Index: src/dird/backup.c
===================================================================
---- src/dird/backup.c (revision 7725)
-+++ src/dird/backup.c (working copy)
-@@ -193,6 +193,7 @@
- Jmsg(jcr, M_INFO, 0, _("Start Backup JobId %s, Job=%s\n"),
- edit_uint64(jcr->JobId, ed1), jcr->Job);
+--- src/dird/backup.c (révision 7736)
++++ src/dird/backup.c (copie de travail)
+@@ -326,24 +326,26 @@
+ }
+ return false;
-+ jcr->set_owner(); /* we are responsible for this job */
- set_jcr_job_status(jcr, JS_Running);
- Dmsg2(100, "JobId=%d JobLevel=%c\n", jcr->jr.JobId, jcr->jr.JobLevel);
- if (!db_update_job_start_record(jcr, jcr->db, &jcr->jr)) {
-@@ -361,8 +362,10 @@
- if (timeout) {
- tid = start_bsock_timer(fd, timeout); /* TODO: New timeout directive??? */
- }
-+ Dmsg0(1, "======== Wait for a message\n");
- /* Wait for Client to terminate */
- while ((n = bget_dirmsg(fd)) >= 0) {
-+ Dmsg1(1, "======= Get client message=%s\n", fd->msg);
- if (!fd_ok &&
- (sscanf(fd->msg, EndJob, &jcr->FDJobStatus, &JobFiles,
- &ReadBytes, &JobBytes, &Errors, &VSS, &Encrypt) == 7 ||
-Index: src/dird/job.c
-===================================================================
---- src/dird/job.c (revision 7731)
-+++ src/dird/job.c (working copy)
-@@ -390,6 +390,14 @@
- jobq_remove(&job_queue, jcr); /* attempt to remove it from queue */
- return true;
-
-+ case JS_Running:
-+ if (jcr->get_JobType() == JT_BACKUP && jcr->file_bsock) {
-+ /* When in JS_Running state, the main thread can wait for
-+ * EndJob message from the Client. We send a signal to the job
-+ * thread to cancel the read()
-+ */
-+ jcr->send_signal_to_owner(TIMEOUT_SIGNAL);
-+ }
- default:
- /* Cancel File daemon */
- if (jcr->file_bsock) {
-Index: src/jcr.h
-===================================================================
---- src/jcr.h (revision 7725)
-+++ src/jcr.h (working copy)
-@@ -180,6 +180,8 @@
- bool is_job_canceled() {return job_canceled(this); };
- int32_t get_JobType() { return m_JobType; };
- int32_t get_JobLevel() { return m_JobLevel; };
-+ void set_owner();
-+ int send_signal_to_owner(int signal);
-
- const char *get_OperationName(); /* in lib/jcr.c */
- const char *get_ActionName(bool past); /* in lib/jcr.c */
-Index: src/lib/jcr.c
-===================================================================
---- src/lib/jcr.c (revision 7725)
-+++ src/lib/jcr.c (working copy)
-@@ -303,6 +303,31 @@
+-/* Come here only after starting SD thread */
++/* Come here only after starting SD thread
++ * and we don't expect any EndJob message because the
++ * the client don't have recieve the "backup" command.
++ */
+ bail_out:
+ set_jcr_job_status(jcr, JS_ErrorTerminated);
+- Dmsg1(400, "wait for sd. use=%d\n", jcr->use_count());
+- /* Cancel SD */
+- wait_for_job_termination(jcr, FDConnectTimeout);
+- Dmsg1(400, "after wait for sd. use=%d\n", jcr->use_count());
++ Dmsg1(400, "wait for sd and fd. use=%d\n", jcr->use_count());
++ /* Get status from SD and FD */
++ wait_for_job_termination(jcr, false /* don't expect EndJob message*/);
++ Dmsg1(400, "after wait for sd and fd. use=%d\n", jcr->use_count());
+ return false;
}
+-
/*
-+ * Update the my_thread_id variable with the current thread id
-+ */
-+void JCR::set_owner()
-+{
-+ Dmsg1(0, "JCR::set_owner(%p)\n", pthread_self());
-+ this->my_thread_id = pthread_self();
-+}
-+
-+/*
-+ * Send a signal to the JCR thread owner. (to break a system call)
-+ * (Must have use set_owner() before)
-+ */
-+int JCR::send_signal_to_owner(int sig)
-+{
-+ int ret=0;
-+ Dmsg2(0, "Sending TIMEOUT ?? to %p by %p\n", this->my_thread_id, pthread_self());
-+
-+ if (!pthread_equal(this->my_thread_id, pthread_self())) {
-+ Dmsg2(0, "Sending TIMEOUT to %p by %p\n", this->my_thread_id, pthread_self());
-+ ret = pthread_kill(this->my_thread_id, sig);
-+ }
-+ return ret;
-+}
-+
-+/*
- * Push a subroutine address into the job end callback stack
+ * Here we wait for the File daemon to signal termination,
+ * then we wait for the Storage daemon. When both
+ * are done, we return the job status.
