Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/bp/bp (9ada9fd5) · Commits · 戴 / test

drivers/edac/Kconfig

+4 −4

Original line number	Diff line number	Diff line
		@@ -42,10 +42,10 @@ config EDAC_LEGACY_SYSFS
		config EDAC_DEBUG
		bool "Debugging"
		help
		This turns on debugging information for the entire EDAC
		sub-system. You can insert module with "debug_level=x", current
		there're four debug levels (x=0,1,2,3 from low to high).
		Usually you should select 'N'.
		This turns on debugging information for the entire EDAC subsystem.
		You do so by inserting edac_module with "edac_debug_level=x." Valid
		levels are 0-4 (from low to high) and by default it is set to 2.
		Usually you should select 'N' here.

		config EDAC_DECODE_MCE
		tristate "Decode MCEs in human-readable form (only on AMD for now)"

drivers/edac/amd64_edac.c

+113 −184

Original line number	Diff line number	Diff line
		@@ -60,7 +60,7 @@ struct scrubrate {
		{ 0x00, 0UL}, /* scrubbing off */
		};

		static int __amd64_read_pci_cfg_dword(struct pci_dev *pdev, int offset,
		int __amd64_read_pci_cfg_dword(struct pci_dev *pdev, int offset,
		u32 val, const char func)
		{
		int err = 0;
		@@ -423,7 +423,6 @@ int amd64_get_dram_hole_info(struct mem_ctl_info mci, u64 hole_base,
		u64 hole_offset, u64 hole_size)
		{
		struct amd64_pvt *pvt = mci->pvt_info;
		u64 base;

		/* only revE and later have the DRAM Hole Address Register */
		if (boot_cpu_data.x86 == 0xf && pvt->ext_model < K8_REV_E) {
		@@ -462,10 +461,8 @@ int amd64_get_dram_hole_info(struct mem_ctl_info mci, u64 hole_base,
		* addresses in the hole so that they start at 0x100000000.
		*/

		base = dhar_base(pvt);

		*hole_base = base;
		*hole_size = (0x1ull << 32) - base;
		*hole_base = dhar_base(pvt);
		hole_size = (1ULL << 32) - hole_base;

		if (boot_cpu_data.x86 > 0xf)
		*hole_offset = f10_dhar_offset(pvt);
		@@ -513,15 +510,15 @@ static u64 sys_addr_to_dram_addr(struct mem_ctl_info *mci, u64 sys_addr)
		{
		struct amd64_pvt *pvt = mci->pvt_info;
		u64 dram_base, hole_base, hole_offset, hole_size, dram_addr;
		int ret = 0;
		int ret;

		dram_base = get_dram_base(pvt, pvt->mc_node_id);

		ret = amd64_get_dram_hole_info(mci, &hole_base, &hole_offset,
		&hole_size);
		if (!ret) {
		if ((sys_addr >= (1ull << 32)) &&
		(sys_addr < ((1ull << 32) + hole_size))) {
		if ((sys_addr >= (1ULL << 32)) &&
		(sys_addr < ((1ULL << 32) + hole_size))) {
		/* use DHAR to translate SysAddr to DramAddr */
		dram_addr = sys_addr - hole_offset;

		@@ -712,10 +709,10 @@ static inline u64 input_addr_to_sys_addr(struct mem_ctl_info *mci,

		/* Map the Error address to a PAGE and PAGE OFFSET. */
		static inline void error_address_to_page_and_offset(u64 error_address,
		u32 page, u32 offset)
		struct err_info *err)
		{
		*page = (u32) (error_address >> PAGE_SHIFT);
		*offset = ((u32) error_address) & ~PAGE_MASK;
		err->page = (u32) (error_address >> PAGE_SHIFT);
		err->offset = ((u32) error_address) & ~PAGE_MASK;
		}

		/*
		@@ -1026,59 +1023,44 @@ static void read_dram_base_limit_regs(struct amd64_pvt *pvt, unsigned range)
		}

		static void k8_map_sysaddr_to_csrow(struct mem_ctl_info *mci, u64 sys_addr,
		u16 syndrome)
		struct err_info *err)
		{
		struct mem_ctl_info *src_mci;
		struct amd64_pvt *pvt = mci->pvt_info;
		int channel, csrow;
		u32 page, offset;

		error_address_to_page_and_offset(sys_addr, &page, &offset);
		error_address_to_page_and_offset(sys_addr, err);

