From 4ec656bdf43a13a655a8259b79dd63bc1f0b1e41 Mon Sep 17 00:00:00 2001 From: Tony Luck Date: Sat, 20 Aug 2016 16:27:58 -0700 Subject: [PATCH] EDAC, skx_edac: Add EDAC driver for Skylake This is an entirely new driver instead of yet another set of patches to sb_edac.c because: 1) Mapping from PCI devices to socket/memory controller is significantly different. Skylake scatters devices on a socket across a number of PCI buses. 2) There is an extra level of interleaving via the "mcroute" register that would be a little messy to squeeze into the old driver. 3) Validation is getting too expensive. Changes to sb_edac need to be checked against Sandy Bridge, Ivy Bridge, Haswell, Broadwell and Knights Landing. Acked-by: Aristeu Rozanski Acked-by: Borislav Petkov Signed-off-by: Tony Luck Signed-off-by: Linus Torvalds --- MAINTAINERS | 6 + drivers/edac/Kconfig | 8 + drivers/edac/Makefile | 1 + drivers/edac/skx_edac.c | 1121 +++++++++++++++++++++++++++++++++++++++ 4 files changed, 1136 insertions(+) create mode 100644 drivers/edac/skx_edac.c diff --git a/MAINTAINERS b/MAINTAINERS index a306795a7450..0bbe4b105c34 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -4525,6 +4525,12 @@ L: linux-edac@vger.kernel.org S: Maintained F: drivers/edac/sb_edac.c +EDAC-SKYLAKE +M: Tony Luck +L: linux-edac@vger.kernel.org +S: Maintained +F: drivers/edac/skx_edac.c + EDAC-XGENE APPLIED MICRO (APM) X-GENE SOC EDAC M: Loc Ho diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig index d0c1dab9b435..dff1a4a6dc1b 100644 --- a/drivers/edac/Kconfig +++ b/drivers/edac/Kconfig @@ -251,6 +251,14 @@ config EDAC_SBRIDGE Support for error detection and correction the Intel Sandy Bridge, Ivy Bridge and Haswell Integrated Memory Controllers. +config EDAC_SKX + tristate "Intel Skylake server Integrated MC" + depends on EDAC_MM_EDAC && PCI && X86_64 && X86_MCE_INTEL + depends on PCI_MMCONFIG + help + Support for error detection and correction the Intel + Skylake server Integrated Memory Controllers. + config EDAC_MPC85XX tristate "Freescale MPC83xx / MPC85xx" depends on EDAC_MM_EDAC && FSL_SOC diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile index f9e4a3e0e6e9..986049925b08 100644 --- a/drivers/edac/Makefile +++ b/drivers/edac/Makefile @@ -31,6 +31,7 @@ obj-$(CONFIG_EDAC_I5400) += i5400_edac.o obj-$(CONFIG_EDAC_I7300) += i7300_edac.o obj-$(CONFIG_EDAC_I7CORE) += i7core_edac.o obj-$(CONFIG_EDAC_SBRIDGE) += sb_edac.o +obj-$(CONFIG_EDAC_SKX) += skx_edac.o obj-$(CONFIG_EDAC_E7XXX) += e7xxx_edac.o obj-$(CONFIG_EDAC_E752X) += e752x_edac.o obj-$(CONFIG_EDAC_I82443BXGX) += i82443bxgx_edac.o diff --git a/drivers/edac/skx_edac.c b/drivers/edac/skx_edac.c new file mode 100644 index 000000000000..0ff4878c2aa1 --- /dev/null +++ b/drivers/edac/skx_edac.c @@ -0,0 +1,1121 @@ +/* + * EDAC driver for Intel(R) Xeon(R) Skylake processors + * Copyright (c) 2016, Intel Corporation. + * + * This program is free software; you can redistribute it