/*
Bacula® - The Network Backup Solution
- Copyright (C) 2000-2007 Free Software Foundation Europe e.V.
+ Copyright (C) 2000-2008 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.
This program is Free Software; you can redistribute it and/or
modify it under the terms of version two of the GNU General Public
- License as published by the Free Software Foundation plus additions
- that are listed in the file LICENSE.
+ License as published by the Free Software Foundation and included
+ in the file LICENSE.
This program is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
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.
#include "bacula.h"
#include "jcr.h"
+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+ * when the jcr is no longer valid.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;
static pthread_mutex_t last_jobs_mutex = PTHREAD_MUTEX_INITIALIZER;
+static pthread_key_t jcr_key; /* Pointer to jcr for each thread */
+
+pthread_once_t key_once = PTHREAD_ONCE_INIT;
+
void lock_jobs()
{
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);
}
+/* Set Job type in JCR and also set appropriate read flag */
+void JCR::set_JobType(int32_t JobType)
+{
+ m_JobType = JobType;
+}
+
+/* Set Job level in JCR and also set appropriate read flag */
+void JCR::set_JobLevel(int32_t JobLevel)
+{
+ m_JobLevel = JobLevel;
+}
+
+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
*/
}
}
+void create_jcr_key()
+{
+ int status = pthread_key_create(&jcr_key, NULL);
+ if (status != 0) {
+ berrno be;
+ Jmsg1(NULL, M_ABORT, 0, _("pthread key create failed: ERR=%s\n"),
+ be.bstrerror(status));
+ }
+}
+
/*
* Create a Job Control Record and link it into JCR chain
* Returns newly allocated JCR
JCR *jcr;
MQUEUE_ITEM *item = NULL;
struct sigaction sigtimer;
+ int status;
- Dmsg0(3400, "Enter new_jcr\n");
+ Dmsg0(dbglvl, "Enter new_jcr\n");
+ status = pthread_once(&key_once, create_jcr_key);
+ if (status != 0) {
+ berrno be;
+ Jmsg1(NULL, M_ABORT, 0, _("pthread_once failed. ERR=%s\n"), be.bstrerror(status));
+ }
jcr = (JCR *)malloc(size);
memset(jcr, 0, size);
jcr->my_thread_id = pthread_self();
/* 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;
+ jcr->set_JobType(JT_SYSTEM); /* internal job until defined */
+ jcr->set_JobLevel(L_NONE);
set_jcr_job_status(jcr, JS_Created); /* ready to run */
-
+ set_jcr_in_tsd(jcr);
sigtimer.sa_flags = 0;
sigtimer.sa_handler = timeout_handler;
sigfillset(&sigtimer.sa_mask);
*/
static void remove_jcr(JCR *jcr)
{
- Dmsg0(3400, "Enter remove_jcr\n");
+ Dmsg0(dbglvl, "Enter remove_jcr\n");
if (!jcr) {
Emsg0(M_ABORT, 0, _("NULL jcr.\n"));
}
jcrs->remove(jcr);
- Dmsg0(3400, "Leave remove_jcr\n");
+ Dmsg0(dbglvl, "Leave remove_jcr\n");
}
/*
*/
static void free_common_jcr(JCR *jcr)
{
- struct s_last_job *je, last_job;
-
- /* Keep some statistics */
- switch (jcr->JobType) {
- 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) {
- je = (struct s_last_job *)malloc(sizeof(struct s_last_job));
- memcpy((char *)je, (char *)&last_job, sizeof(last_job));
- if (!last_jobs) {
- init_last_jobs_list();
- }
- last_jobs->append(je);
- if (last_jobs->size() > max_last_jobs) {
- je = (struct s_last_job *)last_jobs->first();
- last_jobs->remove(je);
- free(je);
- }
- }
- break;
- default:
- break;
- }
jcr->destroy_mutex();
if (jcr->msg_queue) {
jcr->cached_path = NULL;
jcr->cached_pnl = 0;
}
- free_getuser_cache();
- free_getgroup_cache();
+ if (jcr->id_list) {
+ free_guid_list(jcr->id_list);
+ jcr->id_list = NULL;
+ }
+ /* Invalidate the tsd jcr data */
+ set_jcr_in_tsd(INVALID_JCR);
free(jcr);
}
#ifdef DEBUG
void b_free_jcr(const char *file, int line, JCR *jcr)
{
- Dmsg3(3400, "Enter free_jcr 0x%x from %s:%d\n", jcr, file, line);
+ struct s_last_job *je;
+
+ Dmsg3(dbglvl, "Enter free_jcr jid=%u from %s:%d\n", jcr->JobId, file, line);
#else
void free_jcr(JCR *jcr)
{
+ struct s_last_job *je;
- Dmsg2(3400, "Enter free_jcr 0x%x job=%d\n", jcr, jcr->JobId);
+ Dmsg3(dbglvl, "Enter free_jcr jid=%u use_count=%d Job=%s\n",
+ jcr->JobId, jcr->use_count(), jcr->Job);
