amd64_edac: Add ECC decoding support for newer F15h models

On newer models, support has been included for upto 4 DCT's, however,
only DCT0 and DCT3 are currently configured (cf BKDG Section 2.10).
Also, the routing DRAM Requests algorithm is different for F15h M30h.
Thus it is cleaner to use a brand new function rather than adding quirks
to the more generic f1x_match_to_this_node(). Refer to "2.10.5 DRAM
Routing Requests" in the BKDG for further info.

Tested on Fam15h M30h with ECC turned on using mce_amd_inj facility and
verified to be functionally correct.

While at it, verify if erratum workarounds for E505 and E637 still hold.
From email conversations within AMD, the current status of the errata
is:

      * Erratum 505: fixed in model 0x1, stepping 0x1 and later.
      * Erratum 637: not fixed.

Signed-off-by: Aravind Gopalakrishnan <Aravind.Gopalakrishnan@amd.com>
[ Cleanups, corrections ]
Signed-off-by: Borislav Petkov <bp@suse.de>

1 files changed, 216 insertions, 30 deletions

diff --git a/drivers/edac/amd64_edac.c b/drivers/edac/amd64_edac.c
index 8b6a034..a5d6348 100644
--- a/drivers/edac/amd64_edac.c
+++ b/drivers/edac/amd64_edac.c
@@ -123,7 +123,7 @@ static void f15h_select_dct(struct amd64_pvt *pvt, u8 dct)
 	u32 reg = 0;
 
 	amd64_read_pci_cfg(pvt->F1, DCT_CFG_SEL, &reg);
-	reg &= 0xfffffffe;
+	reg &= (pvt->model >= 0x30) ? ~3 : ~1;
 	reg |= dct;
 	amd64_write_pci_cfg(pvt->F1, DCT_CFG_SEL, reg);
 }
@@ -133,8 +133,9 @@ static int f15_read_dct_pci_cfg(struct amd64_pvt *pvt, int addr, u32 *val,
 {
 	u8 dct  = 0;
 
+	/* For F15 M30h, the second dct is DCT 3, refer to BKDG Section 2.10 */
 	if (addr >= 0x140 && addr <= 0x1a0) {
-		dct   = 1;
+		dct   = (pvt->model >= 0x30) ? 3 : 1;
 		addr -= 0x100;
 	}
 
@@ -205,8 +206,10 @@ static int amd64_set_scrub_rate(struct mem_ctl_info *mci, u32 bw)
 	if (boot_cpu_data.x86 == 0xf)
 		min_scrubrate = 0x0;
 
-	/* F15h Erratum #505 */
-	if (boot_cpu_data.x86 == 0x15)
+	/* Erratum #505 for F15h Model 0x00 - Model 0x01, Stepping 0 */
+	if (boot_cpu_data.x86 == 0x15 &&
+		boot_cpu_data.x86_model <= 0x01 &&
+		boot_cpu_data.x86_mask < 0x1)
 		f15h_select_dct(pvt, 0);
 
 	return __amd64_set_scrub_rate(pvt->F3, bw, min_scrubrate);
@@ -218,8 +221,10 @@ static int amd64_get_scrub_rate(struct mem_ctl_info *mci)
 	u32 scrubval = 0;
 	int i, retval = -EINVAL;
 
-	/* F15h Erratum #505 */
-	if (boot_cpu_data.x86 == 0x15)
+	/* Erratum #505 for F15h Model 0x00 - Model 0x01, Stepping 0 */
+	if (boot_cpu_data.x86 == 0x15 &&
+		boot_cpu_data.x86_model <= 0x01 &&
+		boot_cpu_data.x86_mask < 0x1)
 		f15h_select_dct(pvt, 0);
 
 	amd64_read_pci_cfg(pvt->F3, SCRCTRL, &scrubval);
@@ -335,7 +340,7 @@ static void get_cs_base_and_mask(struct amd64_pvt *pvt, int csrow, u8 dct,
 	u64 csbase, csmask, base_bits, mask_bits;
 	u8 addr_shift;
 