+ * Also used by restore.c
*/
- void job_end_push(JCR *jcr, void job_end_cb(JCR *jcr,void *), void *ctx)
-@@ -719,7 +744,7 @@
- bool set_waittime = false;
- int oldJobStatus = jcr->JobStatus;
+-int wait_for_job_termination(JCR *jcr, int timeout)
++int wait_for_job_termination(JCR *jcr, bool expect_EndJob)
+ {
+ int32_t n = 0;
+ BSOCK *fd = jcr->file_bsock;
+@@ -353,40 +355,49 @@
+ uint64_t JobBytes = 0;
+ int VSS = 0;
+ int Encrypt = 0;
+- btimer_t *tid=NULL;
+-
+ set_jcr_job_status(jcr, JS_Running);
-- Dmsg2(800, "set_jcr_job_status(%s, %c)\n", jcr->Job, JobStatus);
-+ Dmsg2(2, "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:
-@@ -740,7 +765,7 @@
- * 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,
-+ Dmsg3(2, "jid=%u OnEntry JobStatus=%c set=%c\n", (uint32_t)jcr->JobId,
- jcr->JobStatus, JobStatus);
- switch (jcr->JobStatus) {
- case JS_ErrorTerminated:
-@@ -781,7 +806,7 @@
+ if (fd) {
+- if (timeout) {
+- tid = start_bsock_timer(fd, timeout); /* TODO: New timeout directive??? */
+- }
+- /* Wait for Client to terminate */
+- while ((n = bget_dirmsg(fd)) >= 0) {
+- if (!fd_ok &&
+- (sscanf(fd->msg, EndJob, &jcr->FDJobStatus, &JobFiles,
+- &ReadBytes, &JobBytes, &Errors, &VSS, &Encrypt) == 7 ||
+- sscanf(fd->msg, OldEndJob, &jcr->FDJobStatus, &JobFiles,
+- &ReadBytes, &JobBytes, &Errors) == 5)) {
+- fd_ok = true;
+- set_jcr_job_status(jcr, jcr->FDJobStatus);
+- Dmsg1(100, "FDStatus=%c\n", (char)jcr->JobStatus);
+- } else {
+- Jmsg(jcr, M_WARNING, 0, _("Unexpected Client Job message: %s\n"),
+- fd->msg);
++ /* Wait for Client to terminate
++ * In some conditions, the client isn't able to send
++ * any messages and we should not wait for ages
++ */
++ int OK=true;
++ int ret;
++ while (OK && expect_EndJob) {
++
++ /* Even if the job is canceled, we let a chance to FD to
++ * send EndJob message
++ */
++ if (job_canceled(jcr)) {
++ OK=false;
+ }
+- if (job_canceled(jcr)) {
+- break;
++
++ /* wait for data few minutes */
++ ret = fd->wait_data_intr(3*60, 0);
++ if (ret == 1) { /* get data */
++ n = bget_dirmsg(fd);
++ if (n >= 0 &&
++ (sscanf(fd->msg, EndJob, &jcr->FDJobStatus, &JobFiles,
++ &ReadBytes, &JobBytes, &Errors, &VSS, &Encrypt) == 7 ||
++ sscanf(fd->msg, OldEndJob, &jcr->FDJobStatus, &JobFiles,
++ &ReadBytes, &JobBytes, &Errors) == 5)) {
++ fd_ok = true;
++ set_jcr_job_status(jcr, jcr->FDJobStatus);
++ OK=false; /* end of loop */
++ } else {
++ Jmsg(jcr, M_WARNING, 0, _("Unexpected Client Job message: %s\n"),
++ fd->msg);
++ }
++ } /* else get timeout or network error */
++
++ if (is_bnet_error(fd)) {
++ Jmsg(jcr, M_FATAL, 0, _("Network error with FD during %s: ERR=%s\n"),
++ job_type_to_str(jcr->get_JobType()), fd->bstrerror());
++ OK=false;
+ }
}
+- if (tid) {
+- stop_bsock_timer(tid);
+- }
+
+- if (is_bnet_error(fd)) {
+- Jmsg(jcr, M_FATAL, 0, _("Network error with FD during %s: ERR=%s\n"),
+- job_type_to_str(jcr->get_JobType()), fd->bstrerror());
+- }
+ fd->signal(BNET_TERMINATE); /* tell Client we are terminating */
}
- if (oldJobStatus != jcr->JobStatus) {
-- Dmsg3(200, "jid=%u leave set_old_job_status=%c new_set=%c\n", (uint32_t)jcr->JobId,
-+ Dmsg3(2, "jid=%u leave set_old_job_status=%c new_set=%c\n", (uint32_t)jcr->JobId,
- oldJobStatus, JobStatus);
- // generate_plugin_event(jcr, bEventStatusChange, NULL);
- }
+
+Index: src/dird/protos.h
+===================================================================
+--- src/dird/protos.h (révision 7736)
++++ src/dird/protos.h (copie de travail)
+@@ -52,7 +52,7 @@
+ extern bool find_recycled_volume(JCR *jcr, bool InChanger, MEDIA_DBR *mr);
+
+ /* backup.c */
+-extern int wait_for_job_termination(JCR *jcr, int timeout=0);
++extern int wait_for_job_termination(JCR *jcr, bool expect_EndJob=true);
+ extern bool do_backup_init(JCR *jcr);
+ extern bool do_backup(JCR *jcr);
+ extern void backup_cleanup(JCR *jcr, int TermCode);
projects / bacula / bacula / blobdiff