Merge tag 'ras_core_for_v6.5' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull RAS updates from Borislav Petkov:

 - Add initial support for RAS hardware found on AMD server GPUs (MI200).

   Those GPUs and CPUs are connected together through the coherent
   fabric and the GPU memory controllers report errors through x86's MCA
   so EDAC needs to support them. The amd64_edac driver supports now HBM
   (High Bandwidth Memory) and thus such heterogeneous memory controller
   systems

 - Other small cleanups and improvements

* tag 'ras_core_for_v6.5' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  EDAC/amd64: Cache and use GPU node map
  EDAC/amd64: Add support for AMD heterogeneous Family 19h Model 30h-3Fh
  EDAC/amd64: Document heterogeneous system enumeration
  x86/MCE/AMD, EDAC/mce_amd: Decode UMC_V2 ECC errors
  x86/amd_nb: Re-sort and re-indent PCI defines
  x86/amd_nb: Add MI200 PCI IDs
  ras/debugfs: Fix error checking for debugfs_create_dir()
  x86/MCE: Check a hw error's address to determine proper recovery action

Documentation/driver-api/edac.rst

@@ -106,6 +106,16 @@ will occupy those chip-select rows.
 This term is avoided because it is unclear when needing to distinguish
 between chip-select rows and socket sets.
 
+* High Bandwidth Memory (HBM)
+
+HBM is a new memory type with low power consumption and ultra-wide
+communication lanes. It uses vertically stacked memory chips (DRAM dies)
+interconnected by microscopic wires called "through-silicon vias," or
+TSVs.
+
+Several stacks of HBM chips connect to the CPU or GPU through an ultra-fast
+interconnect called the "interposer". Therefore, HBM's characteristics
+are nearly indistinguishable from on-chip integrated RAM.
 
 Memory Controllers
 ------------------
@@ -176,3 +186,113 @@ nodes::
 	the L1 and L2 directories would be "edac_device_block's"
 
