From: Jack Steiner <steiner@sgi.com>
If the GRU malfunctions, print a message instead of panicing the system.
This simplifies debugging since some of the debug tools can be used on
a live system. Flush the cache on instruction timeouts in case the
malfunction is related to a coherency issue (never seen this but I'm paranoid).
Signed-off-by: Jack Steiner <steiner@sgi.com>
---
drivers/misc/sgi-gru/gruhandles.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
Index: linux/drivers/misc/sgi-gru/gruhandles.c
===================================================================
--- linux.orig/drivers/misc/sgi-gru/gruhandles.c 2010-06-09 08:11:43.724081727 -0500
+++ linux/drivers/misc/sgi-gru/gruhandles.c 2010-06-09 08:11:46.697237522 -0500
@@ -71,7 +71,7 @@ static void report_instruction_timeout(v
else if (TYPE_IS(TFH, goff))
id = "TFH";
- panic(KERN_ALERT "GRU %p (%s) is malfunctioning\n", h, id);
+ printk(KERN_ALERT "GRU:%d %p (%s) is malfunctioning\n", smp_processor_id(), h, id);
}
static int wait_instruction_complete(void *h, enum mcs_op opc)
@@ -85,6 +85,7 @@ static int wait_instruction_complete(voi
if (status != CCHSTATUS_ACTIVE)
break;
if (GRU_OPERATION_TIMEOUT < (get_cycles() - start_time)) {
+ gru_flush_cache(h);
report_instruction_timeout(h);
start_time = get_cycles();
}
--
