IPMI: don't init irq until ready

Patrick found a race at startup.  Interrupts were being enabled for the IPMI
interface before the driver was really ready to handle them.  This could
result in an oops if something was pending on the interface at startup and
interrupt were already enabled (technically shouldn't happen, but need to
cover for this in real life).  So move the IRQ setup to the code that starts
the actual IPMI processing.

Signed-off-by: Corey Minyard <cminyard@mvista.com>
Cc: Patrick Schoeller <Patrick.Schoeller@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/char/ipmi/ipmi_si_intf.c

@@ -1006,6 +1006,10 @@ static int smi_start_processing(void       *send_info,
 
 	new_smi->intf = intf;
 
+	/* Try to claim any interrupts. */
+	if (new_smi->irq_setup)
+		new_smi->irq_setup(new_smi);
+
 	/* Set up the timer that drives the interface. */
 	setup_timer(&new_smi->si_timer, smi_timeout, (long)new_smi);
 	new_smi->last_timeout_jiffies = jiffies;
@@ -2765,10 +2769,6 @@ static int try_smi_init(struct smi_info *new_smi)
 	setup_oem_data_handler(new_smi);
 	setup_xaction_handlers(new_smi);
 
-	/* Try to claim any interrupts. */
-	if (new_smi->irq_setup)
-		new_smi->irq_setup(new_smi);
-
 	INIT_LIST_HEAD(&(new_smi->xmit_msgs));
 	INIT_LIST_HEAD(&(new_smi->hp_xmit_msgs));
 	new_smi->curr_msg = NULL;