[PATCH 0/4] EDAC/intel: Some Intel EDAC driver updates

[PATCH 0/4] EDAC/intel: Some Intel EDAC driver updates

[Tony Luck]

Patch 1&2 add printing of retry_rd_log registers for the HBM (high
bandwidth memory) on Sapphire Rapids CPUs.

Patch 3 prints from an extra set of retry_rd_log registers on Sapphire
Rapids.

Patch 4 adds sub-DIMM decode (bank, row, column) for Broadwell CPUs

When community review is complete, plan is to add to RAS tree after
v5.20-rc1 comes out for the v5.21 (v6.0?) merge window.

Qiuxu Zhuo (3):
  EDAC/skx_common: Add ChipSelect ADXL component
  EDAC/i10nm: Retrieve and print retry_rd_err_log registers for HBM
  EDAC/i10nm: Print an extra register set of retry_rd_err_log

Youquan Song (1):
  x86/sb_edac: Add row column translation for Broadwell

 drivers/edac/skx_common.h |  10 +++
 drivers/edac/i10nm_base.c | 153 ++++++++++++++++++++++++++++++++------
 drivers/edac/sb_edac.c    | 148 +++++++++++++++++++++++++++++++++---
 drivers/edac/skx_common.c |   5 ++
 4 files changed, 284 insertions(+), 32 deletions(-)


base-commit: 32346491ddf24599decca06190ebca03ff9de7f8
-- 
2.35.3

[PATCH 1/4] EDAC/skx_common: Add ChipSelect ADXL component

[Tony Luck]

From: Qiuxu Zhuo <qiuxu.zhuo@intel.com>

Each pseudo channel of HBM has its own retry_rd_err_log registers.
The bit 0 of ChipSelect ADXL component encodes the pseudo channel
number of HBM memory. So add ChipSelect ADXL component to get HBM
pseudo channel number.

Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
---
 drivers/edac/skx_common.h | 4 ++++
 drivers/edac/skx_common.c | 5 +++++
 2 files changed, 9 insertions(+)

diff --git a/drivers/edac/skx_common.h b/drivers/edac/skx_common.h
index 03ac067a80b9..70ec4f41911b 100644
--- a/drivers/edac/skx_common.h
+++ b/drivers/edac/skx_common.h
@@ -108,16 +108,19 @@ enum {
 	INDEX_MEMCTRL,
 	INDEX_CHANNEL,
 	INDEX_DIMM,
+	INDEX_CS,
 	INDEX_NM_FIRST,
 	INDEX_NM_MEMCTRL = INDEX_NM_FIRST,
 	INDEX_NM_CHANNEL,
 	INDEX_NM_DIMM,
+	INDEX_NM_CS,
 	INDEX_MAX
 };
 
 #define BIT_NM_MEMCTRL	BIT_ULL(INDEX_NM_MEMCTRL)
 #define BIT_NM_CHANNEL	BIT_ULL(INDEX_NM_CHANNEL)
 #define BIT_NM_DIMM	BIT_ULL(INDEX_NM_DIMM)
+#define BIT_NM_CS	BIT_ULL(INDEX_NM_CS)
 
 struct decoded_addr {
 	struct skx_dev *dev;
@@ -129,6 +132,7 @@ struct decoded_addr {
 	int	sktways;
 	int	chanways;
 	int	dimm;
+	int	cs;
 	int	rank;
 	int	channel_rank;
 	u64	rank_address;
diff --git a/drivers/edac/skx_common.c b/drivers/edac/skx_common.c
index 19c17c5198c5..ee074fb507d8 100644
--- a/drivers/edac/skx_common.c
+++ b/drivers/edac/skx_common.c
@@ -27,9 +27,11 @@ static const char * const component_names[] = {
 	[INDEX_MEMCTRL]		= "MemoryControllerId",
 	[INDEX_CHANNEL]		= "ChannelId",
 	[INDEX_DIMM]		= "DimmSlotId",
+	[INDEX_CS]		= "ChipSelect",
 	[INDEX_NM_MEMCTRL]	= "NmMemoryControllerId",
 	[INDEX_NM_CHANNEL]	= "NmChannelId",
 	[INDEX_NM_DIMM]		= "NmDimmSlotId",
+	[INDEX_NM_CS]		= "NmChipSelect",
 };
 
 static int component_indices[ARRAY_SIZE(component_names)];
@@ -139,10 +141,13 @@ static bool skx_adxl_decode(struct decoded_addr *res, bool error_in_1st_level_me
 			       (int)adxl_values[component_indices[INDEX_NM_CHANNEL]] : -1;
 		res->dimm    = (adxl_nm_bitmap & BIT_NM_DIMM) ?
 			       (int)adxl_values[component_indices[INDEX_NM_DIMM]] : -1;
+		res->cs      = (adxl_nm_bitmap & BIT_NM_CS) ?
+			       (int)adxl_values[component_indices[INDEX_NM_CS]] : -1;
 	} else {
 		res->imc     = (int)adxl_values[component_indices[INDEX_MEMCTRL]];
 		res->channel = (int)adxl_values[component_indices[INDEX_CHANNEL]];
 		res->dimm    = (int)adxl_values[component_indices[INDEX_DIMM]];
+		res->cs      = (int)adxl_values[component_indices[INDEX_CS]];
 	}
 
 	if (res->imc > NUM_IMC - 1 || res->imc < 0) {
-- 
2.35.3

[PATCH 2/4] EDAC/i10nm: Retrieve and print retry_rd_err_log registers for HBM

[Tony Luck]

