Better cancel + fix Allow Dups code

diff --git a/bacula/src/dird/getmsg.c b/bacula/src/dird/getmsg.c
index 60e7aed439ae9c23daa3b4c5539e7996c9372381..1dd0687e6a423939186582a0d1aaab8b6932666c 100644 (file)
--- a/bacula/src/dird/getmsg.c
+++ b/bacula/src/dird/getmsg.c
@@ -125,7 +125,7 @@ void set_jcr_sd_job_status(JCR *jcr, int SDJobStatus)
  */
 int bget_dirmsg(BSOCK *bs)
 {
-   int32_t n;
+   int32_t n = BNET_TERMINATE;
    char Job[MAX_NAME_LENGTH];
    char MsgType[20];
    int type;
@@ -133,11 +133,11 @@ int bget_dirmsg(BSOCK *bs)
    JCR *jcr = bs->jcr();
    char *msg;
 
-   for (;;) {
+   for ( ; !bs->is_stop() && !bs->is_timed_out(); ) {
       n = bs->recv();
       Dmsg2(100, "bget_dirmsg %d: %s\n", n, bs->msg);
 
-      if (is_bnet_stop(bs)) {
+      if (bs->is_stop() || bs->is_timed_out()) {
          return n;                    /* error or terminate */
       }
       if (n == BNET_SIGNAL) {          /* handle signal */
@@ -328,6 +328,7 @@ int bget_dirmsg(BSOCK *bs)
 #endif
       return n;
    }
+   return n;
 }
 
 static char *find_msg_start(char *msg)

diff --git a/bacula/src/dird/job.c b/bacula/src/dird/job.c
index de75ce7591ce7fe9afd30b03dbbffee524d1f0c4..80e8744c87128a361028ee995253206ef86e2c96 100644 (file)
--- a/bacula/src/dird/job.c
+++ b/bacula/src/dird/job.c
@@ -1,7 +1,7 @@
 /*
    Bacula® - The Network Backup Solution
 
-   Copyright (C) 2000-2009 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.
@@ -31,7 +31,6 @@
  *
  *     Kern Sibbald, October MM
  *
- *    Version $Id$
  */
 
 #include "bacula.h"
@@ -415,6 +414,11 @@ bool cancel_job(UAContext *ua, JCR *jcr)
          fd->signal(BNET_TERMINATE);
          fd->close();
          ua->jcr->file_bsock = NULL;
+         jcr->file_bsock->set_terminated();
+         if (jcr->my_thread_id) {
+            pthread_kill(jcr->my_thread_id, TIMEOUT_SIGNAL);
+            Dmsg1(800, "Send kill to jid=%d\n", jcr->JobId);
+         }
       }
 
       /* Cancel Storage daemon */
@@ -448,6 +452,10 @@ bool cancel_job(UAContext *ua, JCR *jcr)
          sd->signal(BNET_TERMINATE);
          sd->close();
          ua->jcr->store_bsock = NULL;
+         jcr->store_bsock->set_terminated();
+         if (jcr->my_thread_id) {
+            pthread_kill(jcr->my_thread_id, TIMEOUT_SIGNAL);
+         }
       }
       break;
    }
@@ -683,6 +691,7 @@ bool allow_duplicate_job(JCR *jcr)
    if (job->AllowDuplicateJobs) {
       return true;
    }
+   Dmsg0(800, "Enter allow_duplicate_job\n");
    /*
     * After this point, we do not want to allow any duplicate
     * job to run.
@@ -694,7 +703,7 @@ bool allow_duplicate_job(JCR *jcr)
       }
       if (strcmp(job->name(), djcr->job->name()) == 0) {
          bool cancel_dup = false;
-         bool cancel_me = false;
+         bool cancel_me = false; 
          if (job->DuplicateJobProximity > 0) {
             utime_t now = (utime_t)time(NULL);
             if ((now - djcr->start_time) > job->DuplicateJobProximity) {
@@ -753,17 +762,23 @@ bool allow_duplicate_job(JCR *jcr)
          }
          if (cancel_dup || job->CancelRunningDuplicates) {
             /* Zap the duplicated job djcr */
-            UAContext *ua = new_ua_context(djcr);
+            UAContext *ua = new_ua_context(jcr);
             Jmsg(jcr, M_INFO, 0, _("Cancelling duplicate JobId=%d.\n"), djcr->JobId);
-            ua->jcr = djcr;
             cancel_job(ua, djcr);
             free_ua_context(ua);
-            Dmsg2(800, "Have cancelled JCR %p JobId=%d\n", djcr, djcr->JobId);
+            if (djcr->my_thread_id) {
+               pthread_kill(djcr->my_thread_id, TIMEOUT_SIGNAL);
+               Dmsg1(800, "Send kill to jid=%d\n", djcr->JobId);
+            }
+            Dmsg2(800, "Cancel dup %p JobId=%d\n", djcr, djcr->JobId);
          } else {
             /* Zap current job */
             Jmsg(jcr, M_FATAL, 0, _("JobId %d already running. Duplicate job not allowed.\n"),
                djcr->JobId);
+            Dmsg2(800, "Cancel me %p JobId=%d\n", jcr, jcr->JobId);
          }
+         Dmsg4(800, "curJobId=%d use_cnt=%d dupJobId=%d use_cnt=%d\n",
+               jcr->JobId, jcr->use_count(), djcr->JobId, djcr->use_count());
          break;                 /* did our work, get out of foreach loop */
       }
    }