 .. kernel-doc:: drivers/edac/edac_device.h
+
+
+Heterogeneous system support
+----------------------------
+
+An AMD heterogeneous system is built by connecting the data fabrics of
+both CPUs and GPUs via custom xGMI links. Thus, the data fabric on the
+GPU nodes can be accessed the same way as the data fabric on CPU nodes.
+
+The MI200 accelerators are data center GPUs. They have 2 data fabrics,
+and each GPU data fabric contains four Unified Memory Controllers (UMC).
+Each UMC contains eight channels. Each UMC channel controls one 128-bit
+HBM2e (2GB) channel (equivalent to 8 X 2GB ranks).  This creates a total
+of 4096-bits of DRAM data bus.
+
+While the UMC is interfacing a 16GB (8high X 2GB DRAM) HBM stack, each UMC
+channel is interfacing 2GB of DRAM (represented as rank).
+
+Memory controllers on AMD GPU nodes can be represented in EDAC thusly:
+
+	GPU DF / GPU Node -> EDAC MC
+	GPU UMC           -> EDAC CSROW
+	GPU UMC channel   -> EDAC CHANNEL
+
+For example: a heterogeneous system with 1 AMD CPU is connected to
+4 MI200 (Aldebaran) GPUs using xGMI.
+
+Some more heterogeneous hardware details:
+
+- The CPU UMC (Unified Memory Controller) is mostly the same as the GPU UMC.
+  They have chip selects (csrows) and channels. However, the layouts are different
+  for performance, physical layout, or other reasons.
+- CPU UMCs use 1 channel, In this case UMC = EDAC channel. This follows the
+  marketing speak. CPU has X memory channels, etc.
+- CPU UMCs use up to 4 chip selects, So UMC chip select = EDAC CSROW.
+- GPU UMCs use 1 chip select, So UMC = EDAC CSROW.
+- GPU UMCs use 8 channels, So UMC channel = EDAC channel.
+
+The EDAC subsystem provides a mechanism to handle AMD heterogeneous
+systems by calling system specific ops for both CPUs and GPUs.
+
+AMD GPU nodes are enumerated in sequential order based on the PCI
+hierarchy, and the first GPU node is assumed to have a Node ID value
+following those of the CPU nodes after latter are fully populated::
+
+	$ ls /sys/devices/system/edac/mc/
+		mc0   - CPU MC node 0
+		mc1  |
+		mc2  |- GPU card[0] => node 0(mc1), node 1(mc2)
+		mc3  |
+		mc4  |- GPU card[1] => node 0(mc3), node 1(mc4)
+		mc5  |
+		mc6  |- GPU card[2] => node 0(mc5), node 1(mc6)
+		mc7  |
+		mc8  |- GPU card[3] => node 0(mc7), node 1(mc8)
+
+For example, a heterogeneous system with one AMD CPU is connected to
+four MI200 (Aldebaran) GPUs using xGMI. This topology can be represented
+via the following sysfs entries::
+
+	/sys/devices/system/edac/mc/..
+
+	CPU			# CPU node
+	├── mc 0
+
+	GPU Nodes are enumerated sequentially after CPU nodes have been populated
+	GPU card 1		# Each MI200 GPU has 2 nodes/mcs
+	├── mc 1		# GPU node 0 == mc1, Each MC node has 4 UMCs/CSROWs
+	│   ├── csrow 0		# UMC 0
+	│   │   ├── channel 0	# Each UMC has 8 channels
+	│   │   ├── channel 1   # size of each channel is 2 GB, so each UMC has 16 GB
+	│   │   ├── channel 2
+	│   │   ├── channel 3
+	│   │   ├── channel 4
+	│   │   ├── channel 5
+	│   │   ├── channel 6
+	│   │   ├── channel 7
+	│   ├── csrow 1		# UMC 1
+	│   │   ├── channel 0
+	│   │   ├── ..
+	│   │   ├── channel 7
+	│   ├── ..		..
+	│   ├── csrow 3		# UMC 3
+	│   │   ├── channel 0
+	│   │   ├── ..
+	│   │   ├── channel 7
+	│   ├── rank 0
+	│   ├── ..		..
+	│   ├── rank 31		# total 32 ranks/dimms from 4 UMCs
+	├
+	├── mc 2		# GPU node 1 == mc2
+	│   ├── ..		# each GPU has total 64 GB
+
+	GPU card 2
+	├── mc 3
+	│   ├── ..
+	├── mc 4
+	│   ├── ..
+
+	GPU card 3
+	├── mc 5
+	│   ├── ..
+	├── mc 6
+	│   ├── ..
+
+	GPU card 4
+	├── mc 7
+	│   ├── ..
+	├── mc 8
+	│   ├── ..

arch/x86/kernel/amd_nb.c

@@ -21,8 +21,11 @@
 #define PCI_DEVICE_ID_AMD_17H_M60H_ROOT		0x1630
 #define PCI_DEVICE_ID_AMD_17H_MA0H_ROOT		0x14b5
 #define PCI_DEVICE_ID_AMD_19H_M10H_ROOT		0x14a4
+#define PCI_DEVICE_ID_AMD_19H_M40H_ROOT		0x14b5
 #define PCI_DEVICE_ID_AMD_19H_M60H_ROOT		0x14d8
 #define PCI_DEVICE_ID_AMD_19H_M70H_ROOT		0x14e8
+#define PCI_DEVICE_ID_AMD_MI200_ROOT		0x14bb
+
 #define PCI_DEVICE_ID_AMD_17H_DF_F4		0x1464
 #define PCI_DEVICE_ID_AMD_17H_M10H_DF_F4	0x15ec
 #define PCI_DEVICE_ID_AMD_17H_M30H_DF_F4	0x1494
@@ -31,12 +34,12 @@
 #define PCI_DEVICE_ID_AMD_17H_MA0H_DF_F4	0x1728
 #define PCI_DEVICE_ID_AMD_19H_DF_F4		0x1654
 #define PCI_DEVICE_ID_AMD_19H_M10H_DF_F4	0x14b1
-#define PCI_DEVICE_ID_AMD_19H_M40H_ROOT	0x14b5
 #define PCI_DEVICE_ID_AMD_19H_M40H_DF_F4	0x167d
 #define PCI_DEVICE_ID_AMD_19H_M50H_DF_F4	0x166e
 #define PCI_DEVICE_ID_AMD_19H_M60H_DF_F4	0x14e4
 #define PCI_DEVICE_ID_AMD_19H_M70H_DF_F4	0x14f4
 #define PCI_DEVICE_ID_AMD_19H_M78H_DF_F4	0x12fc
+#define PCI_DEVICE_ID_AMD_MI200_DF_F4		0x14d4
 
