diff mbox series

[1/1] EDAC/i10nm: Add driver decoder for Sapphire Rapids server

Message ID 20230206011423.50559-1-qiuxu.zhuo@intel.com (mailing list archive)
State New, archived
Headers show
Series [1/1] EDAC/i10nm: Add driver decoder for Sapphire Rapids server | expand

Commit Message

Zhuo, Qiuxu Feb. 6, 2023, 1:14 a.m. UTC 
From: Youquan Song <youquan.song@intel.com>

Intel SDM (December 2022) vol3B 17.13.2 contains IMC MC error codes
for Sapphire Rapids. Current i10nm_edac only supports firmware decoder
(ACPI DSM methods) for Sapphire Rapids. So add the driver decoder
(decoding DDR memory errors via extracting error information from the
IMC MC error codes) for Sapphire Rapids for better decoding performance.

Co-developed-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Signed-off-by: Youquan Song <youquan.song@intel.com>
---
On top of the branch 'edac-for-next' of RAS git tree

 drivers/edac/i10nm_base.c | 102 ++++++++++++++++++++++++++------------
 1 file changed, 69 insertions(+), 33 deletions(-)


base-commit: 4c5be17ae44513686885ad939471cc2bd758237e

Comments

Tony Luck Feb. 8, 2023, 4:45 p.m. UTC | #1 
> Intel SDM (December 2022) vol3B 17.13.2 contains IMC MC error codes
> for Sapphire Rapids. Current i10nm_edac only supports firmware decoder
> (ACPI DSM methods) for Sapphire Rapids. So add the driver decoder
> (decoding DDR memory errors via extracting error information from the
> IMC MC error codes) for Sapphire Rapids for better decoding performance.

Applied to RAS tree. Thanks.

-Tony
diff mbox series

Patch

diff --git a/drivers/edac/i10nm_base.c b/drivers/edac/i10nm_base.c
index 46034310b78e..0a4691792801 100644
--- a/drivers/edac/i10nm_base.c
+++ b/drivers/edac/i10nm_base.c
@@ -434,20 +434,39 @@  static bool i10nm_check_2lm(struct res_config *cfg)
 }
 
 /*
- * Check whether the error comes from DDRT by ICX/Tremont model specific error code.
- * Refer to SDM vol3B 16.11.3 Intel IMC MC error codes for IA32_MCi_STATUS.
+ * Check whether the error comes from DDRT by ICX/Tremont/SPR model specific error code.
+ * Refer to SDM vol3B 17.11.3/17.13.2 Intel IMC MC error codes for IA32_MCi_STATUS.
  */
 static bool i10nm_mscod_is_ddrt(u32 mscod)
 {
-	switch (mscod) {
-	case 0x0106: case 0x0107:
-	case 0x0800: case 0x0804:
-	case 0x0806 ... 0x0808:
-	case 0x080a ... 0x080e:
-	case 0x0810: case 0x0811:
-	case 0x0816: case 0x081e:
-	case 0x081f:
-		return true;
+	switch (res_cfg->type) {
+	case I10NM:
+		switch (mscod) {
+		case 0x0106: case 0x0107:
+		case 0x0800: case 0x0804:
+		case 0x0806 ... 0x0808:
+		case 0x080a ... 0x080e:
+		case 0x0810: case 0x0811:
+		case 0x0816: case 0x081e:
+		case 0x081f:
+			return true;
+		}
+
+		break;
+	case SPR:
+		switch (mscod) {
+		case 0x0800: case 0x0804:
+		case 0x0806 ... 0x0808:
+		case 0x080a ... 0x080e:
+		case 0x0810: case 0x0811:
+		case 0x0816: case 0x081e:
+		case 0x081f:
+			return true;
+		}
+
+		break;
+	default:
+		return false;
 	}
 
 	return false;
@@ -455,6 +474,7 @@  static bool i10nm_mscod_is_ddrt(u32 mscod)
 
 static bool i10nm_mc_decode_available(struct mce *mce)
 {
+#define ICX_IMCx_CHy		0x06666000
 	u8 bank;
 
