* in the list while traversing it rather than a single lock
* at the beginning of a traversal and one at the end. This
* incurs slightly more overhead, but effectively eliminates
- * the possibilty of race conditions. In addition, with the
+ * the possibilty of race conditions. In addition, with the
* exception of the global locking of the list during the
* re-reading of the config file, no recursion is needed.
*
*/
/*
- Copyright (C) 2000-2005 Kern Sibbald
+ Copyright (C) 2000-2006 Kern Sibbald
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
- version 2 as ammended with additional clauses defined in the
+ version 2 as amended with additional clauses defined in the
file LICENSE in the main source directory.
This program is distributed in the hope that it will be useful,
dlist *last_jobs = NULL;
const int max_last_jobs = 10;
-static dlist *jcrs = NULL; /* JCR chain */
+static dlist *jcrs = NULL; /* JCR chain */
static pthread_mutex_t jcr_lock = PTHREAD_MUTEX_INITIALIZER;
static pthread_mutex_t job_start_mutex = PTHREAD_MUTEX_INITIALIZER;
+static pthread_mutex_t last_jobs_mutex = PTHREAD_MUTEX_INITIALIZER;
+
+
void lock_jobs()
{
P(job_start_mutex);
{
if (last_jobs) {
while (!last_jobs->empty()) {
- void *je = last_jobs->first();
- last_jobs->remove(je);
- free(je);
+ void *je = last_jobs->first();
+ last_jobs->remove(je);
+ free(je);
}
delete last_jobs;
last_jobs = NULL;
}
}
-void read_last_jobs_list(int fd, uint64_t addr)
+bool read_last_jobs_list(int fd, uint64_t addr)
{
struct s_last_job *je, job;
uint32_t num;
Dmsg1(100, "read_last_jobs seek to %d\n", (int)addr);
if (addr == 0 || lseek(fd, (off_t)addr, SEEK_SET) < 0) {
- return;
+ return false;
}
if (read(fd, &num, sizeof(num)) != sizeof(num)) {
- return;
+ return false;
}
Dmsg1(100, "Read num_items=%d\n", num);
if (num > 4 * max_last_jobs) { /* sanity check */
- return;
+ return false;
}
for ( ; num; num--) {
if (read(fd, &job, sizeof(job)) != sizeof(job)) {
Dmsg1(000, "Read job entry. ERR=%s\n", strerror(errno));
- return;
+ return false;
}
if (job.JobId > 0) {
- je = (struct s_last_job *)malloc(sizeof(struct s_last_job));
- memcpy((char *)je, (char *)&job, sizeof(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);
- }
+ je = (struct s_last_job *)malloc(sizeof(struct s_last_job));
+ memcpy((char *)je, (char *)&job, sizeof(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);
+ }
}
}
+ return true;
}
uint64_t write_last_jobs_list(int fd, uint64_t addr)
num = last_jobs->size();
if (write(fd, &num, sizeof(num)) != sizeof(num)) {
Dmsg1(000, "Error writing num_items: ERR=%s\n", strerror(errno));
- return 0;
+ return 0;
}
foreach_dlist(je, last_jobs) {
- if (write(fd, je, sizeof(struct s_last_job)) != sizeof(struct s_last_job)) {
+ if (write(fd, je, sizeof(struct s_last_job)) != sizeof(struct s_last_job)) {
Dmsg1(000, "Error writing job: ERR=%s\n", strerror(errno));
- return 0;
- }
+ return 0;
+ }
}
}
/* Return current address */
void lock_last_jobs_list()
{
- /* Use jcr chain mutex */
- lock_jcr_chain();
+ P(last_jobs_mutex);
}
void unlock_last_jobs_list()
{
- /* Use jcr chain mutex */
- unlock_jcr_chain();
+ V(last_jobs_mutex);
}
/*
jcr->job_end_push.init(1, false);
jcr->sched_time = time(NULL);
jcr->daemon_free_jcr = daemon_free_jcr; /* plug daemon free routine */
- jcr->use_count = 1;
- pthread_mutex_init(&(jcr->mutex), NULL);
- jcr->JobStatus = JS_Created; /* ready to run */
+ jcr->init_mutex();
+ jcr->inc_use_count();
jcr->VolumeName = get_pool_memory(PM_FNAME);
jcr->VolumeName[0] = 0;
jcr->errmsg = get_pool_memory(PM_MESSAGE);
/* Setup some dummy values */
bstrncpy(jcr->Job, "*System*", sizeof(jcr->Job));
jcr->JobId = 0;
- jcr->JobType = JT_SYSTEM; /* internal job until defined */
+ jcr->JobType = JT_SYSTEM; /* internal job until defined */
jcr->JobLevel = L_NONE;
- jcr->JobStatus = JS_Created;
+ set_jcr_job_status(jcr, JS_Created); /* ready to run */
sigtimer.sa_flags = 0;
sigtimer.sa_handler = timeout_handler;
/*
* Remove a JCR from the chain
* NOTE! The chain must be locked prior to calling
- * this routine.
