X-Git-Url: https://git.sur5r.net/?a=blobdiff_plain;f=bacula%2Fsrc%2Fstored%2Fjob.c;h=0b5e6bd4a51e1dc1badec7b6ef5d429786f74a74;hb=cfc2aa18dbf9afb34902a14473152f388836538d;hp=2bd6be9d24943d72dcc08b4a4c1bcfb8da06964d;hpb=22a266f343aa3573413db9089450a87b6e1a766b;p=bacula%2Fbacula

diff --git a/bacula/src/stored/job.c b/bacula/src/stored/job.c
index 2bd6be9d24..0b5e6bd4a5 100644
--- a/bacula/src/stored/job.c
+++ b/bacula/src/stored/job.c
@@ -1,22 +1,14 @@
-/*
- *   Job control and execution for Storage Daemon
- *
- *   Kern Sibbald, MM
- *
- *   Version $Id$
- *
- */
 /*
    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
@@ -33,6 +25,14 @@
    (FSFE), Fiduciary Program, Sumatrastrasse 25, 8006 Zürich,
    Switzerland, email:ftf@fsfeurope.org.
 */
+/*
+ *   Job control and execution for Storage Daemon
+ *
+ *   Kern Sibbald, MM
+ *
+ *   Version $Id$
+ *
+ */
 
 #include "bacula.h"
 #include "stored.h"
@@ -48,10 +48,14 @@ extern bool do_mac(JCR *jcr);
 
 /* Requests from the Director daemon */
 static char jobcmd[] = "JobId=%d job=%127s job_name=%127s client_name=%127s "
+      "type=%d level=%d FileSet=%127s NoAttr=%d SpoolAttr=%d FileSetMD5=%127s "
+      "SpoolData=%d WritePartAfterJob=%d PreferMountedVols=%d SpoolSize=%s\n";
+static char oldjobcmd[] = "JobId=%d job=%127s job_name=%127s client_name=%127s "
       "type=%d level=%d FileSet=%127s NoAttr=%d SpoolAttr=%d FileSetMD5=%127s "
       "SpoolData=%d WritePartAfterJob=%d PreferMountedVols=%d\n";
 
 
+
 /* Responses sent to Director daemon */
 static char OKjob[]     = "3000 OK Job SDid=%u SDtime=%u Authorization=%s\n";
 static char BAD_job[]   = "3915 Bad Job command. stat=%d CMD: %s\n";
@@ -73,6 +77,8 @@ bool job_cmd(JCR *jcr)
 {
    int JobId;
    char auth_key[100];
+   char spool_size[30];
+   char seed[100];
    BSOCK *dir = jcr->dir_bsock;
    POOL_MEM job_name, client_name, job, fileset_name, fileset_md5;
    int JobType, level, spool_attributes, no_attributes, spool_data;
@@ -84,17 +90,26 @@ bool job_cmd(JCR *jcr)
     * Get JobId and permissions from Director
     */
    Dmsg1(100, "<dird: %s", dir->msg);
+   bstrncpy(spool_size, "0", sizeof(spool_size));
    stat = sscanf(dir->msg, jobcmd, &JobId, job.c_str(), job_name.c_str(),
               client_name.c_str(),
               &JobType, &level, fileset_name.c_str(), &no_attributes,
-              &spool_attributes, fileset_md5.c_str(), &spool_data, 
+              &spool_attributes, fileset_md5.c_str(), &spool_data,
+              &write_part_after_job, &PreferMountedVols, spool_size);
+   if (stat != 14) {
+      /* Try old version */
+      stat = sscanf(dir->msg, oldjobcmd, &JobId, job.c_str(), job_name.c_str(),
+              client_name.c_str(),
+              &JobType, &level, fileset_name.c_str(), &no_attributes,
+              &spool_attributes, fileset_md5.c_str(), &spool_data,
               &write_part_after_job, &PreferMountedVols);
-   if (stat != 13) {
-      pm_strcpy(jcr->errmsg, dir->msg);
-      bnet_fsend(dir, BAD_job, stat, jcr->errmsg);
-      Dmsg1(100, ">dird: %s", dir->msg);
-      set_jcr_job_status(jcr, JS_ErrorTerminated);
-      return false;
+      if (stat != 13) {
+         pm_strcpy(jcr->errmsg, dir->msg);
+         dir->fsend(BAD_job, stat, jcr->errmsg);
+         Dmsg1(100, ">dird: %s", dir->msg);
+         set_jcr_job_status(jcr, JS_ErrorTerminated);
+         return false;
+      }
    }
    /*
     * Since this job could be rescheduled, we
@@ -124,6 +139,7 @@ bool job_cmd(JCR *jcr)
    jcr->no_attributes = no_attributes;
    jcr->spool_attributes = spool_attributes;
    jcr->spool_data = spool_data;
+   jcr->spool_size = str_to_int64(spool_size);
    jcr->write_part_after_job = write_part_after_job;
    jcr->fileset_md5 = get_pool_memory(PM_NAME);
    pm_strcpy(jcr->fileset_md5, fileset_md5);
@@ -134,9 +150,10 @@ bool job_cmd(JCR *jcr)
    /*
     * Pass back an authorization key for the File daemon
     */
-   make_session_key(auth_key, NULL, 1);
-   bnet_fsend(dir, OKjob, jcr->VolSessionId, jcr->VolSessionTime, auth_key);
-   Dmsg1(100, ">dird: %s", dir->msg);
+   bsnprintf(seed, sizeof(seed), "%p%d", jcr, JobId);
+   make_session_key(auth_key, seed, 1);
+   dir->fsend(OKjob, jcr->VolSessionId, jcr->VolSessionTime, auth_key);
+   Dmsg2(100, ">dird jid=%u: %s", (uint32_t)jcr->JobId, dir->msg);
    jcr->sd_auth_key = bstrdup(auth_key);
    memset(auth_key, 0, sizeof(auth_key));
    generate_daemon_event(jcr, "JobStart");
@@ -148,13 +165,14 @@ bool run_cmd(JCR *jcr)
    struct timeval tv;
    struct timezone tz;
    struct timespec timeout;
-   int errstat;
+   int errstat = 0;
 
