[Openipmi-developer] [PATCH] Question regarding a change between 1.4.25 and 1.4.26 (still present in 2.0.13)

[Jonathan F. <jon...@gm...>]

Hi,

I have an "issue" I'm able to fix by backtracking a change in OpenIPMI
but I don't understand why it fixes the thing, if possible I would
like to know about that change.

I'm using an older revision of OpenIPMI (1.4.26) and openhpi/libipmi
plugin (2.2.1).

When I generate sensor threshold events I don't see the events in
hpigetevents or hpi_cmd/event enable. But the events are actually
displayed by runing hpiel. I spent some time trying to figure out
where the issue was (openhpi, OpenIPMI, the IPMI driver etc) and
managed to find a change that introduced that behavior (I tried a lot
of hpi/ipmi packages combos from older to latest ones). The change
that introduced that (see the patch to backtrack it) was the commit on
2006-02-28 09:50

2006-02-28 09:50  cminyard

   * ChangeLog (1.527.2.134), lib/chassis.c (1.20.2.2), lib/mc.c
     (1.88.2.22), ChangeLog (1.756), lib/chassis.c (1.22), lib/mc.c
     (1.115): More minor fixes.

2006-02-28 03:54  cminyard

   * lib/: mc.c (1.114), mc.c (1.88.2.21): Missed something in the
     last change.

Any comments on that change? It's still present in 2.0.13. I would
like to know if my IPMC could be doing something wrong? The
combination of the spinlock move and the call to
ipmi_mc_remove_active_handler() introduced the error in my case. By
backtracking a subset of the changes as it was on 2006-02-28 03:54
with that patch the problem goes away.

Regards,

/jonathan

--- CUT HERE ---

Signed-off-by: Jonathan Fournier <jon...@gm...>

--- a/lib/chassis.c
+++ b/lib/chassis.c
@@ -362,13 +362,10 @@ chassis_mc_control_active_handler(ipmi_m
     ipmi_domain_t  *domain = ipmi_mc_get_domain(mc);

     if (active)
   return;

-    ipmi_mc_remove_active_handler(mc,
-             chassis_mc_control_active_handler,
-             control);
     _ipmi_domain_entity_lock(domain);
     _ipmi_entity_get(entity);
     _ipmi_domain_entity_unlock(domain);
     ipmi_control_destroy(control);
     _ipmi_entity_put(entity);
--- a/lib/mc.c
+++ b/lib/mc.c
@@ -1382,12 +1382,12 @@ mc_reread_sel_timeout(void *cb_data, os_
      lock, so just don't start the timer and everything should
      be happy. */
   DEBUG_INFO(info);
   info->processing = 0;
   info->timer_running = 0;
-  ipmi_unlock(info->lock);
     }
+  ipmi_unlock(info->lock);
 }

 typedef struct sel_reread_s
 {
     ipmi_mc_done_cb handler;