From: Qiuxu Zhuo <qiuxu.zhuo@intel.com>

An HBM memory channel is divided into two pseudo channels. Each
pseudo channel has its own retry_rd_err_log registers. Retrieve and
print retry_rd_err_log registers of the HBM pseudo channel if the
memory error is from HBM.

Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
---
 drivers/edac/skx_common.h |  4 ++
 drivers/edac/i10nm_base.c | 84 +++++++++++++++++++++++++++++++--------
 2 files changed, 71 insertions(+), 17 deletions(-)

diff --git a/drivers/edac/skx_common.h b/drivers/edac/skx_common.h
index 70ec4f41911b..dbf8e458ad2b 100644
--- a/drivers/edac/skx_common.h
+++ b/drivers/edac/skx_common.h
@@ -158,7 +158,11 @@ struct res_config {
 	int sad_all_offset;
 	/* Offsets of retry_rd_err_log registers */
 	u32 *offsets_scrub;
+	u32 *offsets_scrub_hbm0;
+	u32 *offsets_scrub_hbm1;
 	u32 *offsets_demand;
+	u32 *offsets_demand_hbm0;
+	u32 *offsets_demand_hbm1;
 };
 
 typedef int (*get_dimm_config_f)(struct mem_ctl_info *mci,
diff --git a/drivers/edac/i10nm_base.c b/drivers/edac/i10nm_base.c
index 6cf50ee0b77c..976d8e8a4d1b 100644
--- a/drivers/edac/i10nm_base.c
+++ b/drivers/edac/i10nm_base.c
@@ -77,18 +77,20 @@ static int retry_rd_err_log;
 
 static u32 offsets_scrub_icx[]  = {0x22c60, 0x22c54, 0x22c5c, 0x22c58, 0x22c28, 0x20ed8};
 static u32 offsets_scrub_spr[]  = {0x22c60, 0x22c54, 0x22f08, 0x22c58, 0x22c28, 0x20ed8};
+static u32 offsets_scrub_spr_hbm0[]  = {0x2860, 0x2854, 0x2b08, 0x2858, 0x2828, 0x0ed8};
+static u32 offsets_scrub_spr_hbm1[]  = {0x2c60, 0x2c54, 0x2f08, 0x2c58, 0x2c28, 0x0fa8};
 static u32 offsets_demand_icx[] = {0x22e54, 0x22e60, 0x22e64, 0x22e58, 0x22e5c, 0x20ee0};
 static u32 offsets_demand_spr[] = {0x22e54, 0x22e60, 0x22f10, 0x22e58, 0x22e5c, 0x20ee0};
+static u32 offsets_demand_spr_hbm0[] = {0x2a54, 0x2a60, 0x2b10, 0x2a58, 0x2a5c, 0x0ee0};
+static u32 offsets_demand_spr_hbm1[] = {0x2e54, 0x2e60, 0x2f10, 0x2e58, 0x2e5c, 0x0fb0};
 
-static void __enable_retry_rd_err_log(struct skx_imc *imc, int chan, bool enable)
+static void __enable_retry_rd_err_log(struct skx_imc *imc, int chan, bool enable,
+				      u32 *offsets_scrub, u32 *offsets_demand)
 {
 	u32 s, d;
 
-	if (!imc->mbase)
-		return;
-
-	s = I10NM_GET_REG32(imc, chan, res_cfg->offsets_scrub[0]);
-	d = I10NM_GET_REG32(imc, chan, res_cfg->offsets_demand[0]);
+	s = I10NM_GET_REG32(imc, chan, offsets_scrub[0]);
+	d = I10NM_GET_REG32(imc, chan, offsets_demand[0]);
 
 	if (enable) {
 		/* Save default configurations */
@@ -115,21 +117,39 @@ static void __enable_retry_rd_err_log(struct skx_imc *imc, int chan, bool enable
 			d &= ~RETRY_RD_ERR_LOG_EN;
 	}
 