 /* Protect the PCI config register pairs used for SMN. */
 static DEFINE_MUTEX(smn_mutex);
@@ -53,6 +56,7 @@ static const struct pci_device_id amd_root_ids[] = {
 	{ PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_19H_M40H_ROOT) },
 	{ PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_19H_M60H_ROOT) },
 	{ PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_19H_M70H_ROOT) },
+	{ PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_MI200_ROOT) },
 	{}
 };
 
@@ -81,6 +85,7 @@ static const struct pci_device_id amd_nb_misc_ids[] = {
 	{ PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_19H_M60H_DF_F3) },
 	{ PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_19H_M70H_DF_F3) },
 	{ PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_19H_M78H_DF_F3) },
+	{ PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_MI200_DF_F3) },
 	{}
 };
 
@@ -101,6 +106,7 @@ static const struct pci_device_id amd_nb_link_ids[] = {
 	{ PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_19H_M40H_DF_F4) },
 	{ PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_19H_M50H_DF_F4) },
 	{ PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_CNB17H_F4) },
+	{ PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_MI200_DF_F4) },
 	{}
 };
 

arch/x86/kernel/cpu/mce/amd.c

@@ -715,11 +715,13 @@ void mce_amd_feature_init(struct cpuinfo_x86 *c)
 
 bool amd_mce_is_memory_error(struct mce *m)
 {
+	enum smca_bank_types bank_type;
 	/* ErrCodeExt[20:16] */
 	u8 xec = (m->status >> 16) & 0x1f;
 
+	bank_type = smca_get_bank_type(m->extcpu, m->bank);
 	if (mce_flags.smca)
-		return smca_get_bank_type(m->extcpu, m->bank) == SMCA_UMC && xec == 0x0;
+		return (bank_type == SMCA_UMC || bank_type == SMCA_UMC_V2) && xec == 0x0;
 
 	return m->bank == 4 && xec == 0x8;
 }
@@ -1050,7 +1052,7 @@ static const char *get_name(unsigned int cpu, unsigned int bank, struct threshol
 	if (bank_type >= N_SMCA_BANK_TYPES)
 		return NULL;
 
-	if (b && bank_type == SMCA_UMC) {
+	if (b && (bank_type == SMCA_UMC || bank_type == SMCA_UMC_V2)) {
 		if (b->block < ARRAY_SIZE(smca_umc_block_names))
 			return smca_umc_block_names[b->block];
 		return NULL;

arch/x86/kernel/cpu/mce/core.c

@@ -1533,7 +1533,7 @@ noinstr void do_machine_check(struct pt_regs *regs)
 		/* If this triggers there is no way to recover. Die hard. */
 		BUG_ON(!on_thread_stack() || !user_mode(regs));
 
-		if (kill_current_task)
+		if (!mce_usable_address(&m))
 			queue_task_work(&m, msg, kill_me_now);
 		else
 			queue_task_work(&m, msg, kill_me_maybe);

drivers/edac/amd64_edac.c

@@ -975,6 +975,74 @@ static int sys_addr_to_csrow(struct mem_ctl_info *mci, u64 sys_addr)
 	return csrow;
 }
 