		/*
		* Find out which node the error address belongs to. This may be
		* different from the node that detected the error.
		*/
		src_mci = find_mc_by_sys_addr(mci, sys_addr);
		if (!src_mci) {
		err->src_mci = find_mc_by_sys_addr(mci, sys_addr);
		if (!err->src_mci) {
		amd64_mc_err(mci, "failed to map error addr 0x%lx to a node\n",
		(unsigned long)sys_addr);
		edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
		page, offset, syndrome,
		-1, -1, -1,
		"failed to map error addr to a node",
		"");
		err->err_code = ERR_NODE;
		return;
		}

		/* Now map the sys_addr to a CSROW */
		csrow = sys_addr_to_csrow(src_mci, sys_addr);
		if (csrow < 0) {
		edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
		page, offset, syndrome,
		-1, -1, -1,
		"failed to map error addr to a csrow",
		"");
		err->csrow = sys_addr_to_csrow(err->src_mci, sys_addr);
		if (err->csrow < 0) {
		err->err_code = ERR_CSROW;
		return;
		}

		/* CHIPKILL enabled */
		if (pvt->nbcfg & NBCFG_CHIPKILL) {
		channel = get_channel_from_ecc_syndrome(mci, syndrome);
		if (channel < 0) {
		err->channel = get_channel_from_ecc_syndrome(mci, err->syndrome);
		if (err->channel < 0) {
		/*
		* Syndrome didn't map, so we don't know which of the
		* 2 DIMMs is in error. So we need to ID 'both' of them
		* as suspect.
		*/
		amd64_mc_warn(src_mci, "unknown syndrome 0x%04x - "
		amd64_mc_warn(err->src_mci, "unknown syndrome 0x%04x - "
		"possible error reporting race\n",
		syndrome);
		edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
		page, offset, syndrome,
		csrow, -1, -1,
		"unknown syndrome - possible error reporting race",
		"");
		err->syndrome);
		err->err_code = ERR_CHANNEL;
		return;
		}
		} else {
		@@ -1090,13 +1072,8 @@ static void k8_map_sysaddr_to_csrow(struct mem_ctl_info *mci, u64 sys_addr,
		* was obtained from email communication with someone at AMD.
		* (Wish the email was placed in this comment - norsk)
		*/
		channel = ((sys_addr & BIT(3)) != 0);
		err->channel = ((sys_addr & BIT(3)) != 0);
		}

		edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, src_mci, 1,
		page, offset, syndrome,
		csrow, channel, -1,
		"", "");
		}

		static int ddr2_cs_size(unsigned i, bool dct_width)
		@@ -1482,7 +1459,7 @@ static u64 f1x_swap_interleaved_region(struct amd64_pvt *pvt, u64 sys_addr)

		/* For a given @dram_range, check if @sys_addr falls within it. */
		static int f1x_match_to_this_node(struct amd64_pvt *pvt, unsigned range,
		u64 sys_addr, int nid, int chan_sel)
		u64 sys_addr, int *chan_sel)
		{
		int cs_found = -EINVAL;
		u64 chan_addr;
		@@ -1555,15 +1532,14 @@ static int f1x_match_to_this_node(struct amd64_pvt *pvt, unsigned range,

		cs_found = f1x_lookup_addr_in_dct(chan_addr, node_id, channel);

		if (cs_found >= 0) {
		*nid = node_id;
		if (cs_found >= 0)
		*chan_sel = channel;
		}

		return cs_found;
		}

		static int f1x_translate_sysaddr_to_cs(struct amd64_pvt *pvt, u64 sys_addr,
		int node, int chan_sel)
		int *chan_sel)
		{
		int cs_found = -EINVAL;
		unsigned range;
		@@ -1577,8 +1553,7 @@ static int f1x_translate_sysaddr_to_cs(struct amd64_pvt *pvt, u64 sys_addr,
		(get_dram_limit(pvt, range) >= sys_addr)) {

		cs_found = f1x_match_to_this_node(pvt, range,
		sys_addr, node,
		chan_sel);
		sys_addr, chan_sel);
		if (cs_found >= 0)
		break;
		}
		@@ -1594,22 +1569,15 @@ static int f1x_translate_sysaddr_to_cs(struct amd64_pvt *pvt, u64 sys_addr,
		* (MCX_ADDR).
		*/
		static void f1x_map_sysaddr_to_csrow(struct mem_ctl_info *mci, u64 sys_addr,
		u16 syndrome)
		struct err_info *err)
		{
		struct amd64_pvt *pvt = mci->pvt_info;
		u32 page, offset;
		int nid, csrow, chan = 0;

		error_address_to_page_and_offset(sys_addr, &page, &offset);

		csrow = f1x_translate_sysaddr_to_cs(pvt, sys_addr, &nid, &chan);
		error_address_to_page_and_offset(sys_addr, err);

		if (csrow < 0) {
		edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
		page, offset, syndrome,
		-1, -1, -1,
		"failed to map error addr to a csrow",
		"");
		err->csrow = f1x_translate_sysaddr_to_cs(pvt, sys_addr, &err->channel);
		if (err->csrow < 0) {
		err->err_code = ERR_CSROW;
		return;
		}

