From 08860d511edd47ec732e2ca528a59e1a7ad5cd7a Mon Sep 17 00:00:00 2001 From: Eric Bollengier Date: Thu, 6 Nov 2008 08:59:25 +0000 Subject: [PATCH] ebl update debug after signal git-svn-id: https://bacula.svn.sourceforge.net/svnroot/bacula/trunk@7987 91ce42f0-d328-0410-95d8-f526ca767f89 --- bacula/patches/testing/debug_lock.patch | 98 +++++++++++++++++++------ 1 file changed, 75 insertions(+), 23 deletions(-) diff --git a/bacula/patches/testing/debug_lock.patch b/bacula/patches/testing/debug_lock.patch index 3b1e3fe479..477154d3a4 100644 --- a/bacula/patches/testing/debug_lock.patch +++ b/bacula/patches/testing/debug_lock.patch @@ -1,8 +1,8 @@ Index: src/cats/sql.c =================================================================== ---- src/cats/sql.c (revision 7972) -+++ src/cats/sql.c (working copy) -@@ -311,6 +311,34 @@ +--- src/cats/sql.c (révision 7981) ++++ src/cats/sql.c (copie de travail) +@@ -311,6 +311,36 @@ return mdb->errmsg; } @@ -10,11 +10,14 @@ Index: src/cats/sql.c + if (mdb->allow_transactions) { /* batch connection */ + return; + } -+ if (_db_lock_jcr && _db_lock_jcr != mdb->jcr) { -+ Dmsg2(1, "ERROR: not the same jcr %p != %p", _db_lock_jcr, mdb->jcr); ++ if (_db_lock_recurse_count && !pthread_equal(_db_lock_threadid, pthread_self())) { ++ Dmsg2(1, "ERROR: not the same threadif %p != %p\n", _db_lock_threadid, pthread_self()); + } ++// if (_db_lock_jcr && _db_lock_jcr != mdb->jcr) { ++// Dmsg2(1, "ERROR: not the same jcr %p != %p", _db_lock_jcr, mdb->jcr); ++// } + _db_lock_recurse_count++; -+ _db_lock_jobid = mdb->jcr->JobId; ++// _db_lock_jobid = mdb->jcr->JobId; + _db_lock_time = (utime_t) time(NULL); + _db_lock_threadid = pthread_self(); +} @@ -29,15 +32,14 @@ Index: src/cats/sql.c + _db_lock_recurse_count--; + if (!_db_lock_recurse_count) { + _db_lock_jobid = 0; -+ _db_lock_time = 0; -+ memset(0, _db_lock_threadid, sizeof(_db_lock_threadid)); ++ memset(&_db_lock_threadid, 0, sizeof(_db_lock_threadid)); + } +} + /* * Lock database, this can be called multiple times by the same * thread without blocking, but must be unlocked the number of -@@ -324,6 +352,7 @@ +@@ -324,6 +354,7 @@ e_msg(file, line, M_FATAL, 0, "rwl_writelock failure. stat=%d: ERR=%s\n", errstat, be.bstrerror(errstat)); } @@ -45,7 +47,7 @@ Index: src/cats/sql.c } /* -@@ -334,6 +363,7 @@ +@@ -334,6 +365,7 @@ void _db_unlock(const char *file, int line, B_DB *mdb) { int errstat; @@ -55,9 +57,9 @@ Index: src/cats/sql.c e_msg(file, line, M_FATAL, 0, "rwl_writeunlock failure. stat=%d: ERR=%s\n", Index: src/lib/message.c =================================================================== ---- src/lib/message.c (revision 7972) -+++ src/lib/message.c (working copy) -@@ -62,6 +62,26 @@ +--- src/lib/message.c (révision 7981) ++++ src/lib/message.c (copie de travail) +@@ -62,6 +62,25 @@ FILE *con_fd = NULL; /* Console file descriptor */ brwlock_t con_lock; /* Console lock structure */ @@ -66,7 +68,7 @@ Index: src/lib/message.c + * Global variables to get information about lock/unlock db access + */ +JCR *_db_lock_jcr = NULL; -+JobId_t _db_lock_jobid = 0; ++uint32_t _db_lock_jobid = 0; +utime_t _db_lock_time = 0; +int _db_lock_recurse_count = 0; +pthread_t _db_lock_threadid; @@ -76,18 +78,17 @@ Index: src/lib/message.c + char