#endif
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);
}
- Dmsg3(3400, "Dec free_jcr 0x%x use_count=%d jobid=%d\n", jcr, jcr->use_count(), jcr->JobId);
+ if (jcr->JobId > 0) {
+ Dmsg3(dbglvl, "Dec free_jcr jid=%u use_count=%d Job=%s\n",
+ jcr->JobId, jcr->use_count(), jcr->Job);
+ }
if (jcr->use_count() > 0) { /* if in use */
unlock_jcr_chain();
- Dmsg3(3400, "free_jcr 0x%x job=%d use_count=%d\n", jcr, jcr->JobId, jcr->use_count());
return;
}
-
+ if (jcr->JobId > 0) {
+ Dmsg3(dbglvl, "remove jcr jid=%u use_count=%d Job=%s\n",
+ jcr->JobId, jcr->use_count(), jcr->Job);
+ }
remove_jcr(jcr); /* remove Jcr from chain */
- unlock_jcr_chain();
job_end_pop(jcr); /* pop and call hooked routines */
- Dmsg1(3400, "End job=%d\n", jcr->JobId);
+ Dmsg1(dbglvl, "End job=%d\n", jcr->JobId);
+
+ /* Keep some statistics */
+ switch (jcr->get_JobType()) {
+ case JT_BACKUP:
+ case JT_VERIFY:
+ case JT_RESTORE:
+ case JT_MIGRATE:
+ case JT_COPY:
+ case JT_ADMIN:
+ /* Keep list of last jobs, but not Console where JobId==0 */
+ if (jcr->JobId > 0) {
+ lock_last_jobs_list();
+ num_jobs_run++;
+ je = (struct s_last_job *)malloc(sizeof(struct s_last_job));
+ memset(je, 0, sizeof(struct s_last_job)); /* zero in case unset fields */
+ je->Errors = jcr->Errors;
+ je->JobType = jcr->get_JobType();
+ 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->get_JobLevel();
+ je->start_time = jcr->start_time;
+ je->end_time = time(NULL);
+
+ if (!last_jobs) {
+ init_last_jobs_list();
+ }
+ last_jobs->append(je);
+ if (last_jobs->size() > max_last_jobs) {
+ je = (struct s_last_job *)last_jobs->first();
+ 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(3400, "Exit free_jcr\n");
+ Dmsg0(dbglvl, "Exit free_jcr\n");
+}
+
+void set_jcr_in_tsd(JCR *jcr)
+{
+ int status = pthread_setspecific(jcr_key, (void *)jcr);
+ if (status != 0) {
+ berrno be;
+ Jmsg1(jcr, M_ABORT, 0, _("pthread_setspecific failed: ERR=%s\n"), be.bstrerror(status));
+ }
+}
+
+JCR *get_jcr_from_tsd()
+{
+ JCR *jcr = (JCR *)pthread_getspecific(jcr_key);
+// printf("get_jcr_from_tsd: jcr=%p\n", jcr);
+ /* set any INVALID_JCR to NULL which the rest of Bacula understands */
+ if (jcr == INVALID_JCR) {
+ jcr = NULL;
+ }
+ return jcr;
}
+
+/*
+ * Find which JobId corresponds to the current thread
+ */
+uint32_t get_jobid_from_tsd()
+{
+ JCR *jcr;
+ uint32_t JobId = 0;
+ jcr = get_jcr_from_tsd();
+// printf("get_jobid_from_tsr: jcr=%p\n", jcr);
+ if (jcr) {
+ JobId = (uint32_t)jcr->JobId;
+ }
+ return JobId;
+}
/*
* Given a JobId, find the JCR
foreach_jcr(jcr) {
if (jcr->JobId == JobId) {
jcr->inc_use_count();
- Dmsg2(3400, "Inc get_jcr 0x%x use_count=%d\n", jcr, jcr->use_count());
+ Dmsg3(dbglvl, "Inc get_jcr jid=%u use_count=%d Job=%s\n",
+ jcr->JobId, jcr->use_count(), jcr->Job);
break;
}
}
if (jcr->VolSessionId == SessionId &&
jcr->VolSessionTime == SessionTime) {
jcr->inc_use_count();
- Dmsg2(3400, "Inc get_jcr 0x%x use_count=%d\n", jcr, jcr->use_count());
+ Dmsg3(dbglvl, "Inc get_jcr jid=%u use_count=%d Job=%s\n",
+ jcr->JobId, jcr->use_count(), jcr->Job);
break;
}
}
foreach_jcr(jcr) {
if (strncmp(Job, jcr->Job, len) == 0) {
jcr->inc_use_count();
- Dmsg2(3400, "Inc get_jcr 0x%x use_count=%d\n", jcr, jcr->use_count());
+ Dmsg3(dbglvl, "Inc get_jcr jid=%u use_count=%d Job=%s\n",
+ jcr->JobId, jcr->use_count(), jcr->Job);
break;
}
}
foreach_jcr(jcr) {
if (strcmp(jcr->Job, Job) == 0) {
jcr->inc_use_count();
- Dmsg2(3400, "Inc get_jcr 0x%x use_count=%d\n", jcr, jcr->use_count());
+ Dmsg3(dbglvl, "Inc get_jcr jid=%u use_count=%d Job=%s\n",
+ jcr->JobId, jcr->use_count(), jcr->Job);
break;
}
}
void set_jcr_job_status(JCR *jcr, int JobStatus)
{
+ bool set_waittime=false;
+ Dmsg2(800, "set_jcr_job_status(%s, %c)\n", jcr->Job, JobStatus);
+ /* if wait state is new, we keep current time for watchdog MaxWaitTime */
+ switch (JobStatus) {
+ 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:
+ set_waittime = true;
+ default:
+ break;
+ }
+
/*
* For a set of errors, ... keep the current status
* so it isn't lost. For all others, set it.