		@@ -1619,12 +1587,7 @@ static void f1x_map_sysaddr_to_csrow(struct mem_ctl_info *mci, u64 sys_addr,
		* this point.
		*/
		if (dct_ganging_enabled(pvt))
		chan = get_channel_from_ecc_syndrome(mci, syndrome);

		edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
		page, offset, syndrome,
		csrow, chan, -1,
		"", "");
		err->channel = get_channel_from_ecc_syndrome(mci, err->syndrome);
		}

		/*
		@@ -1633,14 +1596,11 @@ static void f1x_map_sysaddr_to_csrow(struct mem_ctl_info *mci, u64 sys_addr,
		*/
		static void amd64_debug_display_dimm_sizes(struct amd64_pvt *pvt, u8 ctrl)
		{
		int dimm, size0, size1, factor = 0;
		int dimm, size0, size1;
		u32 *dcsb = ctrl ? pvt->csels[1].csbases : pvt->csels[0].csbases;
		u32 dbam = ctrl ? pvt->dbam1 : pvt->dbam0;

		if (boot_cpu_data.x86 == 0xf) {
		if (pvt->dclr0 & WIDTH_128)
		factor = 1;

		/* K8 families < revF not supported yet */
		if (pvt->ext_model < K8_REV_F)
		return;
		@@ -1671,8 +1631,8 @@ static void amd64_debug_display_dimm_sizes(struct amd64_pvt *pvt, u8 ctrl)
		DBAM_DIMM(dimm, dbam));

		amd64_info(EDAC_MC ": %d: %5dMB %d: %5dMB\n",
		dimm * 2, size0 << factor,
		dimm * 2 + 1, size1 << factor);
		dimm * 2, size0,
		dimm * 2 + 1, size1);
		}
		}

		@@ -1893,101 +1853,56 @@ static int get_channel_from_ecc_syndrome(struct mem_ctl_info *mci, u16 syndrome)
		return map_err_sym_to_channel(err_sym, pvt->ecc_sym_sz);
		}

		/*
		* Handle any Correctable Errors (CEs) that have occurred. Check for valid ERROR
		* ADDRESS and process.
		*/
		static void amd64_handle_ce(struct mem_ctl_info mci, struct mce m)
		static void __log_bus_error(struct mem_ctl_info mci, struct err_info err,
		u8 ecc_type)
		{
		struct amd64_pvt *pvt = mci->pvt_info;
		u64 sys_addr;
		u16 syndrome;

		/* Ensure that the Error Address is VALID */
		if (!(m->status & MCI_STATUS_ADDRV)) {
		amd64_mc_err(mci, "HW has no ERROR_ADDRESS available\n");
		edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
		0, 0, 0,
		-1, -1, -1,
		"HW has no ERROR_ADDRESS available",
		"");
		return;
		}

		sys_addr = get_error_address(m);
		syndrome = extract_syndrome(m->status);

		amd64_mc_err(mci, "CE ERROR_ADDRESS= 0x%llx\n", sys_addr);

		pvt->ops->map_sysaddr_to_csrow(mci, sys_addr, syndrome);
		}
		enum hw_event_mc_err_type err_type;
		const char *string;

		/* Handle any Un-correctable Errors (UEs) */
		static void amd64_handle_ue(struct mem_ctl_info mci, struct mce m)
		{
		struct mem_ctl_info log_mci, src_mci = NULL;
		int csrow;
		u64 sys_addr;
		u32 page, offset;

		log_mci = mci;

		if (!(m->status & MCI_STATUS_ADDRV)) {
		amd64_mc_err(mci, "HW has no ERROR_ADDRESS available\n");
		edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
		0, 0, 0,
		-1, -1, -1,
		"HW has no ERROR_ADDRESS available",
		"");
		if (ecc_type == 2)
		err_type = HW_EVENT_ERR_CORRECTED;
		else if (ecc_type == 1)
		err_type = HW_EVENT_ERR_UNCORRECTED;
		else {
		WARN(1, "Something is rotten in the state of Denmark.\n");
		return;
		}

		sys_addr = get_error_address(m);
		error_address_to_page_and_offset(sys_addr, &page, &offset);