+/*
+ * See AMD PPR DF::LclNodeTypeMap
+ *
+ * This register gives information for nodes of the same type within a system.
+ *
+ * Reading this register from a GPU node will tell how many GPU nodes are in the
+ * system and what the lowest AMD Node ID value is for the GPU nodes. Use this
+ * info to fixup the Linux logical "Node ID" value set in the AMD NB code and EDAC.
+ */
+static struct local_node_map {
+	u16 node_count;
+	u16 base_node_id;
+} gpu_node_map;
+
+#define PCI_DEVICE_ID_AMD_MI200_DF_F1		0x14d1
+#define REG_LOCAL_NODE_TYPE_MAP			0x144
+
+/* Local Node Type Map (LNTM) fields */
+#define LNTM_NODE_COUNT				GENMASK(27, 16)
+#define LNTM_BASE_NODE_ID			GENMASK(11, 0)
+
+static int gpu_get_node_map(void)
+{
+	struct pci_dev *pdev;
+	int ret;
+	u32 tmp;
+
+	/*
+	 * Node ID 0 is reserved for CPUs.
+	 * Therefore, a non-zero Node ID means we've already cached the values.
+	 */
+	if (gpu_node_map.base_node_id)
+		return 0;
+
+	pdev = pci_get_device(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_MI200_DF_F1, NULL);
+	if (!pdev) {
+		ret = -ENODEV;
+		goto out;
+	}
+
+	ret = pci_read_config_dword(pdev, REG_LOCAL_NODE_TYPE_MAP, &tmp);
+	if (ret)
+		goto out;
+
+	gpu_node_map.node_count = FIELD_GET(LNTM_NODE_COUNT, tmp);
+	gpu_node_map.base_node_id = FIELD_GET(LNTM_BASE_NODE_ID, tmp);
+
+out:
+	pci_dev_put(pdev);
+	return ret;
+}
+
+static int fixup_node_id(int node_id, struct mce *m)
+{
+	/* MCA_IPID[InstanceIdHi] give the AMD Node ID for the bank. */
+	u8 nid = (m->ipid >> 44) & 0xF;
+
+	if (smca_get_bank_type(m->extcpu, m->bank) != SMCA_UMC_V2)
+		return node_id;
+
+	/* Nodes below the GPU base node are CPU nodes and don't need a fixup. */
+	if (nid < gpu_node_map.base_node_id)
+		return node_id;
+
+	/* Convert the hardware-provided AMD Node ID to a Linux logical one. */
+	return nid - gpu_node_map.base_node_id + 1;
+}
+
 /* Protect the PCI config register pairs used for DF indirect access. */
 static DEFINE_MUTEX(df_indirect_mutex);
 
@@ -1426,12 +1494,47 @@ static int umc_get_cs_mode(int dimm, u8 ctrl, struct amd64_pvt *pvt)
 	return cs_mode;
 }
 
+static int __addr_mask_to_cs_size(u32 addr_mask_orig, unsigned int cs_mode,
+				  int csrow_nr, int dimm)
+{
+	u32 msb, weight, num_zero_bits;
+	u32 addr_mask_deinterleaved;
+	int size = 0;
+
+	/*
+	 * The number of zero bits in the mask is equal to the number of bits
+	 * in a full mask minus the number of bits in the current mask.
+	 *
+	 * The MSB is the number of bits in the full mask because BIT[0] is
+	 * always 0.
+	 *
+	 * In the special 3 Rank interleaving case, a single bit is flipped
+	 * without swapping with the most significant bit. This can be handled
+	 * by keeping the MSB where it is and ignoring the single zero bit.
+	 */
+	msb = fls(addr_mask_orig) - 1;
+	weight = hweight_long(addr_mask_orig);
+	num_zero_bits = msb - weight - !!(cs_mode & CS_3R_INTERLEAVE);
+
+	/* Take the number of zero bits off from the top of the mask. */
+	addr_mask_deinterleaved = GENMASK_ULL(msb - num_zero_bits, 1);
+
+	edac_dbg(1, "CS%d DIMM%d AddrMasks:\n", csrow_nr, dimm);
+	edac_dbg(1, "  Original AddrMask: 0x%x\n", addr_mask_orig);
+	edac_dbg(1, "  Deinterleaved AddrMask: 0x%x\n", addr_mask_deinterleaved);
+
+	/* Register [31:1] = Address [39:9]. Size is in kBs here. */
+	size = (addr_mask_deinterleaved >> 2) + 1;
+
+	/* Return size in MBs. */
+	return size >> 10;
+}
+
 static int umc_addr_mask_to_cs_size(struct amd64_pvt *pvt, u8 umc,
 				    unsigned int cs_mode, int csrow_nr)
 {
-	u32 addr_mask_orig, addr_mask_deinterleaved;
-	u32 msb, weight, num_zero_bits;
 	int cs_mask_nr = csrow_nr;
+	u32 addr_mask_orig;
 	int dimm, size = 0;
 