-	if (boot_cpu_data.x86 == 0xf && pvt->ext_model < K8_REV_F) {
+	if (pvt->fam == 0xf && pvt->ext_model < K8_REV_F) {
 		csbase		= pvt->csels[dct].csbases[csrow];
 		csmask		= pvt->csels[dct].csmasks[csrow];
 		base_bits	= GENMASK(21, 31) | GENMASK(9, 15);
@@ -343,10 +348,11 @@ static void get_cs_base_and_mask(struct amd64_pvt *pvt, int csrow, u8 dct,
 		addr_shift	= 4;
 
 	/*
-	* F16h needs two addr_shift values: 8 for high and 6 for low
-	* (cf. F16h BKDG).
-	*/
-	} else if (boot_cpu_data.x86 == 0x16) {
+	 * F16h and F15h, models 30h and later need two addr_shift values:
+	 * 8 for high and 6 for low (cf. F16h BKDG).
+	 */
+	} else if (pvt->fam == 0x16 ||
+		  (pvt->fam == 0x15 && pvt->model >= 0x30)) {
 		csbase          = pvt->csels[dct].csbases[csrow];
 		csmask          = pvt->csels[dct].csmasks[csrow >> 1];
 
@@ -736,13 +742,16 @@ static void dump_misc_regs(struct amd64_pvt *pvt)
 }
 
 /*
- * see BKDG, F2x[1,0][5C:40], F2[1,0][6C:60]
+ * See BKDG, F2x[1,0][5C:40], F2[1,0][6C:60]
  */
 static void prep_chip_selects(struct amd64_pvt *pvt)
 {
-	if (boot_cpu_data.x86 == 0xf && pvt->ext_model < K8_REV_F) {
+	if (pvt->fam == 0xf && pvt->ext_model < K8_REV_F) {
 		pvt->csels[0].b_cnt = pvt->csels[1].b_cnt = 8;
 		pvt->csels[0].m_cnt = pvt->csels[1].m_cnt = 8;
+	} else if (pvt->fam == 0x15 && pvt->model >= 0x30) {
+		pvt->csels[0].b_cnt = pvt->csels[1].b_cnt = 4;
+		pvt->csels[0].m_cnt = pvt->csels[1].m_cnt = 2;
 	} else {
 		pvt->csels[0].b_cnt = pvt->csels[1].b_cnt = 8;
 		pvt->csels[0].m_cnt = pvt->csels[1].m_cnt = 4;
@@ -916,15 +925,15 @@ static struct pci_dev *pci_get_related_function(unsigned int vendor,
 static void read_dram_base_limit_regs(struct amd64_pvt *pvt, unsigned range)
 {
 	struct amd_northbridge *nb;
-	struct pci_dev *misc, *f1 = NULL;
-	struct cpuinfo_x86 *c = &boot_cpu_data;
+	struct pci_dev *f1 = NULL;
+	unsigned int pci_func;
 	int off = range << 3;
 	u32 llim;
 
 	amd64_read_pci_cfg(pvt->F1, DRAM_BASE_LO + off,  &pvt->ranges[range].base.lo);
 	amd64_read_pci_cfg(pvt->F1, DRAM_LIMIT_LO + off, &pvt->ranges[range].lim.lo);
 
-	if (c->x86 == 0xf)
+	if (pvt->fam == 0xf)
 		return;
 
 	if (!dram_rw(pvt, range))
@@ -934,15 +943,17 @@ static void read_dram_base_limit_regs(struct amd64_pvt *pvt, unsigned range)
 	amd64_read_pci_cfg(pvt->F1, DRAM_LIMIT_HI + off, &pvt->ranges[range].lim.hi);
 
 	/* F15h: factor in CC6 save area by reading dst node's limit reg */
-	if (c->x86 != 0x15)
+	if (pvt->fam != 0x15)
 		return;
 
 	nb = node_to_amd_nb(dram_dst_node(pvt, range));
 	if (WARN_ON(!nb))
 		return;
 
-	misc = nb->misc;
-	f1 = pci_get_related_function(misc->vendor, PCI_DEVICE_ID_AMD_15H_NB_F1, misc);
+	pci_func = (pvt->model == 0x30) ? PCI_DEVICE_ID_AMD_15H_M30H_NB_F1
+					: PCI_DEVICE_ID_AMD_15H_NB_F1;
+
+	f1 = pci_get_related_function(nb->misc->vendor, pci_func, nb->misc);
 	if (WARN_ON(!f1))
 		return;
 
