WaitFD:
+ case JS_WaitSD:
+ case JS_WaitMedia:
+ case JS_WaitMount:
+ case JS_WaitStoreRes:
+ case JS_WaitJobRes:
+ case JS_WaitClientRes:
+ case JS_WaitMaxJobs:
+ case JS_WaitPriority:
+ if (!enter_in_waittime) { /* we get out the wait time */
+ jcr->wait_time_sum += (time(NULL) - jcr->wait_time);
+ jcr->wait_time = 0;
+ }
+ break;
+
+ /* if wait state is new, we keep current time for watchdog MaxWaitTime */
+ default:
+ if (enter_in_waittime) {
+ jcr->wait_time = time(NULL);
+ }
+ break;
+ }
+}
+
+/*
+ * Priority runs from 0 (lowest) to 10 (highest)
+ */
+static int get_status_priority(int JobStatus)
+{
+ int priority = 0;
+ switch (JobStatus) {
case JS_ErrorTerminated:
case JS_FatalError:
case JS_Canceled:
+ case JS_Incomplete:
+ priority = 10;
break;
case JS_Error:
+ priority = 8;
+ break;
case JS_Differences:
- switch (JobStatus) {
- case JS_ErrorTerminated:
- case JS_FatalError:
- case JS_Canceled:
- /* Override more minor status */
- jcr->JobStatus = JobStatus;
- break;
- }
+ priority = 7;
break;
- default:
- jcr->JobStatus = JobStatus;
}
- Dmsg3(100, "jid=%u OnExit JobStatus=%c set=%c\n", (uint32_t)jcr->JobId,
- jcr->JobStatus, JobStatus);
+ return priority;
+}
+
+
+void set_jcr_job_status(JCR *jcr, int JobStatus)
+{
+ jcr->setJobStatus(JobStatus);
+}
+
+void JCR::setJobStatus(int newJobStatus)
+{
+ JCR *jcr = this;
+ int priority, old_priority;
+ int oldJobStatus = jcr->JobStatus;
+ priority = get_status_priority(newJobStatus);
+ old_priority = get_status_priority(oldJobStatus);
+
+ Dmsg2(800, "set_jcr_job_status(%s, %c)\n", Job, newJobStatus);
+
+ /* Update wait_time depending on newJobStatus and oldJobStatus */
+ update_wait_time(jcr, newJobStatus);
+
+ /*
+ * For a set of errors, ... keep the current status
+ * so it isn't lost. For all others, set it.
+ */
+ Dmsg2(800, "OnEntry JobStatus=%c newJobstatus=%c\n", oldJobStatus, newJobStatus);
+ /*
+ * If status priority is > than proposed new status, change it.
+ * If status priority == new priority and both are zero, take
+ * the new status.
+ * If it is not zero, then we keep the first non-zero "error" that
+ * occurred.
+ */
+ if (priority > old_priority || (
+ priority == 0 && old_priority == 0)) {
+ Dmsg4(800, "Set new stat. old: %c,%d new: %c,%d\n",
+ jcr->JobStatus, old_priority, newJobStatus, priority);
+ jcr->JobStatus = newJobStatus; /* replace with new status */
+ }
+
+ if (oldJobStatus != jcr->JobStatus) {
+ Dmsg2(800, "leave set_job_status old=%c new=%c\n", oldJobStatus, newJobStatus);
+// generate_plugin_event(jcr, bEventStatusChange, NULL);
+ }
}
#ifdef TRACE_JCR_CHAIN
V(jcr_lock);
}
-
/*
* Start walk of jcr chain
* The proper way to walk the jcr chain is:
}
}
+/*
+ * Return number of Jobs
+ */
+int job_count()
+{
+ JCR *jcr;
+ int count = 0;
+
+ lock_jcr_chain();
+ for (jcr = (JCR *)jcrs->first(); jcr = (JCR *)jcrs->next(jcr); ) {
+ if (jcr->JobId > 0) {
+ count++;
+ }
+ }
+ unlock_jcr_chain();
+ return count;
+}
+
/*
* Setup to call the timeout check routine every 30 seconds
static void jcr_timeout_check(watchdog_t *self)
{
JCR *jcr;
- BSOCK *fd;
+ BSOCK *bs;
time_t timer_start;
Dmsg0(dbglvl, "Start JCR timeout checks\n");
if (jcr->JobId == 0) {
continue;
}
- fd = jcr->store_bsock;
- if (fd) {
- timer_start = fd->timer_start;
- if (timer_start && (watchdog_time - timer_start) > fd->timeout) {
- fd->timer_start = 0; /* turn off timer */
- fd->set_timed_out();
- Jmsg(jcr, M_ERROR, 0, _(
+ bs = jcr->store_bsock;
+ if (bs) {
+ timer_start = bs->timer_start;
+ if (timer_start && (watchdog_time - timer_start) > bs->timeout) {
+ bs->timer_start = 0; /* turn off timer */
+ bs->set_timed_out();
+ Qmsg(jcr, M_ERROR, 0, _(
"Watchdog sending kill after %d secs to thread stalled reading Storage daemon.\n"),