*/
+ Dmsg3(300, "jid=%u OnEntry JobStatus=%c set=%c\n", (uint32_t)jcr->JobId,
+ jcr->JobStatus, JobStatus);
switch (jcr->JobStatus) {
case JS_ErrorTerminated:
- case JS_Error:
case JS_FatalError:
- case JS_Differences:
case JS_Canceled:
break;
+ case JS_Error:
+ case JS_Differences:
+ switch (JobStatus) {
+ case JS_ErrorTerminated:
+ case JS_FatalError:
+ case JS_Canceled:
+ /* Override more minor status */
+ jcr->JobStatus = JobStatus;
+ break;
+ default:
+ break;
+ }
+ /*
+ * For a set of Wait situation, keep old time.
+ */
+ 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:
+ set_waittime = false; /* keep old time */
default:
jcr->JobStatus = JobStatus;
+ if (set_waittime) {
+ /* set it before JobStatus */
+ Dmsg0(800, "Setting wait_time\n");
+ jcr->wait_time = time(NULL);
+ }
}
+ Dmsg3(200, "jid=%u leave set_jcr_job_status=%c set=%c\n", (uint32_t)jcr->JobId,
+ jcr->JobStatus, JobStatus);
}
#ifdef TRACE_JCR_CHAIN
#endif
{
#ifdef TRACE_JCR_CHAIN
- Dmsg3(3400, "Lock jcr chain %d from %s:%d\n", ++lock_count, fname, line);
+ Dmsg3(dbglvl, "Lock jcr chain %d from %s:%d\n", ++lock_count, fname, line);
#endif
P(jcr_lock);
}
#endif
{
#ifdef TRACE_JCR_CHAIN
- Dmsg3(3400, "Unlock jcr chain %d from %s:%d\n", lock_count--, fname, line);
+ Dmsg3(dbglvl, "Unlock jcr chain %d from %s:%d\n", lock_count--, fname, line);
#endif
V(jcr_lock);
}
jcr = (JCR *)jcrs->first();
if (jcr) {
jcr->inc_use_count();
- Dmsg3(3400, "Inc jcr_walk_start 0x%x job=%d use_count=%d\n", jcr, jcr->JobId, jcr->use_count());
+ if (jcr->JobId > 0) {
+ Dmsg3(dbglvl, "Inc walk_start jid=%u use_count=%d Job=%s\n",
+ jcr->JobId, jcr->use_count(), jcr->Job);
+ }
}
unlock_jcr_chain();
return jcr;
jcr = (JCR *)jcrs->next(prev_jcr);
if (jcr) {
jcr->inc_use_count();
- Dmsg3(3400, "Inc jcr_walk_next 0x%x job=%d use_count=%d\n", jcr, jcr->JobId, jcr->use_count());
+ if (jcr->JobId > 0) {
+ Dmsg3(dbglvl, "Inc walk_next jid=%u use_count=%d Job=%s\n",
+ jcr->JobId, jcr->use_count(), jcr->Job);
+ }
}
unlock_jcr_chain();
if (prev_jcr) {
void jcr_walk_end(JCR *jcr)
{
if (jcr) {
+ if (jcr->JobId > 0) {
+ Dmsg3(dbglvl, "Free walk_end jid=%u use_count=%d Job=%s\n",
+ jcr->JobId, jcr->use_count(), jcr->Job);
+ }
free_jcr(jcr);
}
}
static void jcr_timeout_check(watchdog_t *self)
{
JCR *jcr;
- BSOCK *fd;
+ BSOCK *bs;
time_t timer_start;
- Dmsg0(3400, "Start JCR timeout checks\n");
+ Dmsg0(dbglvl, "Start JCR timeout checks\n");
/* Walk through all JCRs checking if any one is
* blocked for more than specified max time.
*/
foreach_jcr(jcr) {
- Dmsg2(3400, "jcr_timeout_check JobId=%u jcr=0x%x\n", jcr->JobId, jcr);
+ Dmsg2(dbglvl, "jcr_timeout_check JobId=%u jcr=0x%x\n", jcr->JobId, jcr);
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->m_timed_out = true;
- 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->m_timed_out = true;
- 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->m_timed_out = true;
- 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);
}
endeach_jcr(jcr);
- Dmsg0(3400, "Finished JCR timeout checks\n");
+ Dmsg0(dbglvl, "Finished JCR timeout checks\n");
}
/*