and/or modify it + * under the terms and conditions of the GNU General Public License, + * version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "edac_core.h" + +#define SKX_REVISION " Ver: 1.0 " + +/* + * Debug macros + */ +#define skx_printk(level, fmt, arg...) \ + edac_printk(level, "skx", fmt, ##arg) + +#define skx_mc_printk(mci, level, fmt, arg...) \ + edac_mc_chipset_printk(mci, level, "skx", fmt, ##arg) + +/* + * Get a bit field at register value , from bit to bit + */ +#define GET_BITFIELD(v, lo, hi) \ + (((v) & GENMASK_ULL((hi), (lo))) >> (lo)) + +static LIST_HEAD(skx_edac_list); + +static u64 skx_tolm, skx_tohm; + +#define NUM_IMC 2 /* memory controllers per socket */ +#define NUM_CHANNELS 3 /* channels per memory controller */ +#define NUM_DIMMS 2 /* Max DIMMS per channel */ + +#define MASK26 0x3FFFFFF /* Mask for 2^26 */ +#define MASK29 0x1FFFFFFF /* Mask for 2^29 */ + +/* + * Each cpu socket contains some pci devices that provide global + * information, and also some that are local to each of the two + * memory controllers on the die. + */ +struct skx_dev { + struct list_head list; + u8 bus[4]; + struct pci_dev *sad_all; + struct pci_dev *util_all; + u32 mcroute; + struct skx_imc { + struct mem_ctl_info *mci; + u8 mc; /* system wide mc# */ + u8 lmc; /* socket relative mc# */ + u8 src_id, node_id; + struct skx_channel { + struct pci_dev *cdev; + struct skx_dimm { + u8 close_pg; + u8 bank_xor_enable; + u8 fine_grain_bank; + u8 rowbits; + u8 colbits; + } dimms[NUM_DIMMS]; + } chan[NUM_CHANNELS]; + } imc[NUM_IMC]; +}; +static int skx_num_sockets; + +struct skx_pvt { + struct skx_imc *imc; +}; + +struct decoded_addr { + struct skx_dev *dev; + u64 addr; + int socket; + int imc; + int channel; + u64 chan_addr; + int sktways; + int chanways; + int dimm; + int rank; + int channel_rank; + u64 rank_address; + int row; + int column; + int bank_address; + int bank_group; +}; + +static struct skx_dev *get_skx_dev(u8 bus, u8 idx) +{ + struct skx_dev *d; + + list_for_each_entry(d, &skx_edac_list, list) { + if (d->bus[idx] == bus) + return d; + } + + return NULL; +} + +enum munittype { + CHAN0, CHAN1, CHAN2, SAD_ALL, UTIL_ALL, SAD +}; + +struct munit { + u16 did; + u16 devfn[NUM_IMC]; + u8 busidx; + u8 per_socket; + enum munittype mtype; +}; + +/* + * List of PCI device ids that we need together with some device + * number and function numbers to tell which memory controller the + * device belongs to. + */ +static const struct munit skx_all_munits[] = { + { 0x2054, { }, 1, 1, SAD_ALL }, + { 0x2055, { }, 1, 1, UTIL_ALL }, + { 0x2040, { PCI_DEVFN(10, 0), PCI_DEVFN(12, 0) }, 2, 2, CHAN0 }, + { 0x2044, { PCI_DEVFN(10, 4), PCI_DEVFN(12, 4) }, 2, 2, CHAN1 }, + { 0x2048, { PCI_DEVFN(11, 0), PCI_DEVFN(13, 0) }, 2, 2, CHAN2 }, + { 0x208e, { }, 1, 0, SAD }, + { } +}; + +/* + * We use the per-socket device 0x2016 to count how many sockets are present, + * and to detemine which PCI buses are associated with