diff --git a/bacula/src/lib/jcr.c b/bacula/src/lib/jcr.c
index 7bf2ac08e59d5983999fb7b700ccf647209d2f2f..4e5b4c06e82c22d1b51cccdb0e34bd87b44a628b 100644 (file)
--- a/bacula/src/lib/jcr.c
+++ b/bacula/src/lib/jcr.c
@@ -1,7 +1,7 @@
 /*
    Bacula® - The Network Backup Solution
 
-   Copyright (C) 2000-2009 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.
@@ -31,8 +31,6 @@
  *
  *  Kern E. Sibbald, December 2000
  *
- *  Version $Id$
- *
  *  These routines are thread safe.
  *
  *  The job list routines were re-written in May 2005 to

diff --git a/bacula/src/stored/dircmd.c b/bacula/src/stored/dircmd.c
index 147a1da1397e963845169f0d8c7dd2d5440cfae8..72da68d8e08a1824ca83f2ff1b3b9137c233d755 100644 (file)
--- a/bacula/src/stored/dircmd.c
+++ b/bacula/src/stored/dircmd.c
@@ -1,7 +1,7 @@
 /*
    Bacula® - The Network Backup Solution
 
-   Copyright (C) 2001-2009 Free Software Foundation Europe e.V.
+   Copyright (C) 2001-2010 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.
@@ -43,8 +43,6 @@
  *
  *     Kern Sibbald, May MMI
  *
- *   Version $Id$
- *
  */
 
 #include "bacula.h"
@@ -58,7 +56,7 @@ extern struct s_last_job last_job;
 extern bool init_done;
 
 /* Static variables */
-static char derrmsg[]     = "3900 Invalid command\n";
+static char derrmsg[]     = "3900 Invalid command:";
 static char OKsetdebug[]  = "3000 OK setdebug=%d\n";
 static char invalid_cmd[] = "3997 Invalid command for a Director with Monitor directive enabled.\n";
 static char OK_bootstrap[]    = "3000 OK bootstrap\n";
@@ -241,7 +239,9 @@ void *handle_connection_request(void *arg)
         }
       }
       if (!found) {                   /* command not found */
-        bs->fsend(derrmsg);
+        POOL_MEM err_msg;
+        Mmsg(err_msg, "%s %s\n", derrmsg, bs->msg);
+        bs->fsend(err_msg.c_str());
         break;
       }
    }
@@ -316,14 +316,19 @@ static bool cancel_cmd(JCR *cjcr)
       } else {
          oldStatus = jcr->JobStatus;
          set_jcr_job_status(jcr, JS_Canceled);
+         Dmsg2(800, "Cancel JobId=%d %p\n", jcr->JobId, jcr);
          if (!jcr->authenticated && oldStatus == JS_WaitFD) {
             pthread_cond_signal(&jcr->job_start_wait); /* wake waiting thread */
          }
          if (jcr->file_bsock) {
-            bnet_sig(jcr->file_bsock, BNET_TERMINATE);
+            jcr->file_bsock->signal(BNET_TERMINATE);
+            jcr->file_bsock->set_terminated();
+            Dmsg2(800, "Term bsock jid=%d %p\n", jcr->JobId, jcr);
          } else {
             /* Still waiting for FD to connect, release it */
+            bmicrosleep(0, 50000);
             pthread_cond_signal(&jcr->job_start_wait); /* wake waiting job */
+            Dmsg2(800, "Signal FD connect jid=%d %p\n", jcr->JobId, jcr);
          }
          /* If thread waiting on mount, wake him */
          if (jcr->dcr && jcr->dcr->dev && jcr->dcr->dev->waiting_for_mount()) {