@@ -1173,7 +1184,7 @@ static int f15_dbam_to_chip_select(struct amd64_pvt *pvt, u8 dct,
 }
 
 /*
- * F16h has only limited cs_modes
+ * F16h and F15h model 30h have only limited cs_modes.
  */
 static int f16_dbam_to_chip_select(struct amd64_pvt *pvt, u8 dct,
 				unsigned cs_mode)
@@ -1218,6 +1229,29 @@ static void read_dram_ctl_register(struct amd64_pvt *pvt)
 }
 
 /*
+ * Determine channel (DCT) based on the interleaving mode (see F15h M30h BKDG,
+ * 2.10.12 Memory Interleaving Modes).
+ */
+static u8 f15_m30h_determine_channel(struct amd64_pvt *pvt, u64 sys_addr,
+				     u8 intlv_en, int num_dcts_intlv,
+				     u32 dct_sel)
+{
+	u8 channel = 0;
+	u8 select;
+
+	if (!(intlv_en))
+		return (u8)(dct_sel);
+
+	if (num_dcts_intlv == 2) {
+		select = (sys_addr >> 8) & 0x3;
+		channel = select ? 0x3 : 0;
+	} else if (num_dcts_intlv == 4)
+		channel = (sys_addr >> 8) & 0x7;
+
+	return channel;
+}
+
+/*
  * Determine channel (DCT) based on the interleaving mode: F10h BKDG, 2.8.9 Memory
  * Interleaving Modes.
  */
@@ -1366,6 +1400,10 @@ static int f1x_lookup_addr_in_dct(u64 in_addr, u8 nid, u8 dct)
 			 (in_addr & cs_mask), (cs_base & cs_mask));
 
 		if ((in_addr & cs_mask) == (cs_base & cs_mask)) {
+			if (pvt->fam == 0x15 && pvt->model >= 0x30) {
+				cs_found =  csrow;
+				break;
+			}
 			cs_found = f10_process_possible_spare(pvt, dct, csrow);
 
 			edac_dbg(1, " MATCH csrow=%d\n", cs_found);
@@ -1492,20 +1530,142 @@ static int f1x_match_to_this_node(struct amd64_pvt *pvt, unsigned range,
 	return cs_found;
 }
 