-	I10NM_SET_REG32(imc, chan, res_cfg->offsets_scrub[0], s);
-	I10NM_SET_REG32(imc, chan, res_cfg->offsets_demand[0], d);
+	I10NM_SET_REG32(imc, chan, offsets_scrub[0], s);
+	I10NM_SET_REG32(imc, chan, offsets_demand[0], d);
 }
 
 static void enable_retry_rd_err_log(bool enable)
 {
+	struct skx_imc *imc;
 	struct skx_dev *d;
 	int i, j;
 
 	edac_dbg(2, "\n");
 
 	list_for_each_entry(d, i10nm_edac_list, list)
-		for (i = 0; i < I10NM_NUM_IMC; i++)
-			for (j = 0; j < I10NM_NUM_CHANNELS; j++)
-				__enable_retry_rd_err_log(&d->imc[i], j, enable);
+		for (i = 0; i < I10NM_NUM_IMC; i++) {
+			imc = &d->imc[i];
+			if (!imc->mbase)
+				continue;
+
+			for (j = 0; j < I10NM_NUM_CHANNELS; j++) {
+				if (imc->hbm_mc) {
+					__enable_retry_rd_err_log(imc, j, enable,
+								  res_cfg->offsets_scrub_hbm0,
+								  res_cfg->offsets_demand_hbm0);
+					__enable_retry_rd_err_log(imc, j, enable,
+								  res_cfg->offsets_scrub_hbm1,
+								  res_cfg->offsets_demand_hbm1);
+				} else {
+					__enable_retry_rd_err_log(imc, j, enable,
+								  res_cfg->offsets_scrub,
+								  res_cfg->offsets_demand);
+				}
+			}
+	}
 }
 
 static void show_retry_rd_err_log(struct decoded_addr *res, char *msg,
@@ -140,12 +160,24 @@ static void show_retry_rd_err_log(struct decoded_addr *res, char *msg,
 	u32 corr0, corr1, corr2, corr3;
 	u64 log2a, log5;
 	u32 *offsets;
-	int n;
+	int n, pch;
 
 	if (!imc->mbase)
 		return;
 
-	offsets = scrub_err ? res_cfg->offsets_scrub : res_cfg->offsets_demand;
+	if (imc->hbm_mc) {
+		pch = res->cs & 1;
+
+		if (pch)
+			offsets = scrub_err ? res_cfg->offsets_scrub_hbm1 :
+					      res_cfg->offsets_demand_hbm1;
+		else
+			offsets = scrub_err ? res_cfg->offsets_scrub_hbm0 :
+					      res_cfg->offsets_demand_hbm0;
+	} else {
+		offsets = scrub_err ? res_cfg->offsets_scrub :
+				      res_cfg->offsets_demand;
+	}
 
 	log0 = I10NM_GET_REG32(imc, res->channel, offsets[0]);
 	log1 = I10NM_GET_REG32(imc, res->channel, offsets[1]);
@@ -163,10 +195,24 @@ static void show_retry_rd_err_log(struct decoded_addr *res, char *msg,
 			     log0, log1, log2, log3, log4, log5);
 	}
 
-	corr0 = I10NM_GET_REG32(imc, res->channel, 0x22c18);
-	corr1 = I10NM_GET_REG32(imc, res->channel, 0x22c1c);
-	corr2 = I10NM_GET_REG32(imc, res->channel, 0x22c20);
-	corr3 = I10NM_GET_REG32(imc, res->channel, 0x22c24);
+	if (imc->hbm_mc) {
+		if (pch) {
+			corr0 = I10NM_GET_REG32(imc, res->channel, 0x2c18);
+			corr1 = I10NM_GET_REG32(imc, res->channel, 0x2c1c);
+			corr2 = I10NM_GET_REG32(imc, res->channel, 0x2c20);
+			corr3 = I10NM_GET_REG32(imc, res->channel, 0x2c24);
+		} else {
+			corr0 = I10NM_GET_REG32(imc, res->channel, 0x2818);
+			corr1 = I10NM_GET_REG32(imc, res->channel, 0x281c);
+			corr2 = I10NM_GET_REG32(imc, res->channel, 0x2820);
+			corr3 = I10NM_GET_REG32(imc, res->channel, 0x2824);
+		}
+	} else {
+		corr0 = I10NM_GET_REG32(imc, res->channel, 0x22c18);
+		corr1 = I10NM_GET_REG32(imc, res->channel, 0x22c1c);
+		corr2 = I10NM_GET_REG32(imc, res->channel, 0x22c20);
+		corr3 = I10NM_GET_REG32(imc, res->channel, 0x22c24);
+	}
 
 	if (len - n > 0)
 		snprintf(msg + n, len - n,
@@ -420,7 +466,11 @@ static struct res_config spr_cfg = {
 	.sad_all_devfn		= PCI_DEVFN(10, 0),
 	.sad_all_offset		= 0x300,
 	.offsets_scrub		= offsets_scrub_spr,
+	.offsets_scrub_hbm0	= offsets_scrub_spr_hbm0,
+	.offsets_scrub_hbm1	= offsets_scrub_spr_hbm1,
 	.offsets_demand		= offsets_demand_spr,
+	.offsets_demand_hbm0	= offsets_demand_spr_hbm0,
+	.offsets_demand_hbm1	= offsets_demand_spr_hbm1,
 };
 
 static const struct x86_cpu_id i10nm_cpuids[] = {
-- 
2.35.3

[PATCH 3/4] EDAC/i10nm: Print an extra register set of retry_rd_err_log

[Tony Luck]

From: Qiuxu Zhuo <qiuxu.zhuo@intel.com>

Sapphire Rapids server adds an extra register set for logging more
retry_rd_err_log data. So add code to print the extra register set.

Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
---
 drivers/edac/skx_common.h |  2 +
 drivers/edac/i10nm_base.c | 81 +++++++++++++++++++++++++++++++++------
 2 files changed, 72 insertions(+), 11 deletions(-)

diff --git a/drivers/edac/skx_common.h b/drivers/edac/skx_common.h
index dbf8e458ad2b..e7226e935718 100644
--- a/drivers/edac/skx_common.h
+++ b/drivers/edac/skx_common.h
@@ -82,6 +82,7 @@ struct skx_dev {
 			struct pci_dev	*edev;
 			u32 retry_rd_err_log_s;
 			u32 retry_rd_err_log_d;
+			u32 retry_rd_err_log_d2;
 			struct skx_dimm {
 				u8 close_pg;
 				u8 bank_xor_enable;
@@ -161,6 +162,7 @@ struct res_config {
 	u32 *offsets_scrub_hbm0;
 	u32 *offsets_scrub_hbm1;
 	u32 *offsets_demand;
+	u32 *offsets_demand2;
 	u32 *offsets_demand_hbm0;
 	u32 *offsets_demand_hbm1;
 };
diff --git a/drivers/edac/i10nm_base.c b/drivers/edac/i10nm_base.c
index 976d8e8a4d1b..610ab8cce873 100644
--- a/drivers/edac/i10nm_base.c
+++ b/drivers/edac/i10nm_base.c
@@ -81,26 +81,38 @@ static u32 offsets_scrub_spr_hbm0[]  = {0x2860, 0x2854, 0x2b08, 0x2858, 0x2828,
 static u32 offsets_scrub_spr_hbm1[]  = {0x2c60, 0x2c54, 0x2f08, 0x2c58, 0x2c28, 0x0fa8};
 static u32 offsets_demand_icx[] = {0x22e54, 0x22e60, 0x22e64, 0x22e58, 0x22e5c, 0x20ee0};
 static u32 offsets_demand_spr[] = {0x22e54, 0x22e60, 0x22f10, 0x22e58, 0x22e5c, 0x20ee0};
+static u32 offsets_demand2_spr[] = {0x22c70, 0x22d80, 0x22f18, 0x22d58, 0x22c64, 0x20f10};
 static u32 offsets_demand_spr_hbm0[] = {0x2a54, 0x2a60, 0x2b10, 0x2a58, 0x2a5c, 0x0ee0};
 static u32 offsets_demand_spr_hbm1[] = {0x2e54, 0x2e60, 0x2f10, 0x2e58, 0x2e5c, 0x0fb0};
 
 static void __enable_retry_rd_err_log(struct skx_imc *imc, int chan, bool enable,
-				      u32 *offsets_scrub, u32 *offsets_demand)
+				      u32 *offsets_scrub, u32 *offsets_demand,
+				      u32 *offsets_demand2)
 {
-	u32 s, d;
+	u32 s, d, d2;
 
 	s = I10NM_GET_REG32(imc, chan, offsets_scrub[0]);
 	d = I10NM_GET_REG32(imc, chan, offsets_demand[0]);
+	if (offsets_demand2)
+		d2 = I10NM_GET_REG32(imc, chan, offsets_demand2[0]);
 
 	if (enable) {
 		/* Save default configurations */
 		imc->chan[chan].retry_rd_err_log_s = s;
 		imc->chan[chan].retry_rd_err_log_d = d;
+		if (offsets_demand2)
+			imc->chan[chan].retry_rd_err_log_d2 = d2;
 