 	/* No Chip Selects are enabled. */
@@ -1475,33 +1578,7 @@ static int umc_addr_mask_to_cs_size(struct amd64_pvt *pvt, u8 umc,
 	else
 		addr_mask_orig = pvt->csels[umc].csmasks[cs_mask_nr];
 
-	/*
-	 * The number of zero bits in the mask is equal to the number of bits
-	 * in a full mask minus the number of bits in the current mask.
-	 *
-	 * The MSB is the number of bits in the full mask because BIT[0] is
-	 * always 0.
-	 *
-	 * In the special 3 Rank interleaving case, a single bit is flipped
-	 * without swapping with the most significant bit. This can be handled
-	 * by keeping the MSB where it is and ignoring the single zero bit.
-	 */
-	msb = fls(addr_mask_orig) - 1;
-	weight = hweight_long(addr_mask_orig);
-	num_zero_bits = msb - weight - !!(cs_mode & CS_3R_INTERLEAVE);
-
-	/* Take the number of zero bits off from the top of the mask. */
-	addr_mask_deinterleaved = GENMASK_ULL(msb - num_zero_bits, 1);
-
-	edac_dbg(1, "CS%d DIMM%d AddrMasks:\n", csrow_nr, dimm);
-	edac_dbg(1, "  Original AddrMask: 0x%x\n", addr_mask_orig);
-	edac_dbg(1, "  Deinterleaved AddrMask: 0x%x\n", addr_mask_deinterleaved);
-
-	/* Register [31:1] = Address [39:9]. Size is in kBs here. */
-	size = (addr_mask_deinterleaved >> 2) + 1;
-
-	/* Return size in MBs. */
-	return size >> 10;
+	return __addr_mask_to_cs_size(addr_mask_orig, cs_mode, csrow_nr, dimm);
 }
 
 static void umc_debug_display_dimm_sizes(struct amd64_pvt *pvt, u8 ctrl)
@@ -2992,6 +3069,8 @@ static void decode_umc_error(int node_id, struct mce *m)
 	struct err_info err;
 	u64 sys_addr;
 
+	node_id = fixup_node_id(node_id, m);
+
 	mci = edac_mc_find(node_id);
 	if (!mci)
 		return;
@@ -3675,6 +3754,227 @@ static int umc_hw_info_get(struct amd64_pvt *pvt)
 	return 0;
 }
 
