[Linux-ha-dev] Hb-2.08/stable: crmd crashes on startup under solaris10/i386

[Andrew Beekhof]

On 5/31/07, Otte, Joerg <joerg.otte at nsn.com> wrote:
>
> I must report anather crash. This time in crmd.
> After a reboot the following crash occures every
> minute again and again.
>
> This crash prevents Heartbeat from starting up.
>
> fead4c7c strlen   (807254e, 8047698, 8046230, 0) + c
> feb2d3cb vsnprintf (8046270, 1400, 8072548, 8047698) + 73
> fef885b8 cl_log   (6, 8072548, 806e14d, 0, 80713fa, 8071412) + 58
> 080571af do_state_transition (1000000, 0, 8, 6, 8091b10, 0) + 21b
> 080579d2 s_crmd_fsa (d, 8090c58, 8047738, 805fc1b) + 20e
> 0805fc3d crm_fsa_trigger (0, 0, 8047778, fef9380f) + 2d
> fef93857 G_TRIG_dispatch (8090c58, 0, 0, 0) + a7
> fedbc77f g_main_context_dispatch (808eb80, ffffff9c, 8089d90, 0) + 1e7
> fedbe065 g_main_context_iterate (1, 808c2c0, 8047858, fedbe141, 80554d7, 1) + 41d
> fedbe2c0 g_main_loop_run (8089d78, 806f228, 806ca84, 806f1fd, 0, 0) + 19c
> 080554d7 crmd_init (80478a0, 80553f1, 8088fcc, 8088fdc, 0, 806ca02) + b3
> 08055748 main     (1, 80478d0, 80478d8) + f0
> 080552cc _start   (1, 8047a40, 0, 8047a66, 8047a78, 8047a87) + 80

Is this the best stack trace thats available?
It'd be nice to know which variable is the problem.

But based on the small offset, i _think_ this patch might help.
They're the only variables that I can imagine might be NULL - even
though neither should be.

diff -r b8dc94304c0c crm/crmd/fsa.c
--- a/crm/crmd/fsa.c    Tue May 29 15:03:42 2007 +0200
+++ b/crm/crmd/fsa.c    Fri Jun 01 10:49:07 2007 +0200
@@ -599,9 +599,9 @@ do_state_transition(long long actions,
                  state_from, state_to, input, fsa_cause2string(cause),
                  msg_data->origin);

-       crm_info("%s: State transition %s -> %s [ input=%s cause=%s
origin=%s ]",
-                fsa_our_uname, state_from, state_to, input,
-                fsa_cause2string(cause), msg_data->origin);
+       crm_info("State transition %s -> %s [ input=%s cause=%s origin=%s ]",
+                state_from, state_to, input, fsa_cause2string(cause),
+                msg_data->origin);

        /* the last two clauses might cause trouble later */
        if(election_timeout != NULL
diff -r b8dc94304c0c crm/crmd/messages.c
--- a/crm/crmd/messages.c       Tue May 29 15:03:42 2007 +0200
+++ b/crm/crmd/messages.c       Fri Jun 01 10:48:16 2007 +0200
@@ -90,6 +90,7 @@ register_fsa_input_adv(
        fsa_data_t *fsa_data = NULL;

        last_data_id++;
+       CRM_CHECK(raised_from != NULL, raised_from = "<unknown>");

        crm_debug("%s %s FSA input %d (%s) (cause=%s) %s data",
                  raised_from,
prepend?"prepended":"appended",last_data_id, fsa_input2string(input),