watchdog_time - timer_start);
pthread_kill(jcr->my_thread_id, TIMEOUT_SIGNAL);
}
}
- fd = jcr->file_bsock;
- if (fd) {
- timer_start = fd->timer_start;
- if (timer_start && (watchdog_time - timer_start) > fd->timeout) {
- fd->timer_start = 0; /* turn off timer */
- fd->set_timed_out();
- Jmsg(jcr, M_ERROR, 0, _(
+ bs = jcr->file_bsock;
+ if (bs) {
+ timer_start = bs->timer_start;
+ if (timer_start && (watchdog_time - timer_start) > bs->timeout) {
+ bs->timer_start = 0; /* turn off timer */
+ bs->set_timed_out();
+ Qmsg(jcr, M_ERROR, 0, _(
"Watchdog sending kill after %d secs to thread stalled reading File daemon.\n"),
watchdog_time - timer_start);
pthread_kill(jcr->my_thread_id, TIMEOUT_SIGNAL);
}
}
- fd = jcr->dir_bsock;
- if (fd) {
- timer_start = fd->timer_start;
- if (timer_start && (watchdog_time - timer_start) > fd->timeout) {
- fd->timer_start = 0; /* turn off timer */
- fd->set_timed_out();
- Jmsg(jcr, M_ERROR, 0, _(
+ bs = jcr->dir_bsock;
+ if (bs) {
+ timer_start = bs->timer_start;
+ if (timer_start && (watchdog_time - timer_start) > bs->timeout) {
+ bs->timer_start = 0; /* turn off timer */
+ bs->set_timed_out();
+ Qmsg(jcr, M_ERROR, 0, _(
"Watchdog sending kill after %d secs to thread stalled reading Director.\n"),
watchdog_time - timer_start);
pthread_kill(jcr->my_thread_id, TIMEOUT_SIGNAL);
Dmsg0(dbglvl, "Finished JCR timeout checks\n");
}
+/*
+ * Return next JobId from comma separated list
+ *
+ * Returns:
+ * 1 if next JobId returned
+ * 0 if no more JobIds are in list
+ * -1 there is an error
+ */
+int get_next_jobid_from_list(char **p, uint32_t *JobId)
+{
+ const int maxlen = 30;
+ char jobid[maxlen+1];
+ char *q = *p;
+
+ jobid[0] = 0;
+ for (int i=0; i<maxlen; i++) {
+ if (*q == 0) {
+ break;
+ } else if (*q == ',') {
+ q++;
+ break;
+ }
+ jobid[i] = *q++;
+ jobid[i+1] = 0;
+ }
+ if (jobid[0] == 0) {
+ return 0;
+ } else if (!is_a_number(jobid)) {
+ return -1; /* error */
+ }
+ *p = q;
+ *JobId = str_to_int64(jobid);
+ return 1;
+}
+
/*
* Timeout signal comes here
*/
{
return; /* thus interrupting the function */
}
+
+/* Used to display specific daemon information after a fatal signal
+ * (like B_DB in the director)
+ */
+#define MAX_DBG_HOOK 10
+static dbg_jcr_hook_t *dbg_jcr_hooks[MAX_DBG_HOOK];
+static int dbg_jcr_handler_count;
+
+void dbg_jcr_add_hook(dbg_jcr_hook_t *hook)
+{
+ ASSERT(dbg_jcr_handler_count < MAX_DBG_HOOK);
+ dbg_jcr_hooks[dbg_jcr_handler_count++] = hook;
+}
+
+/*
+ * !!! WARNING !!!
+ *
+ * This function should be used ONLY after a fatal signal. We walk through the
+ * JCR chain without doing any lock, Bacula should not be running.
+ */
+void dbg_print_jcr(FILE *fp)
+{
+ char buf1[128], buf2[128], buf3[128], buf4[128];
+ if (!jcrs) {
+ return;
+ }
+
+ fprintf(fp, "Attempt to dump current JCRs\n");
+
+ for (JCR *jcr = (JCR *)jcrs->first(); jcr ; jcr = (JCR *)jcrs->next(jcr)) {
+ fprintf(fp, "JCR=%p JobId=%d name=%s JobStatus=%c\n",
+ jcr, (int)jcr->JobId, jcr->Job, jcr->JobStatus);
+ fprintf(fp, "\tuse_count=%i\n", jcr->use_count());
+ fprintf(fp, "\tJobType=%c JobLevel=%c\n",
+ jcr->getJobType(), jcr->getJobLevel());
+ bstrftime(buf1, sizeof(buf1), jcr->sched_time);
+ bstrftime(buf2, sizeof(buf2), jcr->start_time);
+ bstrftime(buf3, sizeof(buf3), jcr->end_time);
+ bstrftime(buf4, sizeof(buf4), jcr->wait_time);
+ fprintf(fp, "\tsched_time=%s start_time=%s\n\tend_time=%s wait_time=%s\n",
+ buf1, buf2, buf3, buf4);
+ fprintf(fp, "\tdb=%p db_batch=%p batch_started=%i\n",
+ jcr->db, jcr->db_batch, jcr->batch_started);
+
+ /*
+ * Call all the jcr debug hooks
+ */
+ for(int i=0; i < dbg_jcr_handler_count; i++) {
+ dbg_jcr_hook_t *hook = dbg_jcr_hooks[i];
+ hook(jcr, fp);
+ }
+ }
+}
projects / bacula / bacula / blobdiff