buf[128]; + bstrutime(buf, sizeof(buf), _db_lock_time); + -+ Dmsg4(1, "lock info: recuse_count=%i jobid=%i time=%s threadid=%p\n", -+ _db_lock_recurse_count, _db_lock_jobid, _db_lock_time, _db_lock_threadid); ++ fprintf(stderr, "lock info: recurse_count=%i jobid=%i threadid=%i time=%s\n", ++ _db_lock_recurse_count, _db_lock_jobid, (int)_db_lock_threadid, buf); +} -+ + /* Forward referenced functions */ /* Imported functions */ Index: src/lib/message.h =================================================================== ---- src/lib/message.h (revision 7972) -+++ src/lib/message.h (working copy) +--- src/lib/message.h (révision 7981) ++++ src/lib/message.h (copie de travail) @@ -163,3 +163,14 @@ extern DLL_IMP_EXP int console_msg_pending; extern DLL_IMP_EXP FILE * con_fd; /* Console file descriptor */ @@ -98,20 +99,71 @@ Index: src/lib/message.h + */ +void print_lock_dbg(); +extern DLL_IMP_EXP JCR * _db_lock_jcr; -+extern DLL_IMP_EXP JobId_t _db_lock_jobid; ++extern DLL_IMP_EXP uint32_t _db_lock_jobid; +extern DLL_IMP_EXP utime_t _db_lock_time; +extern DLL_IMP_EXP int _db_lock_recurse_count; +extern DLL_IMP_EXP pthread_t _db_lock_threadid; + +Index: src/lib/jcr.c +=================================================================== +--- src/lib/jcr.c (révision 7981) ++++ src/lib/jcr.c (copie de travail) +@@ -836,7 +836,37 @@ + V(jcr_lock); + } + ++/* ++ * This function should be used only after receiving a violent signal ++ * We walk through the JCR chain without doing any lock, bacula should ++ * not be running. ++ */ ++void print_jcr_dbg() ++{ ++ char buf1[128], buf2[128], buf3[128], buf4[128]; ++ FILE *fp = stderr; + ++ if (!jcrs) ++ return; ++ ++ for (JCR *jcr = (JCR *)jcrs->first(); jcr ; jcr = (JCR *)jcrs->next(jcr)) { ++ if (!jcr) { ++ continue; ++ } ++ ++ fprintf(fp, "JCR=%p JobId=%i name=%s JobStatus=%c\n", jcr, jcr->JobId, jcr->Job, jcr->JobStatus); ++ fprintf(fp, "\tuse_count=%i threadid=%i\n", jcr->use_count(), (int)jcr->my_thread_id); ++ fprintf(fp, "\tJobType=%i JobLevel=%i\n", jcr->get_JobType(), jcr->get_JobLevel()); ++ bstrftime(buf1, sizeof(buf1), jcr->sched_time); ++ bstrftime(buf2, sizeof(buf2), jcr->start_time); ++ bstrftime(buf3, sizeof(buf3), jcr->end_time); ++ bstrftime(buf4, sizeof(buf4), jcr->wait_time); ++ fprintf(fp, "\tsched_time=%s start_time=%s end_time=%s wait_time=%s\n", buf1, buf2, buf3, buf4); ++ fprintf(fp, "\tdequeing=%i\n", jcr->dequeuing); ++ fprintf(fp, "\tdb=%p db_batch=%p batch_started=%i\n", jcr->db, jcr->db_batch, jcr->batch_started); ++ } ++} ++ + /* + * Start walk of jcr chain + * The proper way to walk the jcr chain is: Index: src/lib/signal.c =================================================================== ---- src/lib/signal.c (revision 7972) -+++ src/lib/signal.c (working copy) -@@ -104,6 +104,8 @@ +--- src/lib/signal.c (révision 7981) ++++ src/lib/signal.c (copie de travail) +@@ -53,6 +53,7 @@ + extern char my_name[]; + extern char *exepath; + extern char *exename; ++extern void print_jcr_dbg(); + + static const char *sig_names[BA_NSIG+1]; + +@@ -104,6 +105,9 @@ pid_t pid; int exelen = strlen(exepath); + print_lock_dbg(); ++ print_jcr_dbg(); + fprintf(stderr, _("Kaboom! %s, %s got signal %d - %s. Attempting traceback.\n"), exename, my_name, sig, get_signal_name(sig)); -- 2.39.5