+   Dsm_check(1);
    Dmsg1(200, "Run_cmd: %s\n", jcr->dir_bsock->msg);
    /* The following jobs don't need the FD */
    switch (jcr->JobType) {
-   case JT_MIGRATE:
    case JT_COPY:
+   case JT_MIGRATE:
    case JT_ARCHIVE:
       jcr->authenticated = true;
       do_mac(jcr);
@@ -166,21 +184,25 @@ bool run_cmd(JCR *jcr)
 
    gettimeofday(&tv, &tz);
    timeout.tv_nsec = tv.tv_usec * 1000;
-   timeout.tv_sec = tv.tv_sec + 30 * 60;        /* wait 30 minutes */
+   timeout.tv_sec = tv.tv_sec + me->client_wait;
+
+   Dmsg3(050, "%s waiting %d sec for FD to contact SD key=%s\n",
+         jcr->Job, (int)(timeout.tv_sec-time(NULL)), jcr->sd_auth_key);
 
-   Dmsg1(100, "%s waiting on FD to contact SD\n", jcr->Job);
    /*
     * Wait for the File daemon to contact us to start the Job,
     *  when he does, we will be released, unless the 30 minutes
     *  expires.
     */
    P(mutex);
-   for ( ;!job_canceled(jcr); ) {
+   while ( !jcr->authenticated && !job_canceled(jcr) ) {
       errstat = pthread_cond_timedwait(&jcr->job_start_wait, &mutex, &timeout);
-      if (errstat == 0 || errstat == ETIMEDOUT) {
+      if (errstat == ETIMEDOUT || errstat == EINVAL || errstat == EPERM) {
          break;
       }
    }
+   Dmsg3(100, "Auth=%d canceled=%d errstat=%d\n", jcr->authenticated,
+      job_canceled(jcr), errstat);
    V(mutex);
 
    memset(jcr->sd_auth_key, 0, strlen(jcr->sd_auth_key));
@@ -193,28 +215,32 @@ bool run_cmd(JCR *jcr)
 }
 
 /*
- * After receiving a connection (in job.c) if it is
+ * After receiving a connection (in dircmd.c) if it is
  *   from the File daemon, this routine is called.
  */
 void handle_filed_connection(BSOCK *fd, char *job_name)
 {
    JCR *jcr;
 
-   bmicrosleep(0, 50000);             /* wait 50 millisecs */
+/*
+ * With the following bmicrosleep on, running the 
+ * SD under the debugger fails.   
+ */ 
+// bmicrosleep(0, 50000);             /* wait 50 millisecs */
    if (!(jcr=get_jcr_by_full_name(job_name))) {
-      Jmsg1(NULL, M_FATAL, 0, _("Job name not found: %s\n"), job_name);
-      Dmsg1(100, "Job name not found: %s\n", job_name);
+      Jmsg1(NULL, M_FATAL, 0, _("FD connect failed: Job name not found: %s\n"), job_name);
+      Dmsg1(3, "**** Job \"%s\" not found.\n", job_name);
       return;
    }
 
    jcr->file_bsock = fd;
-   jcr->file_bsock->jcr = jcr;
+   jcr->file_bsock->set_jcr(jcr);
 