each socket. Allocate + * and build the full list of all the skx_dev structures that we need here. + */ +static int get_all_bus_mappings(void) +{ + struct pci_dev *pdev, *prev; + struct skx_dev *d; + u32 reg; + int ndev = 0; + + prev = NULL; + for (;;) { + pdev = pci_get_device(PCI_VENDOR_ID_INTEL, 0x2016, prev); + if (!pdev) + break; + ndev++; + d = kzalloc(sizeof(*d), GFP_KERNEL); + if (!d) { + pci_dev_put(pdev); + return -ENOMEM; + } + pci_read_config_dword(pdev, 0xCC, ®); + d->bus[0] = GET_BITFIELD(reg, 0, 7); + d->bus[1] = GET_BITFIELD(reg, 8, 15); + d->bus[2] = GET_BITFIELD(reg, 16, 23); + d->bus[3] = GET_BITFIELD(reg, 24, 31); + edac_dbg(2, "busses: %x, %x, %x, %x\n", + d->bus[0], d->bus[1], d->bus[2], d->bus[3]); + list_add_tail(&d->list, &skx_edac_list); + skx_num_sockets++; + prev = pdev; + } + + return ndev; +} + +static int get_all_munits(const struct munit *m) +{ + struct pci_dev *pdev, *prev; + struct skx_dev *d; + u32 reg; + int i = 0, ndev = 0; + + prev = NULL; + for (;;) { + pdev = pci_get_device(PCI_VENDOR_ID_INTEL, m->did, prev); + if (!pdev) + break; + ndev++; + if (m->per_socket == NUM_IMC) { + for (i = 0; i < NUM_IMC; i++) + if (m->devfn[i] == pdev->devfn) + break; + if (i == NUM_IMC) + goto fail; + } + d = get_skx_dev(pdev->bus->number, m->busidx); + if (!d) + goto fail; + + /* Be sure that the device is enabled */ + if (unlikely(pci_enable_device(pdev) < 0)) { + skx_printk(KERN_ERR, + "Couldn't enable %04x:%04x\n", PCI_VENDOR_ID_INTEL, m->did); + goto fail; + } + + switch (m->mtype) { + case CHAN0: case CHAN1: case CHAN2: + pci_dev_get(pdev); + d->imc[i].chan[m->mtype].cdev = pdev; + break; + case SAD_ALL: + pci_dev_get(pdev); + d->sad_all = pdev; + break; + case UTIL_ALL: + pci_dev_get(pdev); + d->util_all = pdev; + break; + case SAD: + /* + * one of these devices per core, including cores + * that don't exist on this SKU. Ignore any that + * read a route table of zero, make sure all the + * non-zero values match. + */ + pci_read_config_dword(pdev, 0xB4, ®); + if (reg != 0) { + if (d->mcroute == 0) + d->mcroute = reg; + else if (d->mcroute != reg) { + skx_printk(KERN_ERR, + "mcroute mismatch\n"); + goto fail; + } + } + ndev--; + break; + } + + prev = pdev; + } + + return ndev; +fail: + pci_dev_put(pdev); + return -ENODEV; +} + +const struct x86_cpu_id skx_cpuids[] = { + { X86_VENDOR_INTEL, 6, 0x55, 0, 0 }, /* Skylake */ + { } +}; +MODULE_DEVICE_TABLE(x86cpu, skx_cpuids); + +static u8 get_src_id(struct skx_dev *d) +{ + u32 reg; + + pci_read_config_dword(d->util_all, 0xF0, ®); + + return GET_BITFIELD(reg, 12, 14); +} + +static u8 skx_get_node_id(struct skx_dev *d) +{ + u32 reg; + + pci_read_config_dword(d->util_all, 0xF4, ®); + + return GET_BITFIELD(reg, 0, 2); +} + +static int get_dimm_attr(u32 reg, int lobit, int hibit, int add, int minval, + int maxval, char *name) +{ + u32 val = GET_BITFIELD(reg, lobit, hibit); + + if (val < minval || val > maxval) { + edac_dbg(2, "bad %s = %d (raw=%x)\n", name, val, reg); + return -EINVAL; + } + return val + add; +} + +#define IS_DIMM_PRESENT(mtr) GET_BITFIELD((mtr), 15, 15) + +#define numrank(reg) get_dimm_attr((reg), 12, 13, 0, 1, 2, "ranks") +#define numrow(reg) get_dimm_attr((reg), 2, 4, 12, 1, 6, "rows") +#define numcol(reg) get_dimm_attr((reg), 0, 1, 10, 0, 2, "cols") + +static int get_width(u32 mtr) +{ + switch (GET_BITFIELD(mtr, 8, 9)) { + case 0: + return DEV_X4; + case 1: + return DEV_X8; + case 2: + return DEV_X16; + } + return DEV_UNKNOWN; +} + +static int skx_get_hi_lo(void) +{ + struct pci_dev *pdev; + u32 reg; + + pdev = pci_get_device(PCI_VENDOR_ID_INTEL, 0x2034, NULL); + if (!pdev) { + edac_dbg(0, "Can't get tolm/tohm\n"); + return -ENODEV; + } + + pci_read_config_dword(pdev, 0xD0, ®); + skx_tolm = reg; + pci_read_config_dword(pdev, 0xD4, ®); + skx_tohm = reg; + pci_read_config_dword(pdev, 0xD8, ®); + skx_tohm |= (u64)reg << 32; + + pci_dev_put(pdev); + edac_dbg(2, "tolm=%llx tohm=%llx\n", skx_tolm, skx_tohm); + + return 0; +} + +static int get_dimm_info(u32 mtr, u32 amap, struct dimm_info *dimm, + struct skx_imc *imc, int chan, int dimmno) +{ + int banks = 16, ranks, rows, cols, npages; + u64 size; + + if (!IS_DIMM_PRESENT(mtr)) + return 0; + ranks = numrank(mtr); + rows = numrow(mtr); + cols = numcol(mtr); + + /* + * Compute size in 8-byte (2^3) words, then shift to MiB (2^20) + */ + size = ((1ull << (rows + cols + ranks)) * banks) >> (20 - 3); + npages = MiB_TO_PAGES(size); + + edac_dbg(0, "mc#%d: channel %d, dimm %d, %lld Mb (%d pages) bank: %d, rank: %d, row: %#x, col: %#x\n", + imc->mc, chan, dimmno, size, npages, + banks, ranks, rows, cols); + + imc->chan[chan].dimms[dimmno].close_pg = GET_BITFIELD(mtr, 0, 0); + imc->chan[chan].dimms[dimmno].bank_xor_enable = GET_BITFIELD(mtr, 9, 9); + imc->chan[chan].dimms[dimmno].fine_grain_bank = GET_BITFIELD(amap, 0, 0); + imc->chan[chan].dimms[dimmno].rowbits = rows; + imc->chan[chan].dimms[dimmno].colbits = cols; + + dimm->nr_pages = npages; + dimm->grain = 32; + dimm->dtype = get_width(mtr); + dimm->mtype = MEM_DDR4; + dimm->edac_mode = EDAC_SECDED; /* likely better than this */ + snprintf(dimm->label, sizeof(dimm->label), "CPU_SrcID#%u_MC#%u_Chan#%u_DIMM#%u", + imc->src_id, imc->lmc, chan, dimmno); + + return 1; +} + +#define SKX_GET_MTMTR(dev, reg) \ + pci_read_config_dword((dev), 0x87c, ®) + +static bool skx_check_ecc(struct pci_dev *pdev) +{ + u32 mtmtr; + + SKX_GET_MTMTR(pdev, mtmtr); + + return !!GET_BITFIELD(mtmtr, 2, 2); +} + +static int skx_get_dimm_config(struct mem_ctl_info *mci) +{ + struct skx_pvt *pvt = mci->pvt_info; + struct skx_imc *imc = pvt->imc; + struct dimm_info *dimm; + int i, j; + u32 mtr, amap; + int ndimms; + + for (i = 0; i < NUM_CHANNELS; i++) { + ndimms = 0; + pci_read_config_dword(imc->chan[i].cdev, 0x8C, &amap); + for (j = 0; j < NUM_DIMMS; j++) { + dimm = EDAC_DIMM_PTR(mci->layers, mci->dimms, + mci->n_layers, i, j, 