[PATCH 1/3] rasdaemon: Decode bank_type based on InstanceIdHi in MCA_IPID.

From: Muralidhara M K
Date: Tue Jul 04 2023 - 06:58:53 EST


From: Muralidhara M K <muralidhara.mk@xxxxxxx>

On some AMD systems, InstanceIdHi bits b'47:44 of MCA_IPID register
are Reserved, but the same bits are defined on viz. Genoa. So bank type
is erroneously decoded.

Incorrect bank_type which is socket_id value is observed from the
register InstanceIdHi field.
Handle these bit fields appropriately to get correct bank_type.

Signed-off-by: Muralidhara M K <muralidhara.mk@xxxxxxx>
Tested-by: Gupta Akshay <akshay.gupta@xxxxxxx>
---
mce-amd-smca.c | 9 +++++++++
1 file changed, 9 insertions(+)

diff --git a/mce-amd-smca.c b/mce-amd-smca.c
index f69b555..4a2e645 100644
--- a/mce-amd-smca.c
+++ b/mce-amd-smca.c
@@ -790,6 +790,10 @@ static void decode_smca_error(struct mce_event *e, struct mce_priv* m)
const struct smca_hwid *s_hwid;
uint32_t mcatype_hwid = EXTRACT(e->ipid, 32, 63);
uint8_t mcatype_instancehi = EXTRACT(e->ipid, 44, 47);
+ /* MCA_IPID[43:32] HardwareID of IP associated with MCA bank */
+ uint16_t ipid_hwid = EXTRACT(e->ipid, 32, 43);
+ /* MCA_IPID[63:48] McaType of the MCA bank within the IP */
+ uint16_t ipid_mcatype = EXTRACT(e->ipid, 48, 63);
unsigned int csrow = -1, channel = -1;
unsigned int i;

@@ -800,6 +804,11 @@ static void decode_smca_error(struct mce_event *e, struct mce_priv* m)
if (mcatype_hwid == s_hwid->mcatype_hwid) {
bank_type = s_hwid->bank_type;
break;
+ } else if ((mcatype_instancehi == e->socketid) &&
+ (ipid_hwid == EXTRACT(s_hwid->mcatype_hwid, 0, 11)) &&
+ (ipid_mcatype == EXTRACT(s_hwid->mcatype_hwid, 16, 31))) {
+ bank_type = s_hwid->bank_type;
+ break;
}
if (mcatype_instancehi >= NONCPU_NODE_INDEX)
bank_type = SMCA_UMC_V2;
--
2.25.1