-static int f1x_translate_sysaddr_to_cs(struct amd64_pvt *pvt, u64 sys_addr,
-				       int *chan_sel)
+static int f15_m30h_match_to_this_node(struct amd64_pvt *pvt, unsigned range,
+					u64 sys_addr, int *chan_sel)
+{
+	int cs_found = -EINVAL;
+	int num_dcts_intlv = 0;
+	u64 chan_addr, chan_offset;
+	u64 dct_base, dct_limit;
+	u32 dct_cont_base_reg, dct_cont_limit_reg, tmp;
+	u8 channel, alias_channel, leg_mmio_hole, dct_sel, dct_offset_en;
+
+	u64 dhar_offset		= f10_dhar_offset(pvt);
+	u8 intlv_addr		= dct_sel_interleave_addr(pvt);
+	u8 node_id		= dram_dst_node(pvt, range);
+	u8 intlv_en		= dram_intlv_en(pvt, range);
+
+	amd64_read_pci_cfg(pvt->F1, DRAM_CONT_BASE, &dct_cont_base_reg);
+	amd64_read_pci_cfg(pvt->F1, DRAM_CONT_LIMIT, &dct_cont_limit_reg);
+
+	dct_offset_en		= (u8) ((dct_cont_base_reg >> 3) & BIT(0));
+	dct_sel			= (u8) ((dct_cont_base_reg >> 4) & 0x7);
+
+	edac_dbg(1, "(range %d) SystemAddr= 0x%llx Limit=0x%llx\n",
+		 range, sys_addr, get_dram_limit(pvt, range));
+
+	if (!(get_dram_base(pvt, range)  <= sys_addr) &&
+	    !(get_dram_limit(pvt, range) >= sys_addr))
+		return -EINVAL;
+
+	if (dhar_valid(pvt) &&
+	    dhar_base(pvt) <= sys_addr &&
+	    sys_addr < BIT_64(32)) {
+		amd64_warn("Huh? Address is in the MMIO hole: 0x%016llx\n",
+			    sys_addr);
+		return -EINVAL;
+	}
+
+	/* Verify sys_addr is within DCT Range. */
+	dct_base = (dct_sel_baseaddr(pvt) << 27);
+	dct_limit = (((dct_cont_limit_reg >> 11) & 0x1FFF) << 27) | 0x7FFFFFF;
+
+	if (!(dct_cont_base_reg & BIT(0)) &&
+	    !(dct_base <= sys_addr && dct_limit >= sys_addr))
+		return -EINVAL;
+
+	/* Verify number of dct's that participate in channel interleaving. */
+	num_dcts_intlv = (int) hweight8(intlv_en);
+
+	if (!(num_dcts_intlv % 2 == 0) || (num_dcts_intlv > 4))
+		return -EINVAL;
+
+	channel = f15_m30h_determine_channel(pvt, sys_addr, intlv_en,
+					     num_dcts_intlv, dct_sel);
+
+	/* Verify we stay within the MAX number of channels allowed */
+	if (channel > 4 || channel < 0)
+		return -EINVAL;
+
+	leg_mmio_hole = (u8) (dct_cont_base_reg >> 1 & BIT(0));
+
+	/* Get normalized DCT addr */
+	if (leg_mmio_hole && (sys_addr >= BIT_64(32)))
+		chan_offset = dhar_offset;
+	else
+		chan_offset = dct_base;
+
+	chan_addr = sys_addr - chan_offset;
+
+	/* remove channel interleave */
+	if (num_dcts_intlv == 2) {
+		if (intlv_addr == 0x4)
+			chan_addr = ((chan_addr >> 9) << 8) |
+						(chan_addr & 0xff);
+		else if (intlv_addr == 0x5)
+			chan_addr = ((chan_addr >> 10) << 9) |
+						(chan_addr & 0x1ff);
+		else
+			return -EINVAL;
+
+	} else if (num_dcts_intlv == 4) {
+		if (intlv_addr == 0x4)
+			chan_addr = ((chan_addr >> 10) << 8) |
+							(chan_addr & 0xff);
+		else if (intlv_addr == 0x5)
+			chan_addr = ((chan_addr >> 11) << 9) |
+							(chan_addr & 0x1ff);
+		else
+			return -EINVAL;
+	}
+
+	if (dct_offset_en) {
+		amd64_read_pci_cfg(pvt->F1,
+				   DRAM_CONT_HIGH_OFF + (int) channel * 4,
+				   &tmp);
+		chan_addr +=  ((tmp >> 11) & 0xfff) << 27;
+	}
+
+	f15h_select_dct(pvt, channel);
+
+	edac_dbg(1, "   Normalized DCT addr: 0x%llx\n", chan_addr);
+
+	/*
+	 * Find Chip select:
+	 * if channel = 3, then alias it to 1. This is because, in F15 M30h,
+	 * there is support for 4 DCT's, but only 2 are currently functional.
+	 * They are DCT0 and DCT3. But we have read all registers of DCT3 into
+	 * pvt->csels[1]. So we need to use '1' here to get correct info.
+	 * Refer F15 M30h BKDG Section 2.10 and 2.10.3 for clarifications.
+	 */
+	alias_channel =  (channel == 3) ? 1 : channel;
+
+	cs_found = f1x_lookup_addr_in_dct(chan_addr, node_id, alias_channel);
+
+	if (cs_found >= 0)
+		*chan_sel = alias_channel;
+
+	return cs_found;
+}
+
+static int f1x_translate_sysaddr_to_cs(struct amd64_pvt *pvt,
+					u64 sys_addr,
+					int *chan_sel)
 {
 	int cs_found = -EINVAL;
 	unsigned range;
 
 	for (range = 0; range < DRAM_RANGES; range++) {
-
 		if (!dram_rw(pvt, range))
 			continue;
 