		/*
		* Find out which node the error address belongs to. This may be
		* different from the node that detected the error.
		*/
		src_mci = find_mc_by_sys_addr(mci, sys_addr);
		if (!src_mci) {
		amd64_mc_err(mci, "ERROR ADDRESS (0x%lx) NOT mapped to a MC\n",
		(unsigned long)sys_addr);
		edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
		page, offset, 0,
		-1, -1, -1,
		"ERROR ADDRESS NOT mapped to a MC",
		"");
		return;
		switch (err->err_code) {
		case DECODE_OK:
		string = "";
		break;
		case ERR_NODE:
		string = "Failed to map error addr to a node";
		break;
		case ERR_CSROW:
		string = "Failed to map error addr to a csrow";
		break;
		case ERR_CHANNEL:
		string = "unknown syndrome - possible error reporting race";
		break;
		default:
		string = "WTF error";
		break;
		}

		log_mci = src_mci;

		csrow = sys_addr_to_csrow(log_mci, sys_addr);
		if (csrow < 0) {
		amd64_mc_err(mci, "ERROR_ADDRESS (0x%lx) NOT mapped to CS\n",
		(unsigned long)sys_addr);
		edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
		page, offset, 0,
		-1, -1, -1,
		"ERROR ADDRESS NOT mapped to CS",
		"");
		} else {
		edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
		page, offset, 0,
		csrow, -1, -1,
		"", "");
		}
		edac_mc_handle_error(err_type, mci, 1,
		err->page, err->offset, err->syndrome,
		err->csrow, err->channel, -1,
		string, "");
		}

		static inline void __amd64_decode_bus_error(struct mem_ctl_info *mci,
		struct mce *m)
		{
		u16 ec = EC(m->status);
		u8 xec = XEC(m->status, 0x1f);
		struct amd64_pvt *pvt = mci->pvt_info;
		u8 ecc_type = (m->status >> 45) & 0x3;
		u8 xec = XEC(m->status, 0x1f);
		u16 ec = EC(m->status);
		u64 sys_addr;
		struct err_info err;

		/* Bail early out if this was an 'observed' error */
		/* Bail out early if this was an 'observed' error */
		if (PP(ec) == NBSL_PP_OBS)
		return;

		@@ -1995,10 +1910,16 @@ static inline void __amd64_decode_bus_error(struct mem_ctl_info *mci,
		if (xec && xec != F10_NBSL_EXT_ERR_ECC)
		return;

		memset(&err, 0, sizeof(err));

		sys_addr = get_error_address(m);

		if (ecc_type == 2)
		amd64_handle_ce(mci, m);
		else if (ecc_type == 1)
		amd64_handle_ue(mci, m);
		err.syndrome = extract_syndrome(m->status);

		pvt->ops->map_sysaddr_to_csrow(mci, sys_addr, &err);

		__log_bus_error(mci, &err, ecc_type);
		}

		void amd64_decode_bus_error(int node_id, struct mce *m)
		@@ -2166,6 +2087,7 @@ static u32 amd64_csrow_nr_pages(struct amd64_pvt *pvt, u8 dct, int csrow_nr)
		u32 cs_mode, nr_pages;
		u32 dbam = dct ? pvt->dbam1 : pvt->dbam0;


		/*
		* The math on this doesn't look right on the surface because x/2*4 can
		* be simplified to x*2 but this expression makes use of the fact that
		@@ -2173,13 +2095,13 @@ static u32 amd64_csrow_nr_pages(struct amd64_pvt *pvt, u8 dct, int csrow_nr)
		* number of bits to shift the DBAM register to extract the proper CSROW
		* field.
		*/
		cs_mode = (dbam >> ((csrow_nr / 2) * 4)) & 0xF;
		cs_mode = DBAM_DIMM(csrow_nr / 2, dbam);

		nr_pages = pvt->ops->dbam_to_cs(pvt, dct, cs_mode) << (20 - PAGE_SHIFT);

		edac_dbg(0, " (csrow=%d) DBAM map index= %d\n", csrow_nr, cs_mode);
		edac_dbg(0, " nr_pages/channel= %u channel-count = %d\n",
		nr_pages, pvt->channel_count);
		edac_dbg(0, "csrow: %d, channel: %d, DBAM idx: %d\n",
		csrow_nr, dct, cs_mode);
		edac_dbg(0, "nr_pages/channel: %u\n", nr_pages);