0); + pci_read_config_dword(imc->chan[i].cdev, + 0x80 + 4*j, &mtr); + ndimms += get_dimm_info(mtr, amap, dimm, imc, i, j); + } + if (ndimms && !skx_check_ecc(imc->chan[0].cdev)) { + skx_printk(KERN_ERR, "ECC is disabled on imc %d\n", imc->mc); + return -ENODEV; + } + } + + return 0; +} + +static void skx_unregister_mci(struct skx_imc *imc) +{ + struct mem_ctl_info *mci = imc->mci; + + if (!mci) + return; + + edac_dbg(0, "MC%d: mci = %p\n", imc->mc, mci); + + /* Remove MC sysfs nodes */ + edac_mc_del_mc(mci->pdev); + + edac_dbg(1, "%s: free mci struct\n", mci->ctl_name); + kfree(mci->ctl_name); + edac_mc_free(mci); +} + +static int skx_register_mci(struct skx_imc *imc) +{ + struct mem_ctl_info *mci; + struct edac_mc_layer layers[2]; + struct pci_dev *pdev = imc->chan[0].cdev; + struct skx_pvt *pvt; + int rc; + + /* allocate a new MC control structure */ + layers[0].type = EDAC_MC_LAYER_CHANNEL; + layers[0].size = NUM_CHANNELS; + layers[0].is_virt_csrow = false; + layers[1].type = EDAC_MC_LAYER_SLOT; + layers[1].size = NUM_DIMMS; + layers[1].is_virt_csrow = true; + mci = edac_mc_alloc(imc->mc, ARRAY_SIZE(layers), layers, + sizeof(struct skx_pvt)); + + if (unlikely(!mci)) + return -ENOMEM; + + edac_dbg(0, "MC#%d: mci = %p\n", imc->mc, mci); + + /* Associate skx_dev and mci for future usage */ + imc->mci = mci; + pvt = mci->pvt_info; + pvt->imc = imc; + + mci->ctl_name = kasprintf(GFP_KERNEL, "Skylake Socket#%d IMC#%d", + imc->node_id, imc->lmc); + mci->mtype_cap = MEM_FLAG_DDR4; + mci->edac_ctl_cap = EDAC_FLAG_NONE; + mci->edac_cap = EDAC_FLAG_NONE; + mci->mod_name = "skx_edac.c"; + mci->dev_name = pci_name(imc->chan[0].cdev); + mci->mod_ver = SKX_REVISION; + mci->ctl_page_to_phys = NULL; + + rc = skx_get_dimm_config(mci); + if (rc < 0) + goto fail; + + /* record ptr to the generic device */ + mci->pdev = &pdev->dev; + + /* add this new MC control structure to EDAC's list of MCs */ + if (unlikely(edac_mc_add_mc(mci))) { + edac_dbg(0, "MC: failed edac_mc_add_mc()\n"); + rc = -EINVAL; + goto fail; + } + + return 0; + +fail: + kfree(mci->ctl_name); + edac_mc_free(mci); + imc->mci = NULL; + return rc; +} + +#define SKX_MAX_SAD 24 + +#define SKX_GET_SAD(d, i, reg) \ + pci_read_config_dword((d)->sad_all, 0x60 + 8 * (i), ®) +#define SKX_GET_ILV(d, i, reg) \ + pci_read_config_dword((d)->sad_all, 0x64 + 8 * (i), ®) + +#define SKX_SAD_MOD3MODE(sad) GET_BITFIELD((sad), 30, 31) +#define SKX_SAD_MOD3(sad) GET_BITFIELD((sad), 27, 27) +#define SKX_SAD_LIMIT(sad) (((u64)GET_BITFIELD((sad), 7, 26) << 26) | MASK26) +#define SKX_SAD_MOD3ASMOD2(sad) GET_BITFIELD((sad), 5, 6) +#define SKX_SAD_ATTR(sad) GET_BITFIELD((sad), 3, 4) +#define SKX_SAD_INTERLEAVE(sad) GET_BITFIELD((sad), 1, 2) +#define SKX_SAD_ENABLE(sad) GET_BITFIELD((sad), 0, 0) + +#define SKX_ILV_REMOTE(tgt) (((tgt) & 8) == 0) +#define SKX_ILV_TARGET(tgt) ((tgt) & 7) + +static bool skx_sad_decode(struct decoded_addr *res) +{ + struct skx_dev *d = list_first_entry(&skx_edac_list, typeof(*d), list); + u64 