Skip to contentSkip to conversation

EDAC, sb_edac: Early return on ADDRV bit and address type test

PATCH New Luck, Tony 1 participant 1 message Sep 4, 2018

Subsystems (2)
EDAC-core EDAC-sbridge
lore
Share & tools
lore

Luck, Tony

author maintainer
#0 lore raw

PATCH 2/3 EDAC, sb_edac: Early return on ADDRV bit and address type test

From: Qiuxu Zhuo <[email protected]>

Current EDAC driver only decodes valid memory physical addresses. So move the test for bit 58(ADDRV) into sbridge_mce_check_error(), and add a test that the address is the right type (looking as MISC bits{8:6}). Get rid of the duplicated test whether it's a memory controller error in sbridge_mce_output_error().

Signed-off-by Qiuxu Zhuo <[email protected]>
Signed-off-by Tony Luck <[email protected]>
---
 drivers/edac/sb_edac.c | 68 ++++++++++++++++++++++--------------------
 1 file changed, 35 insertions(+), 33 deletions(-)
drivers/edac/sb_edac.c +35 -33
index 07726fb00321..f3678cdada83 100644
@@ -2911,35 +2911,27 @@ static void sbridge_mce_output_error(struct mem_ctl_info *mci,
 	 *	cccc = channel
 	 * If the mask doesn't match, report an error to the parsing logic
 	 */
-	if (! ((errcode & 0xef80) == 0x80)) {
-		optype = "Can't parse: it is not a mem";
-	} else {
-		switch (optypenum) {
-		case 0:
-			optype = "generic undef request error";
-			break;
-		case 1:
-			optype = "memory read error";
-			break;
-		case 2:
-			optype = "memory write error";
-			break;
-		case 3:
-			optype = "addr/cmd error";
-			break;
-		case 4:
-			optype = "memory scrubbing error";
-			break;
-		default:
-			optype = "reserved";
-			break;
-		}
+	switch (optypenum) {
+	case 0:
+		optype = "generic undef request error";
+		break;
+	case 1:
+		optype = "memory read error";
+		break;
+	case 2:
+		optype = "memory write error";
+		break;
+	case 3:
+		optype = "addr/cmd error";
+		break;
+	case 4:
+		optype = "memory scrubbing error";
+		break;
+	default:
+		optype = "reserved";
+		break;
 	}
 
-	/* Only decode errors with an valid address (ADDRV) */
-	if (!GET_BITFIELD(m->status, 58, 58))
-		return;
-
 	if (pvt->info.type == KNIGHTS_LANDING) {
 		if (channel == 14) {
 			edac_dbg(0, "%s%s err_code:%04x:%04x EDRAM bank %d\n",

@@ -3045,17 +3037,11 @@ static int sbridge_mce_check_error(struct notifier_block *nb, unsigned long val,
 {
 	struct mce *mce = (struct mce *)data;
 	struct mem_ctl_info *mci;
-	struct sbridge_pvt *pvt;
 	char *type;
 
 	if (edac_get_report_status() == EDAC_REPORTING_DISABLED)
 		return NOTIFY_DONE;
 
-	mci = get_mci_for_node_id(mce->socketid, IMC0);
-	if (!mci)
-		return NOTIFY_DONE;
-	pvt = mci->pvt_info;
-
 	/*
 	 * Just let mcelog handle it if the error is
 	 * outside the memory controller. A memory error

@@ -3065,6 +3051,22 @@ static int sbridge_mce_check_error(struct notifier_block *nb, unsigned long val,
 	if ((mce->status & 0xefff) >> 7 != 1)
 		return NOTIFY_DONE;
 
+	/* Check ADDRV bit in STATUS */
+	if (!GET_BITFIELD(mce->status, 58, 58))
+		return NOTIFY_DONE;
+
+	/* Check MISCV bit in STATUS */
+	if (!GET_BITFIELD(mce->status, 59, 59))
+		return NOTIFY_DONE;
+
+	/* Check address type in MISC (physical address only) */
+	if (GET_BITFIELD(mce->misc, 6, 8) != 2)
+		return NOTIFY_DONE;
+
+	mci = get_mci_for_node_id(mce->socketid, IMC0);
+	if (!mci)
+		return NOTIFY_DONE;
+
 	if (mce->mcgstatus & MCG_STATUS_MCIP)
 		type = "Exception";
 	else