+ * this routine.
*/
static void remove_jcr(JCR *jcr)
{
Dmsg0(3400, "Enter remove_jcr\n");
if (!jcr) {
- Emsg0(M_ABORT, 0, "NULL jcr.\n");
+ Emsg0(M_ABORT, 0, _("NULL jcr.\n"));
}
jcrs->remove(jcr);
Dmsg0(3400, "Leave remove_jcr\n");
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.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);
- }
+ 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;
}
- pthread_mutex_destroy(&jcr->mutex);
+ jcr->destroy_mutex();
- delete jcr->msg_queue;
- close_msg(jcr); /* close messages for this job */
+ if (jcr->msg_queue) {
+ delete jcr->msg_queue;
+ jcr->msg_queue = NULL;
+ }
+ close_msg(jcr); /* close messages for this job */
/* do this after closing messages */
if (jcr->client_name) {
void free_jcr(JCR *jcr)
{
- Dmsg1(3400, "Enter free_jcr 0x%x\n", jcr);
+ Dmsg2(3400, "Enter free_jcr 0x%x job=%d\n", jcr, jcr->JobId);
#endif
dequeue_messages(jcr);
lock_jcr_chain();
- jcr->dec_use_count(); /* decrement use count */
- if (jcr->use_count < 0) {
+ jcr->dec_use_count(); /* decrement use count */
+ if (jcr->use_count() < 0) {
Emsg2(M_ERROR, 0, _("JCR use_count=%d JobId=%d\n"),
- jcr->use_count, jcr->JobId);
+ 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->use_count > 0) { /* if in use */
+ Dmsg3(3400, "Dec free_jcr 0x%x use_count=%d jobid=%d\n", jcr, jcr->use_count(), jcr->JobId);
+ if (jcr->use_count() > 0) { /* if in use */
unlock_jcr_chain();
- Dmsg2(3400, "free_jcr 0x%x use_count=%d\n", jcr, jcr->use_count);
+ Dmsg3(3400, "free_jcr 0x%x job=%d use_count=%d\n", jcr, jcr->JobId, jcr->use_count());
return;
}
- remove_jcr(jcr); /* remove Jcr from chain */
+ remove_jcr(jcr); /* remove Jcr from chain */
unlock_jcr_chain();
- job_end_pop(jcr); /* pop and call hooked routines */
+ job_end_pop(jcr); /* pop and call hooked routines */
Dmsg1(3400, "End job=%d\n", jcr->JobId);
if (jcr->daemon_free_jcr) {
jcr->daemon_free_jcr(jcr); /* call daemon free routine */
}
free_common_jcr(jcr);
- close_msg(NULL); /* flush any daemon messages */
+ close_msg(NULL); /* flush any daemon messages */
+ garbage_collect_memory_pool();
Dmsg0(3400, "Exit free_jcr\n");
}
/*
* Given a JobId, find the JCR
* Returns: jcr on success
- * NULL on failure
+ * NULL on failure
*/
JCR *get_jcr_by_id(uint32_t JobId)
{
JCR *jcr;
- lock_jcr_chain(); /* lock chain */
- foreach_dlist(jcr, jcrs) {
+ 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);
- break;
+ jcr->inc_use_count();
+ Dmsg2(3400, "Inc get_jcr 0x%x use_count=%d\n", jcr, jcr->use_count());
+ break;
}
}
- unlock_jcr_chain();
+ endeach_jcr(jcr);
return jcr;
}
/*
* Given a SessionId and SessionTime, find the JCR
* Returns: jcr on success
- * NULL on failure
+ * NULL on failure
*/
JCR *get_jcr_by_session(uint32_t SessionId, uint32_t SessionTime)
{
JCR *jcr;
- lock_jcr_chain();
- foreach_dlist(jcr, jcrs) {
+ foreach_jcr(jcr) {
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);
- break;
+ jcr->VolSessionTime == SessionTime) {
+ jcr->inc_use_count();
+ Dmsg2(3400, "Inc get_jcr 0x%x use_count=%d\n", jcr, jcr->use_count());
+ break;
}
}
- unlock_jcr_chain();
+ endeach_jcr(jcr);
return jcr;
}
* compares on the number of characters in Job
* thus allowing partial matches.