+/*
+ * The CPUs have one channel per UMC, so UMC number is equivalent to a
+ * channel number. The GPUs have 8 channels per UMC, so the UMC number no
+ * longer works as a channel number.
+ *
+ * The channel number within a GPU UMC is given in MCA_IPID[15:12].
+ * However, the IDs are split such that two UMC values go to one UMC, and
+ * the channel numbers are split in two groups of four.
+ *
+ * Refer to comment on gpu_get_umc_base().
+ *
+ * For example,
+ * UMC0 CH[3:0] = 0x0005[3:0]000
+ * UMC0 CH[7:4] = 0x0015[3:0]000
+ * UMC1 CH[3:0] = 0x0025[3:0]000
+ * UMC1 CH[7:4] = 0x0035[3:0]000
+ */
+static void gpu_get_err_info(struct mce *m, struct err_info *err)
+{
+	u8 ch = (m->ipid & GENMASK(31, 0)) >> 20;
+	u8 phy = ((m->ipid >> 12) & 0xf);
+
+	err->channel = ch % 2 ? phy + 4 : phy;
+	err->csrow = phy;
+}
+
+static int gpu_addr_mask_to_cs_size(struct amd64_pvt *pvt, u8 umc,
+				    unsigned int cs_mode, int csrow_nr)
+{
+	u32 addr_mask_orig = pvt->csels[umc].csmasks[csrow_nr];
+
+	return __addr_mask_to_cs_size(addr_mask_orig, cs_mode, csrow_nr, csrow_nr >> 1);
+}
+
+static void gpu_debug_display_dimm_sizes(struct amd64_pvt *pvt, u8 ctrl)
+{
+	int size, cs_mode, cs = 0;
+
+	edac_printk(KERN_DEBUG, EDAC_MC, "UMC%d chip selects:\n", ctrl);
+
+	cs_mode = CS_EVEN_PRIMARY | CS_ODD_PRIMARY;
+
+	for_each_chip_select(cs, ctrl, pvt) {
+		size = gpu_addr_mask_to_cs_size(pvt, ctrl, cs_mode, cs);
+		amd64_info(EDAC_MC ": %d: %5dMB\n", cs, size);
+	}
+}
+
+static void gpu_dump_misc_regs(struct amd64_pvt *pvt)
+{
+	struct amd64_umc *umc;
+	u32 i;
+
+	for_each_umc(i) {
+		umc = &pvt->umc[i];
+
+		edac_dbg(1, "UMC%d UMC cfg: 0x%x\n", i, umc->umc_cfg);
+		edac_dbg(1, "UMC%d SDP ctrl: 0x%x\n", i, umc->sdp_ctrl);
+		edac_dbg(1, "UMC%d ECC ctrl: 0x%x\n", i, umc->ecc_ctrl);
+		edac_dbg(1, "UMC%d All HBMs support ECC: yes\n", i);
+
+		gpu_debug_display_dimm_sizes(pvt, i);
+	}
+}
+
+static u32 gpu_get_csrow_nr_pages(struct amd64_pvt *pvt, u8 dct, int csrow_nr)
+{
+	u32 nr_pages;
+	int cs_mode = CS_EVEN_PRIMARY | CS_ODD_PRIMARY;
+
+	nr_pages   = gpu_addr_mask_to_cs_size(pvt, dct, cs_mode, csrow_nr);
+	nr_pages <<= 20 - PAGE_SHIFT;
+
+	edac_dbg(0, "csrow: %d, channel: %d\n", csrow_nr, dct);
+	edac_dbg(0, "nr_pages/channel: %u\n", nr_pages);
+
+	return nr_pages;
+}
+
+static void gpu_init_csrows(struct mem_ctl_info *mci)
+{
+	struct amd64_pvt *pvt = mci->pvt_info;
+	struct dimm_info *dimm;
+	u8 umc, cs;
+
+	for_each_umc(umc) {
+		for_each_chip_select(cs, umc, pvt) {
+			if (!csrow_enabled(cs, umc, pvt))
+				continue;
+
+			dimm = mci->csrows[umc]->channels[cs]->dimm;
+
+			edac_dbg(1, "MC node: %d, csrow: %d\n",
+				 pvt->mc_node_id, cs);
+
+			dimm->nr_pages = gpu_get_csrow_nr_pages(pvt, umc, cs);
+			dimm->edac_mode = EDAC_SECDED;
+			dimm->mtype = MEM_HBM2;
+			dimm->dtype = DEV_X16;
+			dimm->grain = 64;
+		}
+	}
+}
+
+static void gpu_setup_mci_misc_attrs(struct mem_ctl_info *mci)
+{
+	struct amd64_pvt *pvt = mci->pvt_info;
+
+	mci->mtype_cap		= MEM_FLAG_HBM2;
+	mci->edac_ctl_cap	= EDAC_FLAG_SECDED;
+
+	mci->edac_cap		= EDAC_FLAG_EC;
+	mci->mod_name		= EDAC_MOD_STR;
+	mci->ctl_name		= pvt->ctl_name;
+	mci->dev_name		= pci_name(pvt->F3);
+	mci->ctl_page_to_phys	= NULL;
+
+	gpu_init_csrows(mci);
+}
+
+/* ECC is enabled by default on GPU nodes */
+static bool gpu_ecc_enabled(struct amd64_pvt *pvt)
+{
+	return true;
+}
+
+static inline u32 gpu_get_umc_base(u8 umc, u8 channel)
+{
+	/*
+	 * On CPUs, there is one channel per UMC, so UMC numbering equals
+	 * channel numbering. On GPUs, there are eight channels per UMC,
+	 * so the channel numbering is different from UMC numbering.
+	 *
+	 * On CPU nodes channels are selected in 6th nibble
+	 * UMC chY[3:0]= [(chY*2 + 1) : (chY*2)]50000;
+	 *
+	 * On GPU nodes channels are selected in 3rd nibble
+	 * HBM chX[3:0]= [Y  ]5X[3:0]000;
+	 * HBM chX[7:4]= [Y+1]5X[3:0]000
+	 */
+	umc *= 2;
+
+	if (channel >= 4)
+		umc++;
+
+	return 0x50000 + (umc << 20) + ((channel % 4) << 12);
+}
+
+static void gpu_read_mc_regs(struct amd64_pvt *pvt)
+{
+	u8 nid = pvt->mc_node_id;
+	struct amd64_umc *umc;
+	u32 i, umc_base;
+
+	/* Read registers from each UMC */
+	for_each_umc(i) {
+		umc_base = gpu_get_umc_base(i, 0);
+		umc = &pvt->umc[i];
+
+		amd_smn_read(nid, umc_base + UMCCH_UMC_CFG, &umc->umc_cfg);
+		amd_smn_read(nid, umc_base + UMCCH_SDP_CTRL, &umc->sdp_ctrl);
+		amd_smn_read(nid, umc_base + UMCCH_ECC_CTRL, &umc->ecc_ctrl);
+	}
+}
+
+static void gpu_read_base_mask(struct amd64_pvt *pvt)
+{
+	u32 base_reg, mask_reg;
+	u32 *base, *mask;
+	int umc, cs;
+
+	for_each_umc(umc) {
+		for_each_chip_select(cs, umc, pvt) {
+			base_reg = gpu_get_umc_base(umc, cs) + UMCCH_BASE_ADDR;
+			base = &pvt->csels[umc].csbases[cs];
+
+			if (!amd_smn_read(pvt->mc_node_id, base_reg, base)) {
+				edac_dbg(0, "  DCSB%d[%d]=0x%08x reg: 0x%x\n",
+					 umc, cs, *base, base_reg);
+			}
+
+			mask_reg = gpu_get_umc_base(umc, cs) + UMCCH_ADDR_MASK;
+			mask = &pvt->csels[umc].csmasks[cs];
+
+			if (!amd_smn_read(pvt->mc_node_id, mask_reg, mask)) {
+				edac_dbg(0, "  DCSM%d[%d]=0x%08x reg: 0x%x\n",
+					 umc, cs, *mask, mask_reg);
+			}
+		}
+	}
+}
+
+static void gpu_prep_chip_selects(struct amd64_pvt *pvt)
+{
+	int umc;
+
+	for_each_umc(umc) {
+		pvt->csels[umc].b_cnt = 8;
+		pvt->csels[umc].m_cnt = 8;
+	}
+}
+
+static int gpu_hw_info_get(struct amd64_pvt *pvt)
+{
+	int ret;
+
+	ret = gpu_get_node_map();
+	if (ret)
+		return ret;
+
+	pvt->umc = kcalloc(pvt->max_mcs, sizeof(struct amd64_umc), GFP_KERNEL);
+	if (!pvt->umc)
+		return -ENOMEM;
+
+	gpu_prep_chip_selects(pvt);
+	gpu_read_base_mask(pvt);
+	gpu_read_mc_regs(pvt);
+
+	return 0;
+}
+
 static void hw_info_put(struct amd64_pvt *pvt)
 {
 	pci_dev_put(pvt->F1);
@@ -3690,6 +3990,14 @@ static struct low_ops umc_ops = {
 	.get_err_info			= umc_get_err_info,
 };
 