-		if ((get_dram_base(pvt, range)  <= sys_addr) &&
-		    (get_dram_limit(pvt, range) >= sys_addr)) {
+		if (pvt->fam == 0x15 && pvt->model >= 0x30)
+			cs_found = f15_m30h_match_to_this_node(pvt, range,
+							       sys_addr,
+							       chan_sel);
 
+		else if ((get_dram_base(pvt, range)  <= sys_addr) &&
+			 (get_dram_limit(pvt, range) >= sys_addr)) {
 			cs_found = f1x_match_to_this_node(pvt, range,
 							  sys_addr, chan_sel);
 			if (cs_found >= 0)
@@ -1624,6 +1784,17 @@ static struct amd64_family_type amd64_family_types[] = {
 			.read_dct_pci_cfg	= f15_read_dct_pci_cfg,
 		}
 	},
+	[F15_M30H_CPUS] = {
+		.ctl_name = "F15h_M30h",
+		.f1_id = PCI_DEVICE_ID_AMD_15H_M30H_NB_F1,
+		.f3_id = PCI_DEVICE_ID_AMD_15H_M30H_NB_F3,
+		.ops = {
+			.early_channel_count	= f1x_early_channel_count,
+			.map_sysaddr_to_csrow	= f1x_map_sysaddr_to_csrow,
+			.dbam_to_cs		= f16_dbam_to_chip_select,
+			.read_dct_pci_cfg	= f15_read_dct_pci_cfg,
+		}
+	},
 	[F16_CPUS] = {
 		.ctl_name = "F16h",
 		.f1_id = PCI_DEVICE_ID_AMD_16H_NB_F1,
@@ -2387,10 +2558,13 @@ static void setup_mci_misc_attrs(struct mem_ctl_info *mci,
  */
 static struct amd64_family_type *amd64_per_family_init(struct amd64_pvt *pvt)
 {
-	u8 fam = boot_cpu_data.x86;
 	struct amd64_family_type *fam_type = NULL;
 
-	switch (fam) {
+	pvt->ext_model  = boot_cpu_data.x86_model >> 4;
+	pvt->model	= boot_cpu_data.x86_model;
+	pvt->fam	= boot_cpu_data.x86;
+
+	switch (pvt->fam) {
 	case 0xf:
 		fam_type		= &amd64_family_types[K8_CPUS];
 		pvt->ops		= &amd64_family_types[K8_CPUS].ops;
@@ -2402,6 +2576,12 @@ static struct amd64_family_type *amd64_per_family_init(struct amd64_pvt *pvt)
 		break;
 
 	case 0x15:
+		if (pvt->model == 0x30) {
+			fam_type	= &amd64_family_types[F15_M30H_CPUS];
+			pvt->ops	= &amd64_family_types[F15_M30H_CPUS].ops;
+			break;
+		}
+
 		fam_type		= &amd64_family_types[F15_CPUS];
 		pvt->ops		= &amd64_family_types[F15_CPUS].ops;
 		break;
@@ -2416,10 +2596,8 @@ static struct amd64_family_type *amd64_per_family_init(struct amd64_pvt *pvt)
 		return NULL;
 	}
 
-	pvt->ext_model = boot_cpu_data.x86_model >> 4;
-
 	amd64_info("%s %sdetected (node %d).\n", fam_type->ctl_name,
-		     (fam == 0xf ?
+		     (pvt->fam == 0xf ?
 				(pvt->ext_model >= K8_REV_F  ? "revF or later "
 							     : "revE or earlier ")
 				 : ""), pvt->mc_node_id);
@@ -2638,6 +2816,14 @@ static DEFINE_PCI_DEVICE_TABLE(amd64_pci_table) = {
 	},
 	{
 		.vendor		= PCI_VENDOR_ID_AMD,
+		.device		= PCI_DEVICE_ID_AMD_15H_M30H_NB_F2,
+		.subvendor	= PCI_ANY_ID,
+		.subdevice	= PCI_ANY_ID,
+		.class		= 0,
+		.class_mask	= 0,
+	},
+	{
+		.vendor		= PCI_VENDOR_ID_AMD,
 		.device		= PCI_DEVICE_ID_AMD_16H_NB_F2,
 		.subvendor	= PCI_ANY_ID,
 		.subdevice	= PCI_ANY_ID,