/* Forward referenced subroutines */
static void terminate_dird(int sig);
static int check_resources();
-static void reload_config(int sig);
/* Exported subroutines */
+extern "C" void reload_config(int sig);
+
/* Imported subroutines */
JCR *wait_for_next_job(char *runjob);
void backup_cleanup(void);
void start_UA_server(char *addr, int port);
void init_job_server(int max_workers);
-void store_jobtype(LEX *lc, struct res_items *item, int index, int pass);
-void store_level(LEX *lc, struct res_items *item, int index, int pass);
-void store_replace(LEX *lc, struct res_items *item, int index, int pass);
+void store_jobtype(LEX *lc, RES_ITEM *item, int index, int pass);
+void store_level(LEX *lc, RES_ITEM *item, int index, int pass);
+void store_replace(LEX *lc, RES_ITEM *item, int index, int pass);
static char *configfile = NULL;
static char *runjob = NULL;
static int background = 1;
+static void init_reload(void);
/* Globals Exported */
DIRRES *director; /* Director resource */
/* Globals Imported */
extern int r_first, r_last; /* first and last resources */
-extern struct res_items job_items[];
+extern RES_TABLE resources[];
+extern RES **res_head;
+extern RES_ITEM job_items[];
extern URES res_all;
init_stack_dump();
my_name_is(argc, argv, "bacula-dir");
- textdomain("bacula-dir");
+ textdomain("bacula");
init_msg(NULL, NULL); /* initialize message handler */
+ init_reload();
daemon_start_time = time(NULL);
while ((ch = getopt(argc, argv, "c:d:fg:r:stu:v?")) != -1) {
/* Create pid must come after we are a daemon -- so we have our final pid */
create_pid_file(director->pid_directory, "bacula-dir", director->DIRport);
+ read_state_file(director->working_directory, "bacula-dir", director->DIRport);
drop(uid, gid); /* reduce priveleges if requested */
-/* signal(SIGHUP, reload_config); */
+ signal(SIGHUP, reload_config);
init_console_msg(working_directory);
exit(1);
}
already_here = TRUE;
- delete_pid_file(director->pid_directory, "bacula-dir",
- director->DIRport);
+ write_state_file(director->working_directory, "bacula-dir", director->DIRport);
+ delete_pid_file(director->pid_directory, "bacula-dir", director->DIRport);
// signal(SIGCHLD, SIG_IGN); /* don't worry about children now */
term_scheduler();
if (runjob) {
term_msg(); /* terminate message handler */
stop_watchdog();
close_memory_pool(); /* release free memory in pool */
- sm_dump(False);
- exit(sig != 0);
+ sm_dump(false);
+ exit(sig);
+}
+
+struct RELOAD_TABLE {
+ int job_count;
+ RES **res_table;
+};
+
+static const int max_reloads = 10;
+static RELOAD_TABLE reload_table[max_reloads];
+
+static void init_reload(void)
+{
+ for (int i=0; i < max_reloads; i++) {
+ reload_table[i].job_count = 0;
+ reload_table[i].res_table = NULL;
+ }
+}
+
+static void free_saved_resources(int table)
+{
+ int num = r_last - r_first + 1;
+ RES **res_tab = reload_table[table].res_table;
+ if (!res_tab) {
+ Dmsg1(100, "res_tab for table %d already released.\n", table);
+ return;
+ }
+ Dmsg1(100, "Freeing resources for table %d\n", table);
+ for (int j=0; j<num; j++) {
+ free_resource(res_tab[j], r_first + j);
+ }
+ free(res_tab);
+ reload_table[table].job_count = 0;
+ reload_table[table].res_table = NULL;
+}
+
+/*
+ * Called here at the end of every job that was
+ * hooked decrementing the active job_count. When
+ * it goes to zero, no one is using the associated
+ * resource table, so free it.
+ */
+static void reload_job_end_cb(JCR *jcr, void *ctx)
+{
+ int reload_id = (int)ctx;
+ Dmsg3(100, "reload job_end JobId=%d table=%d cnt=%d\n", jcr->JobId,
+ reload_id, reload_table[reload_id].job_count);
+ lock_jcr_chain();
+ LockRes();
+ if (--reload_table[reload_id].job_count <= 0) {
+ free_saved_resources(reload_id);
+ }
+ UnlockRes();
+ unlock_jcr_chain();
+}
+
+static int find_free_reload_table_entry()
+{
+ int table = -1;
+ for (int i=0; i < max_reloads; i++) {
+ if (reload_table[i].res_table == NULL) {
+ table = i;
+ break;
+ }
+ }
+ return table;
}
/*
* If we get here, we have received a SIGHUP, which means to
- * reread our configuration file.
+ * reread our configuration file.
*
- * ***FIXME*** Check that there are no jobs running before
- * doing this.