 		s &= ~RETRY_RD_ERR_LOG_NOOVER_UC;
 		s |=  RETRY_RD_ERR_LOG_EN;
 		d &= ~RETRY_RD_ERR_LOG_NOOVER_UC;
 		d |=  RETRY_RD_ERR_LOG_EN;
+
+		if (offsets_demand2) {
+			d2 &= ~RETRY_RD_ERR_LOG_UC;
+			d2 |=  RETRY_RD_ERR_LOG_NOOVER;
+			d2 |=  RETRY_RD_ERR_LOG_EN;
+		}
 	} else {
 		/* Restore default configurations */
 		if (imc->chan[chan].retry_rd_err_log_s & RETRY_RD_ERR_LOG_UC)
@@ -115,10 +127,21 @@ static void __enable_retry_rd_err_log(struct skx_imc *imc, int chan, bool enable
 			d |=  RETRY_RD_ERR_LOG_NOOVER;
 		if (!(imc->chan[chan].retry_rd_err_log_d & RETRY_RD_ERR_LOG_EN))
 			d &= ~RETRY_RD_ERR_LOG_EN;
+
+		if (offsets_demand2) {
+			if (imc->chan[chan].retry_rd_err_log_d2 & RETRY_RD_ERR_LOG_UC)
+				d2 |=  RETRY_RD_ERR_LOG_UC;
+			if (!(imc->chan[chan].retry_rd_err_log_d2 & RETRY_RD_ERR_LOG_NOOVER))
+				d2 &=  ~RETRY_RD_ERR_LOG_NOOVER;
+			if (!(imc->chan[chan].retry_rd_err_log_d2 & RETRY_RD_ERR_LOG_EN))
+				d2 &= ~RETRY_RD_ERR_LOG_EN;
+		}
 	}
 
 	I10NM_SET_REG32(imc, chan, offsets_scrub[0], s);
 	I10NM_SET_REG32(imc, chan, offsets_demand[0], d);
+	if (offsets_demand2)
+		I10NM_SET_REG32(imc, chan, offsets_demand2[0], d2);
 }
 
 static void enable_retry_rd_err_log(bool enable)
@@ -139,14 +162,17 @@ static void enable_retry_rd_err_log(bool enable)
 				if (imc->hbm_mc) {
 					__enable_retry_rd_err_log(imc, j, enable,
 								  res_cfg->offsets_scrub_hbm0,
-								  res_cfg->offsets_demand_hbm0);
+								  res_cfg->offsets_demand_hbm0,
+								  NULL);
 					__enable_retry_rd_err_log(imc, j, enable,
 								  res_cfg->offsets_scrub_hbm1,
-								  res_cfg->offsets_demand_hbm1);
+								  res_cfg->offsets_demand_hbm1,
+								  NULL);
 				} else {
 					__enable_retry_rd_err_log(imc, j, enable,
 								  res_cfg->offsets_scrub,
-								  res_cfg->offsets_demand);
+								  res_cfg->offsets_demand,
+								  res_cfg->offsets_demand2);
 				}
 			}
 	}
@@ -158,7 +184,10 @@ static void show_retry_rd_err_log(struct decoded_addr *res, char *msg,
 	struct skx_imc *imc = &res->dev->imc[res->imc];
 	u32 log0, log1, log2, log3, log4;
 	u32 corr0, corr1, corr2, corr3;
+	u32 lxg0, lxg1, lxg3, lxg4;
+	u32 *xffsets = NULL;
 	u64 log2a, log5;
+	u64 lxg2a, lxg5;
 	u32 *offsets;
 	int n, pch;
 
@@ -175,8 +204,12 @@ static void show_retry_rd_err_log(struct decoded_addr *res, char *msg,
 			offsets = scrub_err ? res_cfg->offsets_scrub_hbm0 :
 					      res_cfg->offsets_demand_hbm0;
 	} else {
-		offsets = scrub_err ? res_cfg->offsets_scrub :
-				      res_cfg->offsets_demand;
+		if (scrub_err) {
+			offsets = res_cfg->offsets_scrub;
+		} else {
+			offsets = res_cfg->offsets_demand;
+			xffsets = res_cfg->offsets_demand2;
+		}
 	}
 
 	log0 = I10NM_GET_REG32(imc, res->channel, offsets[0]);
@@ -185,10 +218,28 @@ static void show_retry_rd_err_log(struct decoded_addr *res, char *msg,
 	log4 = I10NM_GET_REG32(imc, res->channel, offsets[4]);
 	log5 = I10NM_GET_REG64(imc, res->channel, offsets[5]);
 
+	if (xffsets) {
+		lxg0 = I10NM_GET_REG32(imc, res->channel, xffsets[0]);
+		lxg1 = I10NM_GET_REG32(imc, res->channel, xffsets[1]);
+		lxg3 = I10NM_GET_REG32(imc, res->channel, xffsets[3]);
+		lxg4 = I10NM_GET_REG32(imc, res->channel, xffsets[4]);
+		lxg5 = I10NM_GET_REG64(imc, res->channel, xffsets[5]);
+	}
+
 	if (res_cfg->type == SPR) {
 		log2a = I10NM_GET_REG64(imc, res->channel, offsets[2]);
-		n = snprintf(msg, len, " retry_rd_err_log[%.8x %.8x %.16llx %.8x %.8x %.16llx]",
+		n = snprintf(msg, len, " retry_rd_err_log[%.8x %.8x %.16llx %.8x %.8x %.16llx",
 			     log0, log1, log2a, log3, log4, log5);
+
+		if (len - n > 0) {
+			if (xffsets) {
+				lxg2a = I10NM_GET_REG64(imc, res->channel, xffsets[2]);
+				n += snprintf(msg + n, len - n, " %.8x %.8x %.16llx %.8x %.8x %.16llx]",
+					     lxg0, lxg1, lxg2a, lxg3, lxg4, lxg5);
+			} else {
+				n += snprintf(msg + n, len - n, "]");
+			}
+		}
 	} else {
 		log2 = I10NM_GET_REG32(imc, res->channel, offsets[2]);
 		n = snprintf(msg, len, " retry_rd_err_log[%.8x %.8x %.8x %.8x %.8x %.16llx]",
@@ -223,9 +274,16 @@ static void show_retry_rd_err_log(struct decoded_addr *res, char *msg,
 			 corr3 & 0xffff, corr3 >> 16);
 