* Returns: jcr on success
- * NULL on failure
+ * NULL on failure
*/
JCR *get_jcr_by_partial_name(char *Job)
{
if (!Job) {
return NULL;
}
- lock_jcr_chain();
len = strlen(Job);
- foreach_dlist(jcr, jcrs) {
+ 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);
- break;
+ jcr->inc_use_count();
+ Dmsg2(3400, "Inc get_jcr 0x%x use_count=%d\n", jcr, jcr->use_count());
+ break;
}
}
- unlock_jcr_chain();
+ endeach_jcr(jcr);
return jcr;
}
* Given a Job, find the JCR
* requires an exact match of names.
* Returns: jcr on success
- * NULL on failure
+ * NULL on failure
*/
JCR *get_jcr_by_full_name(char *Job)
{
if (!Job) {
return NULL;
}
- lock_jcr_chain();
- foreach_dlist(jcr, jcrs) {
+ 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);
- break;
+ jcr->inc_use_count();
+ Dmsg2(3400, "Inc get_jcr 0x%x use_count=%d\n", jcr, jcr->use_count());
+ break;
}
}
- unlock_jcr_chain();
+ endeach_jcr(jcr);
return jcr;
}
}
-JCR *get_next_jcr(JCR *prev_jcr)
+/*
+ * Start walk of jcr chain
+ * The proper way to walk the jcr chain is:
+ * JCR *jcr;
+ * foreach_jcr(jcr) {
+ * ...
+ * }
+ * endeach_jcr(jcr);
+ *
+ * It is possible to leave out the endeach_jcr(jcr), but
+ * in that case, the last jcr referenced must be explicitly
+ * released with:
+ *
+ * free_jcr(jcr);
+ *
+ */
+JCR *jcr_walk_start()
+{
+ JCR *jcr;
+ lock_jcr_chain();
+ 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());
+ }
+ unlock_jcr_chain();
+ return jcr;
+}
+
+/*
+ * Get next jcr from chain, and release current one
+ */
+JCR *jcr_walk_next(JCR *prev_jcr)
{
JCR *jcr;
jcr = (JCR *)jcrs->next(prev_jcr);
if (jcr) {
jcr->inc_use_count();
- Dmsg2(3400, "Inc get_next_jcr 0x%x use_count=%d\n", jcr, jcr->use_count);
+ Dmsg3(3400, "Inc jcr_walk_next 0x%x job=%d use_count=%d\n", jcr,
+ jcr->JobId, jcr->use_count());
}
unlock_jcr_chain();
+ if (prev_jcr) {
+ free_jcr(prev_jcr);
+ }
return jcr;
}
+/*
+ * Release last jcr referenced
+ */
+void jcr_walk_end(JCR *jcr)
+{
+ if (jcr) {
+ free_jcr(jcr);
+ }
+}
+
+
+/*
+ * Setup to call the timeout check routine every 30 seconds
+ * This routine will check any timers that have been enabled.
+ */
bool init_jcr_subsystem(void)
{
watchdog_t *wd = new_watchdog();
wd->one_shot = false;
- wd->interval = 30; /* FIXME: should be configurable somewhere, even
- if only with a #define */
+ wd->interval = 30; /* FIXME: should be configurable somewhere, even
+ if only with a #define */
wd->callback = jcr_timeout_check;
register_watchdog(wd);
* blocked for more than specified max time.
*/
foreach_jcr(jcr) {
+ Dmsg2(3400, "jcr_timeout_check JobId=%u jcr=0x%x\n", jcr->JobId, jcr);
if (jcr->JobId == 0) {
- free_jcr(jcr);
- continue;
+ 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->timed_out = TRUE;
- Jmsg(jcr, M_ERROR, 0, _(
+ timer_start = fd->timer_start;
+ if (timer_start && (watchdog_time - timer_start) > fd->timeout) {
+ fd->timer_start = 0; /* turn off timer */
+ fd->timed_out = true;
+ Jmsg(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);
- }
+ 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->timed_out = TRUE;
- Jmsg(jcr, M_ERROR, 0, _(
+ timer_start = fd->timer_start;
+ if (timer_start && (watchdog_time - timer_start) > fd->timeout) {
+ fd->timer_start = 0; /* turn off timer */
+ fd->timed_out = true;
+ Jmsg(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);
- }
+ 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->timed_out = TRUE;
- Jmsg(jcr, M_ERROR, 0, _(
+ timer_start = fd->timer_start;
+ if (timer_start && (watchdog_time - timer_start) > fd->timeout) {
+ fd->timer_start = 0; /* turn off timer */
+ fd->timed_out = true;
+ Jmsg(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);
- }
+ watchdog_time - timer_start);
+ pthread_kill(jcr->my_thread_id, TIMEOUT_SIGNAL);
+ }
}
- free_jcr(jcr);
}
+ endeach_jcr(jcr);
Dmsg0(3400, "Finished JCR timeout checks\n");
}
*/
extern "C" void timeout_handler(int sig)
{
- return; /* thus interrupting the function */
+ return; /* thus interrupting the function */
}