		return nr_pages;
		}
		@@ -2190,15 +2112,14 @@ static u32 amd64_csrow_nr_pages(struct amd64_pvt *pvt, u8 dct, int csrow_nr)
		*/
		static int init_csrows(struct mem_ctl_info *mci)
		{
		struct amd64_pvt *pvt = mci->pvt_info;
		struct csrow_info *csrow;
		struct dimm_info *dimm;
		struct amd64_pvt *pvt = mci->pvt_info;
		u64 base, mask;
		u32 val;
		int i, j, empty = 1;
		enum mem_type mtype;
		enum edac_type edac_mode;
		enum mem_type mtype;
		int i, j, empty = 1;
		int nr_pages = 0;
		u32 val;

		amd64_read_pci_cfg(pvt->F3, NBCFG, &val);

		@@ -2208,29 +2129,35 @@ static int init_csrows(struct mem_ctl_info *mci)
		pvt->mc_node_id, val,
		!!(val & NBCFG_CHIPKILL), !!(val & NBCFG_ECC_ENABLE));

		/*
		* We iterate over DCT0 here but we look at DCT1 in parallel, if needed.
		*/
		for_each_chip_select(i, 0, pvt) {
		csrow = mci->csrows[i];
		bool row_dct0 = !!csrow_enabled(i, 0, pvt);
		bool row_dct1 = false;

		if (boot_cpu_data.x86 != 0xf)
		row_dct1 = !!csrow_enabled(i, 1, pvt);

		if (!csrow_enabled(i, 0, pvt) && !csrow_enabled(i, 1, pvt)) {
		edac_dbg(1, "----CSROW %d VALID for MC node %d\n",
		i, pvt->mc_node_id);
		if (!row_dct0 && !row_dct1)
		continue;
		}

		csrow = mci->csrows[i];
		empty = 0;
		if (csrow_enabled(i, 0, pvt))

		edac_dbg(1, "MC node: %d, csrow: %d\n",
		pvt->mc_node_id, i);

		if (row_dct0)
		nr_pages = amd64_csrow_nr_pages(pvt, 0, i);
		if (csrow_enabled(i, 1, pvt))
		nr_pages += amd64_csrow_nr_pages(pvt, 1, i);

		get_cs_base_and_mask(pvt, i, 0, &base, &mask);
		/* 8 bytes of resolution */
		/* K8 has only one DCT */
		if (boot_cpu_data.x86 != 0xf && row_dct1)
		nr_pages += amd64_csrow_nr_pages(pvt, 1, i);

		mtype = amd64_determine_memory_type(pvt, i);

		edac_dbg(1, " for MC node %d csrow %d:\n", pvt->mc_node_id, i);
		edac_dbg(1, " nr_pages: %u\n",
		nr_pages * pvt->channel_count);
		edac_dbg(1, "Total csrow%d pages: %u\n", i, nr_pages);

		/*
		* determine whether CHIPKILL or JUST ECC or NO ECC is operating
		@@ -2247,6 +2174,7 @@ static int init_csrows(struct mem_ctl_info *mci)
		dimm->edac_mode = edac_mode;
		dimm->nr_pages = nr_pages;
		}
		csrow->nr_pages = nr_pages;
		}

		return empty;
		@@ -2591,6 +2519,7 @@ static int amd64_init_one_instance(struct pci_dev *F2)

		mci->pvt_info = pvt;
		mci->pdev = &pvt->F2->dev;
		mci->csbased = 1;

		setup_mci_misc_attrs(mci, fam_type);

drivers/edac/amd64_edac.h

+46 −13

Original line number	Diff line number	Diff line
		@@ -219,7 +219,7 @@
		#define DBAM1 0x180

		/* Extract the DIMM 'type' on the i'th DIMM from the DBAM reg value passed */
		#define DBAM_DIMM(i, reg) ((((reg) >> (4*i))) & 0xF)
		#define DBAM_DIMM(i, reg) ((((reg) >> (4*(i)))) & 0xF)

		#define DBAM_MAX_VALUE 11

		@@ -267,18 +267,20 @@
		#define online_spare_bad_dramcs(pvt, c) (((pvt)->online_spare >> (4 + 4 * (c))) & 0x7)

		#define F10_NB_ARRAY_ADDR 0xB8
		#define F10_NB_ARRAY_DRAM_ECC BIT(31)
		#define F10_NB_ARRAY_DRAM BIT(31)

		/* Bits [2:1] are used to select 16-byte section within a 64-byte cacheline */
		#define SET_NB_ARRAY_ADDRESS(section) (((section) & 0x3) << 1)
		#define SET_NB_ARRAY_ADDR(section) (((section) & 0x3) << 1)