 	/* Clear status bits */
-	if (retry_rd_err_log == 2 && (log0 & RETRY_RD_ERR_LOG_OVER_UC_V)) {
-		log0 &= ~RETRY_RD_ERR_LOG_OVER_UC_V;
-		I10NM_SET_REG32(imc, res->channel, offsets[0], log0);
+	if (retry_rd_err_log == 2) {
+		if (log0 & RETRY_RD_ERR_LOG_OVER_UC_V) {
+			log0 &= ~RETRY_RD_ERR_LOG_OVER_UC_V;
+			I10NM_SET_REG32(imc, res->channel, offsets[0], log0);
+		}
+
+		if (xffsets && (lxg0 & RETRY_RD_ERR_LOG_OVER_UC_V)) {
+			lxg0 &= ~RETRY_RD_ERR_LOG_OVER_UC_V;
+			I10NM_SET_REG32(imc, res->channel, xffsets[0], lxg0);
+		}
 	}
 }
 
@@ -469,6 +527,7 @@ static struct res_config spr_cfg = {
 	.offsets_scrub_hbm0	= offsets_scrub_spr_hbm0,
 	.offsets_scrub_hbm1	= offsets_scrub_spr_hbm1,
 	.offsets_demand		= offsets_demand_spr,
+	.offsets_demand2	= offsets_demand2_spr,
 	.offsets_demand_hbm0	= offsets_demand_spr_hbm0,
 	.offsets_demand_hbm1	= offsets_demand_spr_hbm1,
 };
-- 
2.35.3

[PATCH 4/4] x86/sb_edac: Add row column translation for Broadwell

[Tony Luck]

From: Youquan Song <youquan.song@intel.com>

The sb_edac driver lacks translation for DIMM internal address.

Add memory address translation for row/column/bank/bank_group
on Broadwell.

Signed-off-by: Youquan Song <youquan.song@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
---
 drivers/edac/sb_edac.c | 148 ++++++++++++++++++++++++++++++++++++++---
 1 file changed, 138 insertions(+), 10 deletions(-)

diff --git a/drivers/edac/sb_edac.c b/drivers/edac/sb_edac.c
index 9678ab97c7ac..8e39370fdb5c 100644
--- a/drivers/edac/sb_edac.c
+++ b/drivers/edac/sb_edac.c
@@ -335,6 +335,12 @@ struct sbridge_info {
 struct sbridge_channel {
 	u32		ranks;
 	u32		dimms;
+	struct dimm {
+		u32 rowbits;
+		u32 colbits;
+		u32 bank_xor_enable;
+		u32 amap_fine;
+	} dimm[MAX_DIMMS];
 };
 
 struct pci_id_descr {
@@ -1603,7 +1609,7 @@ static int __populate_dimms(struct mem_ctl_info *mci,
 		banks = 8;
 
 	for (i = 0; i < channels; i++) {
-		u32 mtr;
+		u32 mtr, amap = 0;
 
 		int max_dimms_per_channel;
 
@@ -1615,6 +1621,7 @@ static int __populate_dimms(struct mem_ctl_info *mci,
 			max_dimms_per_channel = ARRAY_SIZE(mtr_regs);
 			if (!pvt->pci_tad[i])
 				continue;
+			pci_read_config_dword(pvt->pci_tad[i], 0x8c, &amap);
 		}
 