 	if (!decoding_via_mca || mem_cfg_2lm)
@@ -468,21 +488,26 @@  static bool i10nm_mc_decode_available(struct mce *mce)
 
 	switch (res_cfg->type) {
 	case I10NM:
-		if (bank < 13 || bank > 26)
-			return false;
-
-		/* DDRT errors can't be decoded from MCA bank registers */
-		if (MCI_MISC_ECC_MODE(mce->misc) == MCI_MISC_ECC_DDRT)
+		/* Check whether the bank is one of {13,14,17,18,21,22,25,26} */
+		if (!(ICX_IMCx_CHy & (1 << bank)))
 			return false;
-
-		if (i10nm_mscod_is_ddrt(MCI_STATUS_MSCOD(mce->status)))
+		break;
+	case SPR:
+		if (bank < 13 || bank > 20)
 			return false;
-
-		/* Check whether one of {13,14,17,18,21,22,25,26} */
-		return ((bank - 13) & BIT(1)) == 0;
+		break;
 	default:
 		return false;
 	}
+
+	/* DDRT errors can't be decoded from MCA bank registers */
+	if (MCI_MISC_ECC_MODE(mce->misc) == MCI_MISC_ECC_DDRT)
+		return false;
+
+	if (i10nm_mscod_is_ddrt(MCI_STATUS_MSCOD(mce->status)))
+		return false;
+
+	return true;
 }
 
 static bool i10nm_mc_decode(struct decoded_addr *res)
@@ -504,9 +529,29 @@  static bool i10nm_mc_decode(struct decoded_addr *res)
 
 	switch (res_cfg->type) {
 	case I10NM:
-		bank = m->bank - 13;
-		res->imc = bank / 4;
-		res->channel = bank % 2;
+		bank              = m->bank - 13;
+		res->imc          = bank / 4;
+		res->channel      = bank % 2;
+		res->column       = GET_BITFIELD(m->misc, 9, 18) << 2;
+		res->row          = GET_BITFIELD(m->misc, 19, 39);
+		res->bank_group   = GET_BITFIELD(m->misc, 40, 41);
+		res->bank_address = GET_BITFIELD(m->misc, 42, 43);
+		res->bank_group  |= GET_BITFIELD(m->misc, 44, 44) << 2;
+		res->rank         = GET_BITFIELD(m->misc, 56, 58);
+		res->dimm         = res->rank >> 2;
+		res->rank         = res->rank % 4;
+		break;
+	case SPR:
+		bank              = m->bank - 13;
+		res->imc          = bank / 2;
+		res->channel      = bank % 2;
+		res->column       = GET_BITFIELD(m->misc, 9, 18) << 2;
+		res->row          = GET_BITFIELD(m->misc, 19, 36);
+		res->bank_group   = GET_BITFIELD(m->misc, 37, 38);
+		res->bank_address = GET_BITFIELD(m->misc, 39, 40);
+		res->bank_group  |= GET_BITFIELD(m->misc, 41, 41) << 2;
+		res->rank         = GET_BITFIELD(m->misc, 57, 57);
+		res->dimm         = GET_BITFIELD(m->misc, 58, 58);
 		break;
 	default:
 		return false;
@@ -518,15 +563,6 @@  static bool i10nm_mc_decode(struct decoded_addr *res)
 		return false;
 	}
 
-	res->column       = GET_BITFIELD(m->misc, 9, 18) << 2;
-	res->row          = GET_BITFIELD(m->misc, 19, 39);
-	res->bank_group   = GET_BITFIELD(m->misc, 40, 41);
-	res->bank_address = GET_BITFIELD(m->misc, 42, 43);
-	res->bank_group  |= GET_BITFIELD(m->misc, 44, 44) << 2;
-	res->rank         = GET_BITFIELD(m->misc, 56, 58);
-	res->dimm         = res->rank >> 2;
-	res->rank         = res->rank % 4;
-
 	return true;
 }