+static struct low_ops gpu_ops = {
+	.hw_info_get			= gpu_hw_info_get,
+	.ecc_enabled			= gpu_ecc_enabled,
+	.setup_mci_misc_attrs		= gpu_setup_mci_misc_attrs,
+	.dump_misc_regs			= gpu_dump_misc_regs,
+	.get_err_info			= gpu_get_err_info,
+};
+
 /* Use Family 16h versions for defaults and adjust as needed below. */
 static struct low_ops dct_ops = {
 	.map_sysaddr_to_csrow		= f1x_map_sysaddr_to_csrow,
@@ -3813,6 +4121,16 @@ static int per_family_init(struct amd64_pvt *pvt)
 		case 0x20 ... 0x2f:
 			pvt->ctl_name			= "F19h_M20h";
 			break;
+		case 0x30 ... 0x3f:
+			if (pvt->F3->device == PCI_DEVICE_ID_AMD_MI200_DF_F3) {
+				pvt->ctl_name		= "MI200";
+				pvt->max_mcs		= 4;
+				pvt->ops		= &gpu_ops;
+			} else {
+				pvt->ctl_name		= "F19h_M30h";
+				pvt->max_mcs		= 8;
+			}
+			break;
 		case 0x50 ... 0x5f:
 			pvt->ctl_name			= "F19h_M50h";
 			break;
@@ -3854,11 +4172,17 @@ static int init_one_instance(struct amd64_pvt *pvt)
 	struct edac_mc_layer layers[2];
 	int ret = -ENOMEM;
 