+ * The algorithm used is as follows: we count how many jobs are
+ * running and mark the running jobs to make a callback on
+ * exiting. The old config is saved with the reload table
+ * id in a reload table. The new config file is read. Now, as
+ * each job exits, it calls back to the reload_job_end_cb(), which
+ * decrements the count of open jobs for the given reload table.
+ * When the count goes to zero, we release those resources.
+ * This allows us to have pointers into the resource table (from
+ * jobs), and once they exit and all the pointers are released, we
+ * release the old table. Note, if no new jobs are running since the
+ * last reload, then the old resources will be immediately release.
+ * A console is considered a job because it may have pointers to
+ * resources, but a SYSTEM job is not since it *should* not have any
+ * permanent pointers to jobs.
*/
-static void reload_config(int sig)
+extern "C"
+void reload_config(int sig)
{
- static int already_here = FALSE;
+ static bool already_here = false;
sigset_t set;
+ JCR *jcr;
+ int njobs = 0; /* number of running jobs */
+ int table, rtable;
if (already_here) {
abort(); /* Oops, recursion -> die */
}
- already_here = TRUE;
- sigfillset(&set);
+ already_here = true;
+ sigemptyset(&set);
+ sigaddset(&set, SIGHUP);
sigprocmask(SIG_BLOCK, &set, NULL);
- free_config_resources();
+// Jmsg(NULL, M_INFO, 0, "Entering experimental reload config code. Bug reports will not be accepted.\n");
+
+ lock_jcr_chain();
+ LockRes();
+
+ table = find_free_reload_table_entry();
+ if (table < 0) {
+ Jmsg(NULL, M_ERROR, 0, _("Too many open reload requests. Request ignored.\n"));
+ goto bail_out;
+ }
+
+ Dmsg1(100, "Reload_config njobs=%d\n", njobs);
+ reload_table[table].res_table = save_config_resources();
+ Dmsg1(100, "Saved old config in table %d\n", table);
parse_config(configfile);
- Dmsg0(200, "check_resources()\n");
+ Dmsg0(100, "Reloaded config file\n");
if (!check_resources()) {
- Jmsg(NULL, M_ERROR_TERM, 0, _("Please correct configuration file: %s\n"), configfile);
+ rtable = find_free_reload_table_entry(); /* save new, bad table */
+ if (rtable < 0) {
+ Jmsg(NULL, M_ERROR, 0, _("Please correct configuration file: %s\n"), configfile);
+ Jmsg(NULL, M_ERROR_TERM, 0, _("Out of reload table entries. Giving up.\n"));
+ } else {
+ Jmsg(NULL, M_ERROR, 0, _("Please correct configuration file: %s\n"), configfile);
+ }
+ reload_table[rtable].res_table = save_config_resources();
+ /* Now restore old resoure values */
+ int num = r_last - r_first + 1;
+ RES **res_tab = reload_table[table].res_table;
+ for (int i=0; i<num; i++) {
+ res_head[i] = res_tab[i];
+ }
+ table = rtable; /* release new, bad, saved table below */
+ } else {
+ /*
+ * Hook all active jobs so that they release this table
+ */
+ foreach_jcr(jcr) {
+ if (jcr->JobType != JT_SYSTEM) {
+ reload_table[table].job_count++;
+ job_end_push(jcr, reload_job_end_cb, (void *)table);
+ njobs++;
+ }
+ free_locked_jcr(jcr);
+ }
}
/* Reset globals */
set_working_directory(director->working_directory);
FDConnectTimeout = director->FDConnectTimeout;
SDConnectTimeout = director->SDConnectTimeout;
-
+ Dmsg0(0, "Director's configuration file reread.\n");
+
+ /* Now release saved resources, if no jobs using the resources */
+ if (njobs == 0) {
+ free_saved_resources(table);
+ }
+
+bail_out:
+ UnlockRes();
+ unlock_jcr_chain();
sigprocmask(SIG_UNBLOCK, &set, NULL);
signal(SIGHUP, reload_config);
- already_here = FALSE;
- Dmsg0(0, "Director's configuration file reread.\n");
+ already_here = false;
}
/*
job->hdr.name, job_items[i].name, *def_svalue, i, offset);
svalue = (char **)((char *)job + offset);
if (*svalue) {
- Dmsg1(000, "Hey something is wrong. p=0x%u\n", (unsigned)*svalue);
+ Pmsg1(000, "Hey something is wrong. p=0x%lu\n", *svalue);
}
*svalue = bstrdup(*def_svalue);
set_bit(i, job->hdr.item_present);
job->hdr.name, job_items[i].name, i, offset);
svalue = (char **)((char *)job + offset);
if (*svalue) {
- Dmsg1(000, "Hey something is wrong. p=0x%u\n", (unsigned)*svalue);
+ Pmsg1(000, "Hey something is wrong. p=0x%lu\n", *svalue);
}
*svalue = *def_svalue;
set_bit(i, job->hdr.item_present);
projects / bacula / bacula / blobdiff