		#define F10_NB_ARRAY_DATA 0xBC
		#define SET_NB_DRAM_INJECTION_WRITE(word, bits) \
		(BIT(((word) & 0xF) + 20) \| \
		BIT(17) \| bits)
		#define SET_NB_DRAM_INJECTION_READ(word, bits) \
		(BIT(((word) & 0xF) + 20) \| \
		BIT(16) \| bits)
		#define F10_NB_ARR_ECC_WR_REQ BIT(17)
		#define SET_NB_DRAM_INJECTION_WRITE(inj) \
		(BIT(((inj.word) & 0xF) + 20) \| \
		F10_NB_ARR_ECC_WR_REQ \| inj.bit_map)
		#define SET_NB_DRAM_INJECTION_READ(inj) \
		(BIT(((inj.word) & 0xF) + 20) \| \
		BIT(16) \| inj.bit_map)


		#define NBCAP 0xE8
		#define NBCAP_CHIPKILL BIT(4)
		@@ -374,6 +376,23 @@ struct amd64_pvt {
		struct error_injection injection;
		};

		enum err_codes {
		DECODE_OK = 0,
		ERR_NODE = -1,
		ERR_CSROW = -2,
		ERR_CHANNEL = -3,
		};

		struct err_info {
		int err_code;
		struct mem_ctl_info *src_mci;
		int csrow;
		int channel;
		u16 syndrome;
		u32 page;
		u32 offset;
		};

		static inline u64 get_dram_base(struct amd64_pvt *pvt, unsigned i)
		{
		u64 addr = ((u64)pvt->ranges[i].base.lo & 0xffff0000) << 8;
		@@ -447,7 +466,7 @@ static inline void amd64_remove_sysfs_inject_files(struct mem_ctl_info *mci)
		struct low_ops {
		int (early_channel_count) (struct amd64_pvt pvt);
		void (map_sysaddr_to_csrow) (struct mem_ctl_info mci, u64 sys_addr,
		u16 syndrome);
		struct err_info *);
		int (dbam_to_cs) (struct amd64_pvt pvt, u8 dct, unsigned cs_mode);
		int (read_dct_pci_cfg) (struct amd64_pvt pvt, int offset,
		u32 val, const char func);
		@@ -459,6 +478,8 @@ struct amd64_family_type {
		struct low_ops ops;
		};

		int __amd64_read_pci_cfg_dword(struct pci_dev *pdev, int offset,
		u32 val, const char func);
		int __amd64_write_pci_cfg_dword(struct pci_dev *pdev, int offset,
		u32 val, const char *func);

		@@ -475,3 +496,15 @@ int amd64_get_dram_hole_info(struct mem_ctl_info mci, u64 hole_base,
		u64 hole_offset, u64 hole_size);

		#define to_mci(k) container_of(k, struct mem_ctl_info, dev)

		/* Injection helpers */
		static inline void disable_caches(void *dummy)
		{
		write_cr0(read_cr0() \| X86_CR0_CD);
		wbinvd();
		}

		static inline void enable_caches(void *dummy)
		{
		write_cr0(read_cr0() & ~X86_CR0_CD);
		}

drivers/edac/amd64_edac_inj.c

+67 −61

Original line number	Diff line number	Diff line
		@@ -22,10 +22,11 @@ static ssize_t amd64_inject_section_store(struct device *dev,
		struct mem_ctl_info *mci = to_mci(dev);
		struct amd64_pvt *pvt = mci->pvt_info;
		unsigned long value;
		int ret = 0;
		int ret;

		ret = strict_strtoul(data, 10, &value);
		if (ret != -EINVAL) {
		if (ret < 0)
		return ret;

		if (value > 3) {
		amd64_warn("%s: invalid section 0x%lx\n", __func__, value);
		@@ -35,8 +36,6 @@ static ssize_t amd64_inject_section_store(struct device *dev,
		pvt->injection.section = (u32) value;
		return count;
		}
		return ret;
		}

		static ssize_t amd64_inject_word_show(struct device *dev,
		struct device_attribute *mattr,
		@@ -60,10 +59,11 @@ static ssize_t amd64_inject_word_store(struct device *dev,
		struct mem_ctl_info *mci = to_mci(dev);
		struct amd64_pvt *pvt = mci->pvt_info;
		unsigned long value;
		int ret = 0;
		int ret;

		ret = strict_strtoul(data, 10, &value);
		if (ret != -EINVAL) {
		if (ret < 0)
		return ret;