 		for (j = 0; j < max_dimms_per_channel; j++) {
@@ -1627,6 +1634,7 @@ static int __populate_dimms(struct mem_ctl_info *mci,
 					mtr_regs[j], &mtr);
 			}
 			edac_dbg(4, "Channel #%d  MTR%d = %x\n", i, j, mtr);
+
 			if (IS_DIMM_PRESENT(mtr)) {
 				if (!IS_ECC_ENABLED(pvt->info.mcmtr)) {
 					sbridge_printk(KERN_ERR, "CPU SrcID #%d, Ha #%d, Channel #%d has DIMMs, but ECC is disabled\n",
@@ -1661,6 +1669,11 @@ static int __populate_dimms(struct mem_ctl_info *mci,
 				dimm->dtype = pvt->info.get_width(pvt, mtr);
 				dimm->mtype = mtype;
 				dimm->edac_mode = mode;
+				pvt->channel[i].dimm[j].rowbits = order_base_2(rows);
+				pvt->channel[i].dimm[j].colbits = order_base_2(cols);
+				pvt->channel[i].dimm[j].bank_xor_enable =
+						GET_BITFIELD(pvt->info.mcmtr, 9, 9);
+				pvt->channel[i].dimm[j].amap_fine = GET_BITFIELD(amap, 0, 0);
 				snprintf(dimm->label, sizeof(dimm->label),
 						 "CPU_SrcID#%u_Ha#%u_Chan#%u_DIMM#%u",
 						 pvt->sbridge_dev->source_id, pvt->sbridge_dev->dom, i, j);
@@ -1922,6 +1935,99 @@ static struct mem_ctl_info *get_mci_for_node_id(u8 node_id, u8 ha)
 	return NULL;
 }
 
+static u8 sb_close_row[] = {
+	15, 16, 17, 18, 20, 21, 22, 28, 10, 11, 12, 13, 29, 30, 31, 32, 33
+};
+
+static u8 sb_close_column[] = {
+	3, 4, 5, 14, 19, 23, 24, 25, 26, 27
+};
+
+static u8 sb_open_row[] = {
+	14, 15, 16, 20, 28, 21, 22, 23, 24, 25, 26, 27, 29, 30, 31, 32, 33
+};
+
+static u8 sb_open_column[] = {
+	3, 4, 5, 6, 7, 8, 9, 10, 11, 12
+};
+
+static u8 sb_open_fine_column[] = {
+	3, 4, 5, 7, 8, 9, 10, 11, 12, 13
+};
+
+static int sb_bits(u64 addr, int nbits, u8 *bits)
+{
+	int i, res = 0;
+
+	for (i = 0; i < nbits; i++)
+		res |= ((addr >> bits[i]) & 1) << i;
+	return res;
+}
+
+static int sb_bank_bits(u64 addr, int b0, int b1, int do_xor, int x0, int x1)
+{
+	int ret = GET_BITFIELD(addr, b0, b0) | (GET_BITFIELD(addr, b1, b1) << 1);
+
+	if (do_xor)
+		ret ^= GET_BITFIELD(addr, x0, x0) | (GET_BITFIELD(addr, x1, x1) << 1);
+
+	return ret;
+}
+
+static bool sb_decode_ddr4(struct mem_ctl_info *mci, int ch, u8 rank,
+			   u64 rank_addr, char *msg)
+{
+	int dimmno = 0;
+	int row, col, bank_address, bank_group;
+	struct sbridge_pvt *pvt;
+	u32 bg0 = 0, rowbits = 0, colbits = 0;
+	u32 amap_fine = 0, bank_xor_enable = 0;
+
+	dimmno = (rank < 12) ? rank / 4 : 2;
+	pvt = mci->pvt_info;
+	amap_fine =  pvt->channel[ch].dimm[dimmno].amap_fine;
+	bg0 = amap_fine ? 6 : 13;
+	rowbits = pvt->channel[ch].dimm[dimmno].rowbits;
+	colbits = pvt->channel[ch].dimm[dimmno].colbits;
+	bank_xor_enable = pvt->channel[ch].dimm[dimmno].bank_xor_enable;
+
+	if (pvt->is_lockstep) {
+		pr_warn_once("LockStep row/column decode is not supported yet!\n");
+		msg[0] = '\0';
+		return false;
+	}
+
+	if (pvt->is_close_pg) {
+		row = sb_bits(rank_addr, rowbits, sb_close_row);
+		col = sb_bits(rank_addr, colbits, sb_close_column);
+		col |= 0x400; /* C10 is autoprecharge, always set */
+		bank_address = sb_bank_bits(rank_addr, 8, 9, bank_xor_enable, 22, 28);
+		bank_group = sb_bank_bits(rank_addr, 6, 7, bank_xor_enable, 20, 21);
+	} else {
+		row = sb_bits(rank_addr, rowbits, sb_open_row);
+		if (amap_fine)
+			col = sb_bits(rank_addr, colbits, sb_open_fine_column);
+		else
+			col = sb_bits(rank_addr, colbits, sb_open_column);
+		bank_address = sb_bank_bits(rank_addr, 18, 19, bank_xor_enable, 22, 23);
+		bank_group = sb_bank_bits(rank_addr, bg0, 17, bank_xor_enable, 20, 21);
+	}
+
+	row &= (1u << rowbits) - 1;
+
+	sprintf(msg, "row:0x%x col:0x%x bank_addr:%d bank_group:%d",
+		row, col, bank_address, bank_group);
+	return true;
+}
+
+static bool sb_decode_ddr3(struct mem_ctl_info *mci, int ch, u8 rank,
+			   u64 rank_addr, char *msg)
+{
+	pr_warn_once("DDR3 row/column decode not support yet!\n");
+	msg[0] = '\0';
+	return false;
+}
+
 static int get_memory_error_data(struct mem_ctl_info *mci,
 				 u64 addr,
 				 u8 *socket, u8 *ha,
@@ -1937,12 +2043,13 @@ static int get_memory_error_data(struct mem_ctl_info *mci,
 	int			interleave_mode, shiftup = 0;
 	unsigned int		sad_interleave[MAX_INTERLEAVE];
 	u32			reg, dram_rule;
-	u8			ch_way, sck_way, pkg, sad_ha = 0;
+	u8			ch_way, sck_way, pkg, sad_ha = 0, rankid = 0;
 	u32			tad_offset;
 	u32			rir_way;
 	u32			mb, gb;
 	u64			ch_addr, offset, limit = 0, prv = 0;
-
+	u64			rank_addr;
+	enum mem_type		mtype;
 