+	/*
+	 * For Heterogeneous family EDAC CHIP_SELECT and CHANNEL layers should
+	 * be swapped to fit into the layers.
+	 */
 	layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
-	layers[0].size = pvt->csels[0].b_cnt;
+	layers[0].size = (pvt->F3->device == PCI_DEVICE_ID_AMD_MI200_DF_F3) ?
+			 pvt->max_mcs : pvt->csels[0].b_cnt;
 	layers[0].is_virt_csrow = true;
 	layers[1].type = EDAC_MC_LAYER_CHANNEL;
-	layers[1].size = pvt->max_mcs;
+	layers[1].size = (pvt->F3->device == PCI_DEVICE_ID_AMD_MI200_DF_F3) ?
+			 pvt->csels[0].b_cnt : pvt->max_mcs;
 	layers[1].is_virt_csrow = false;
 
 	mci = edac_mc_alloc(pvt->mc_node_id, ARRAY_SIZE(layers), layers, 0);

drivers/edac/amd64_edac.h

@@ -16,6 +16,7 @@
 #include <linux/slab.h>
 #include <linux/mmzone.h>
 #include <linux/edac.h>
+#include <linux/bitfield.h>
 #include <asm/cpu_device_id.h>
 #include <asm/msr.h>
 #include "edac_module.h"

drivers/edac/mce_amd.c

@@ -1186,7 +1186,8 @@ static void decode_smca_error(struct mce *m)
 	if (xec < smca_mce_descs[bank_type].num_descs)
 		pr_cont(", %s.\n", smca_mce_descs[bank_type].descs[xec]);
 
-	if (bank_type == SMCA_UMC && xec == 0 && decode_dram_ecc)
+	if ((bank_type == SMCA_UMC || bank_type == SMCA_UMC_V2) &&
+	    xec == 0 && decode_dram_ecc)
 		decode_dram_ecc(topology_die_id(m->extcpu), m);
 }
 

drivers/ras/debugfs.c

@@ -46,7 +46,7 @@ int __init ras_add_daemon_trace(void)
 
 	fentry = debugfs_create_file("daemon_active", S_IRUSR, ras_debugfs_dir,
 				     NULL, &trace_fops);
-	if (!fentry)
+	if (IS_ERR(fentry))
 		return -ENODEV;
 
 	return 0;

include/linux/pci_ids.h

@@ -568,6 +568,7 @@
 #define PCI_DEVICE_ID_AMD_19H_M60H_DF_F3 0x14e3
 #define PCI_DEVICE_ID_AMD_19H_M70H_DF_F3 0x14f3
 #define PCI_DEVICE_ID_AMD_19H_M78H_DF_F3 0x12fb
+#define PCI_DEVICE_ID_AMD_MI200_DF_F3	0x14d3
 #define PCI_DEVICE_ID_AMD_CNB17H_F3	0x1703
 #define PCI_DEVICE_ID_AMD_LANCE		0x2000
 #define PCI_DEVICE_ID_AMD_LANCE_HOME	0x2001