		if (value > 8) {
		amd64_warn("%s: invalid word 0x%lx\n", __func__, value);
		@@ -73,8 +73,6 @@ static ssize_t amd64_inject_word_store(struct device *dev,
		pvt->injection.word = (u32) value;
		return count;
		}
		return ret;
		}

		static ssize_t amd64_inject_ecc_vector_show(struct device *dev,
		struct device_attribute *mattr,
		@@ -97,22 +95,20 @@ static ssize_t amd64_inject_ecc_vector_store(struct device *dev,
		struct mem_ctl_info *mci = to_mci(dev);
		struct amd64_pvt *pvt = mci->pvt_info;
		unsigned long value;
		int ret = 0;
		int ret;

		ret = strict_strtoul(data, 16, &value);
		if (ret != -EINVAL) {
		if (ret < 0)
		return ret;

		if (value & 0xFFFF0000) {
		amd64_warn("%s: invalid EccVector: 0x%lx\n",
		__func__, value);
		amd64_warn("%s: invalid EccVector: 0x%lx\n", __func__, value);
		return -EINVAL;
		}

		pvt->injection.bit_map = (u32) value;
		return count;
		}
		return ret;
		}

		/*
		* Do a DRAM ECC read. Assemble staged values in the pvt area, format into
		@@ -126,29 +122,26 @@ static ssize_t amd64_inject_read_store(struct device *dev,
		struct amd64_pvt *pvt = mci->pvt_info;
		unsigned long value;
		u32 section, word_bits;
		int ret = 0;
		int ret;

		ret = strict_strtoul(data, 10, &value);
		if (ret != -EINVAL) {
		if (ret < 0)
		return ret;

		/* Form value to choose 16-byte section of cacheline */
		section = F10_NB_ARRAY_DRAM_ECC \|
		SET_NB_ARRAY_ADDRESS(pvt->injection.section);
		section = F10_NB_ARRAY_DRAM \| SET_NB_ARRAY_ADDR(pvt->injection.section);

		amd64_write_pci_cfg(pvt->F3, F10_NB_ARRAY_ADDR, section);

		word_bits = SET_NB_DRAM_INJECTION_READ(pvt->injection.word,
		pvt->injection.bit_map);
		word_bits = SET_NB_DRAM_INJECTION_READ(pvt->injection);

		/* Issue 'word' and 'bit' along with the READ request */
		amd64_write_pci_cfg(pvt->F3, F10_NB_ARRAY_DATA, word_bits);

		edac_dbg(0, "section=0x%x word_bits=0x%x\n",
		section, word_bits);
		edac_dbg(0, "section=0x%x word_bits=0x%x\n", section, word_bits);

		return count;
		}
		return ret;
		}

		/*
		* Do a DRAM ECC write. Assemble staged values in the pvt area and format into
		@@ -160,31 +153,44 @@ static ssize_t amd64_inject_write_store(struct device *dev,
		{
		struct mem_ctl_info *mci = to_mci(dev);
		struct amd64_pvt *pvt = mci->pvt_info;
		u32 section, word_bits, tmp;
		unsigned long value;
		u32 section, word_bits;
		int ret = 0;
		int ret;

		ret = strict_strtoul(data, 10, &value);
		if (ret != -EINVAL) {
		if (ret < 0)
		return ret;

		/* Form value to choose 16-byte section of cacheline */
		section = F10_NB_ARRAY_DRAM_ECC \|
		SET_NB_ARRAY_ADDRESS(pvt->injection.section);
		section = F10_NB_ARRAY_DRAM \| SET_NB_ARRAY_ADDR(pvt->injection.section);

		amd64_write_pci_cfg(pvt->F3, F10_NB_ARRAY_ADDR, section);

		word_bits = SET_NB_DRAM_INJECTION_WRITE(pvt->injection.word,
		pvt->injection.bit_map);
		word_bits = SET_NB_DRAM_INJECTION_WRITE(pvt->injection);

		pr_notice_once("Don't forget to decrease MCE polling interval in\n"
		"/sys/bus/machinecheck/devices/machinecheck<CPUNUM>/check_interval\n"
		"so that you can get the error report faster.\n");

		on_each_cpu(disable_caches, NULL, 1);