    Dmsg1(110, "Found Job %s\n", job_name);
 
    if (jcr->authenticated) {
       Jmsg2(jcr, M_FATAL, 0, _("Hey!!!! JobId %u Job %s already authenticated.\n"),
-         jcr->JobId, jcr->Job);
+         (uint32_t)jcr->JobId, jcr->Job);
       free_jcr(jcr);
       return;
    }
@@ -227,7 +253,7 @@ void handle_filed_connection(BSOCK *fd, char *job_name)
       Jmsg(jcr, M_FATAL, 0, _("Unable to authenticate File daemon\n"));
    } else {
       jcr->authenticated = true;
-      Dmsg1(110, "OK Authentication Job %s\n", jcr->Job);
+      Dmsg2(110, "OK Authentication jid=%u Job %s\n", (uint32_t)jcr->JobId, jcr->Job);
    }
 
    if (!jcr->authenticated) {
@@ -263,7 +289,7 @@ bool query_cmd(JCR *jcr)
       unbash_spaces(dev_name);
       foreach_res(device, R_DEVICE) {
          /* Find resource, and make sure we were able to open it */
-         if (fnmatch(dev_name.c_str(), device->hdr.name, 0) == 0) {
+         if (strcmp(dev_name.c_str(), device->hdr.name) == 0) {
             if (!device->dev) {
                device->dev = init_dev(jcr, device);
             }
@@ -272,24 +298,24 @@ bool query_cmd(JCR *jcr)
             }  
             ok = dir_update_device(jcr, device->dev);
             if (ok) {
-               ok = bnet_fsend(dir, OK_query);
+               ok = dir->fsend(OK_query);
             } else {
-               bnet_fsend(dir, NO_query);
+               dir->fsend(NO_query);
             }
             return ok;
          }
       }
       foreach_res(changer, R_AUTOCHANGER) {
          /* Find resource, and make sure we were able to open it */
-         if (fnmatch(dev_name.c_str(), changer->hdr.name, 0) == 0) {
+         if (strcmp(dev_name.c_str(), changer->hdr.name) == 0) {
             if (!changer->device || changer->device->size() == 0) {
                continue;              /* no devices */
             }
             ok = dir_update_changer(jcr, changer);
             if (ok) {
-               ok = bnet_fsend(dir, OK_query);
+               ok = dir->fsend(OK_query);
             } else {
-               bnet_fsend(dir, NO_query);
+               dir->fsend(NO_query);
             }
             return ok;
          }
@@ -297,12 +323,12 @@ bool query_cmd(JCR *jcr)
       /* If we get here, the device/autochanger was not found */
       unbash_spaces(dir->msg);
       pm_strcpy(jcr->errmsg, dir->msg);
-      bnet_fsend(dir, NO_device, dev_name.c_str());
+      dir->fsend(NO_device, dev_name.c_str());
       Dmsg1(100, ">dird: %s\n", dir->msg);
    } else {
       unbash_spaces(dir->msg);
       pm_strcpy(jcr->errmsg, dir->msg);
-      bnet_fsend(dir, BAD_query, jcr->errmsg);
+      dir->fsend(BAD_query, jcr->errmsg);
       Dmsg1(100, ">dird: %s\n", dir->msg);
    }
 
@@ -320,7 +346,7 @@ void stored_free_jcr(JCR *jcr)
 {
    Dmsg1(900, "stored_free_jcr JobId=%u\n", jcr->JobId);
    if (jcr->file_bsock) {
-      bnet_close(jcr->file_bsock);
+      jcr->file_bsock->close();
       jcr->file_bsock = NULL;
    }
    if (jcr->job_name) {
@@ -354,6 +380,10 @@ void stored_free_jcr(JCR *jcr)
    }
    jcr->dcrs = NULL;
 
+   /* Avoid a double free */
+   if (jcr->dcr == jcr->read_dcr) {
+      jcr->read_dcr = NULL;
+   }
    if (jcr->dcr) {
       free_dcr(jcr->dcr);
       jcr->dcr = NULL;
@@ -381,5 +411,10 @@ void stored_free_jcr(JCR *jcr)
       delete jcr->write_store;
       jcr->write_store = NULL;
    }
+   Dsm_check(1);
+
+   if (jcr->JobId != 0)
+      write_state_file(me->working_directory, "bacula-sd", get_first_port_host_order(me->sdaddrs));
+
    return;
 }