 	/*
 	 * Step 0) Check if the address is at special memory ranges
@@ -2226,6 +2333,28 @@ static int get_memory_error_data(struct mem_ctl_info *mci,
 	pci_read_config_dword(pvt->pci_tad[base_ch], rir_offset[n_rir][idx], &reg);
 	*rank = RIR_RNK_TGT(pvt->info.type, reg);
 
+	if (pvt->info.type == BROADWELL) {
+		if (pvt->is_close_pg)
+			shiftup = 6;
+		else
+			shiftup = 13;
+
+		rank_addr = ch_addr >> shiftup;
+		rank_addr /= (1 << rir_way);
+		rank_addr <<= shiftup;
+		rank_addr |= ch_addr & GENMASK_ULL(shiftup - 1, 0);
+		rank_addr -= RIR_OFFSET(pvt->info.type, reg);
+
+		mtype = pvt->info.get_memory_type(pvt);
+		rankid = *rank;
+		if (mtype == MEM_DDR4 || mtype == MEM_RDDR4)
+			sb_decode_ddr4(mci, base_ch, rankid, rank_addr, msg);
+		else
+			sb_decode_ddr3(mci, base_ch, rankid, rank_addr, msg);
+	} else {
+		msg[0] = '\0';
+	}
+
 	edac_dbg(0, "RIR#%d: channel address 0x%08Lx < 0x%08Lx, RIR interleave %d, index %d\n",
 		 n_rir,
 		 ch_addr,
@@ -2950,7 +3079,7 @@ static void sbridge_mce_output_error(struct mem_ctl_info *mci,
 	struct mem_ctl_info *new_mci;
 	struct sbridge_pvt *pvt = mci->pvt_info;
 	enum hw_event_mc_err_type tp_event;
-	char *optype, msg[256];
+	char *optype, msg[256], msg_full[512];
 	bool ripv = GET_BITFIELD(m->mcgstatus, 0, 0);
 	bool overflow = GET_BITFIELD(m->status, 62, 62);
 	bool uncorrected_error = GET_BITFIELD(m->status, 61, 61);
@@ -3089,18 +3218,17 @@ static void sbridge_mce_output_error(struct mem_ctl_info *mci,
 	 */
 	if (!pvt->is_lockstep && !pvt->is_cur_addr_mirrored && !pvt->is_close_pg)
 		channel = first_channel;
-
-	snprintf(msg, sizeof(msg),
-		 "%s%s area:%s err_code:%04x:%04x socket:%d ha:%d channel_mask:%ld rank:%d",
+	snprintf(msg_full, sizeof(msg_full),
+		 "%s%s area:%s err_code:%04x:%04x socket:%d ha:%d channel_mask:%ld rank:%d %s",
 		 overflow ? " OVERFLOW" : "",
 		 (uncorrected_error && recoverable) ? " recoverable" : "",
 		 area_type,
 		 mscod, errcode,
 		 socket, ha,
 		 channel_mask,
-		 rank);
+		 rank, msg);
 
-	edac_dbg(0, "%s\n", msg);
+	edac_dbg(0, "%s\n", msg_full);
 
 	/* FIXME: need support for channel mask */
 
@@ -3111,7 +3239,7 @@ static void sbridge_mce_output_error(struct mem_ctl_info *mci,
 	edac_mc_handle_error(tp_event, mci, core_err_cnt,
 			     m->addr >> PAGE_SHIFT, m->addr & ~PAGE_MASK, 0,
 			     channel, dimm, -1,
-			     optype, msg);
+			     optype, msg_full);
 	return;
 err_parsing:
 	edac_mc_handle_error(tp_event, mci, core_err_cnt, 0, 0, 0,
-- 
2.35.3