		/* Issue 'word' and 'bit' along with the READ request */
		amd64_write_pci_cfg(pvt->F3, F10_NB_ARRAY_DATA, word_bits);

		edac_dbg(0, "section=0x%x word_bits=0x%x\n",
		section, word_bits);
		retry:
		/* wait until injection happens */
		amd64_read_pci_cfg(pvt->F3, F10_NB_ARRAY_DATA, &tmp);
		if (tmp & F10_NB_ARR_ECC_WR_REQ) {
		cpu_relax();
		goto retry;
		}

		on_each_cpu(enable_caches, NULL, 1);

		edac_dbg(0, "section=0x%x word_bits=0x%x\n", section, word_bits);

		return count;
		}
		return ret;
		}

		/*
		* update NUM_INJ_ATTRS in case you add new members

drivers/edac/edac_mc.c

+27 −24

Original line number	Diff line number	Diff line
		@@ -974,20 +974,22 @@ static void edac_ce_error(struct mem_ctl_info *mci,
		long grain)
		{
		unsigned long remapped_page;
		char *msg_aux = "";

		if (*msg)
		msg_aux = " ";

		if (edac_mc_get_log_ce()) {
		if (other_detail && *other_detail)
		edac_mc_printk(mci, KERN_WARNING,
		"%d CE %s on %s (%s %s - %s)\n",
		error_count,
		msg, label, location,
		detail, other_detail);
		"%d CE %s%son %s (%s %s - %s)\n",
		error_count, msg, msg_aux, label,
		location, detail, other_detail);
		else
		edac_mc_printk(mci, KERN_WARNING,
		"%d CE %s on %s (%s %s)\n",
		error_count,
		msg, label, location,
		detail);
		"%d CE %s%son %s (%s %s)\n",
		error_count, msg, msg_aux, label,
		location, detail);
		}
		edac_inc_ce_error(mci, enable_per_layer_report, pos, error_count);

		@@ -1022,27 +1024,31 @@ static void edac_ue_error(struct mem_ctl_info *mci,
		const char *other_detail,
		const bool enable_per_layer_report)
		{
		char *msg_aux = "";

		if (*msg)
		msg_aux = " ";

		if (edac_mc_get_log_ue()) {
		if (other_detail && *other_detail)
		edac_mc_printk(mci, KERN_WARNING,
		"%d UE %s on %s (%s %s - %s)\n",
		error_count,
		msg, label, location, detail,
		other_detail);
		"%d UE %s%son %s (%s %s - %s)\n",
		error_count, msg, msg_aux, label,
		location, detail, other_detail);
		else
		edac_mc_printk(mci, KERN_WARNING,
		"%d UE %s on %s (%s %s)\n",
		error_count,
		msg, label, location, detail);
		"%d UE %s%son %s (%s %s)\n",
		error_count, msg, msg_aux, label,
		location, detail);
		}

		if (edac_mc_get_panic_on_ue()) {
		if (other_detail && *other_detail)
		panic("UE %s on %s (%s%s - %s)\n",
		msg, label, location, detail, other_detail);
		panic("UE %s%son %s (%s%s - %s)\n",
		msg, msg_aux, label, location, detail, other_detail);
		else
		panic("UE %s on %s (%s%s)\n",
		msg, label, location, detail);
		panic("UE %s%son %s (%s%s)\n",
		msg, msg_aux, label, location, detail);
		}

		edac_inc_ue_error(mci, enable_per_layer_report, pos, error_count);
		@@ -1101,10 +1107,6 @@ void edac_mc_handle_error(const enum hw_event_mc_err_type type,
		*/
		for (i = 0; i < mci->n_layers; i++) {
		if (pos[i] >= (int)mci->layers[i].size) {
		if (type == HW_EVENT_ERR_CORRECTED)
		p = "CE";
		else
		p = "UE";

		edac_mc_printk(mci, KERN_ERR,
		"INTERNAL ERROR: %s value is out of range (%d >= %d)\n",
		@@ -1136,6 +1138,7 @@ void edac_mc_handle_error(const enum hw_event_mc_err_type type,
		grain = 0;
		p = label;
		*p = '\0';

		for (i = 0; i < mci->tot_dimms; i++) {
		struct dimm_info *dimm = mci->dimms[i];

		@@ -1203,6 +1206,7 @@ void edac_mc_handle_error(const enum hw_event_mc_err_type type,

		/* Fill the RAM location data */
		p = location;

		for (i = 0; i < mci->n_layers; i++) {
		if (pos[i] < 0)
		continue;
		@@ -1215,7 +1219,6 @@ void edac_mc_handle_error(const enum hw_event_mc_err_type type,
		*(p - 1) = '\0';

		/* Report the error via the trace interface */

		grain_bits = fls_long(grain) + 1;
		trace_mc_event(type, msg, label, error_count,
		mci->mc_idx, top_layer, mid_layer, low_layer,

Admin message