addr = res->addr; + int i, idx, tgt, lchan, shift; + u32 sad, ilv; + u64 limit, prev_limit; + int remote = 0; + + /* Simple sanity check for I/O space or out of range */ + if (addr >= skx_tohm || (addr >= skx_tolm && addr < BIT_ULL(32))) { + edac_dbg(0, "Address %llx out of range\n", addr); + return false; + } + +restart: + prev_limit = 0; + for (i = 0; i < SKX_MAX_SAD; i++) { + SKX_GET_SAD(d, i, sad); + limit = SKX_SAD_LIMIT(sad); + if (SKX_SAD_ENABLE(sad)) { + if (addr >= prev_limit && addr <= limit) + goto sad_found; + } + prev_limit = limit + 1; + } + edac_dbg(0, "No SAD entry for %llx\n", addr); + return false; + +sad_found: + SKX_GET_ILV(d, i, ilv); + + switch (SKX_SAD_INTERLEAVE(sad)) { + case 0: + idx = GET_BITFIELD(addr, 6, 8); + break; + case 1: + idx = GET_BITFIELD(addr, 8, 10); + break; + case 2: + idx = GET_BITFIELD(addr, 12, 14); + break; + case 3: + idx = GET_BITFIELD(addr, 30, 32); + break; + } + + tgt = GET_BITFIELD(ilv, 4 * idx, 4 * idx + 3); + + /* If point to another node, find it and start over */ + if (SKX_ILV_REMOTE(tgt)) { + if (remote) { + edac_dbg(0, "Double remote!\n"); + return false; + } + remote = 1; + list_for_each_entry(d, &skx_edac_list, list) { + if (d->imc[0].src_id == SKX_ILV_TARGET(tgt)) + goto restart; + } + edac_dbg(0, "Can't find node %d\n", SKX_ILV_TARGET(tgt)); + return false; + } + + if (SKX_SAD_MOD3(sad) == 0) + lchan = SKX_ILV_TARGET(tgt); + else { + switch (SKX_SAD_MOD3MODE(sad)) { + case 0: + shift = 6; + break; + case 1: + shift = 8; + break; + case 2: + shift = 12; + break; + default: + edac_dbg(0, "illegal mod3mode\n"); + return false; + } + switch (SKX_SAD_MOD3ASMOD2(sad)) { + case 0: + lchan = (addr >> shift) % 3; + break; + case 1: + lchan = (addr >> shift) % 2; + break; + case 2: + lchan = (addr >> shift) % 2; + lchan = (lchan << 1) | ~lchan; + break; + case 3: + lchan = ((addr >> shift) % 2) << 1; + break; + } + lchan = (lchan << 1) | (SKX_ILV_TARGET(tgt) & 1); + } + + res->dev = d; + res->socket = d->imc[0].src_id; + res->imc = GET_BITFIELD(d->mcroute, lchan * 3, lchan * 3 + 2); + res->channel = GET_BITFIELD(d->mcroute, lchan * 2 + 18, lchan * 2 + 19); + + edac_dbg(2, "%llx: socket=%d imc=%d channel=%d\n", + res->addr, res->socket, res->imc, res->channel); + return true; +} + +#define SKX_MAX_TAD 8 + +#define SKX_GET_TADBASE(d, mc, i, reg) \ + pci_read_config_dword((d)->imc[mc].chan[0].cdev, 0x850 + 4 * (i), ®) +#define SKX_GET_TADWAYNESS(d, mc, i, reg) \ + pci_read_config_dword((d)->imc[mc].chan[0].cdev, 0x880 + 4 * (i), ®) +#define SKX_GET_TADCHNILVOFFSET(d, mc, ch, i, reg) \ + pci_read_config_dword((d)->imc[mc].chan[ch].cdev, 0x90 + 4 * (i), ®) + +#define SKX_TAD_BASE(b) ((u64)GET_BITFIELD((b), 12, 31) << 26) +#define SKX_TAD_SKT_GRAN(b) GET_BITFIELD((b), 4, 5) +#define SKX_TAD_CHN_GRAN(b) GET_BITFIELD((b), 6, 7) +#define SKX_TAD_LIMIT(b) (((u64)GET_BITFIELD((b), 12, 31) << 26) | MASK26) +#define SKX_TAD_OFFSET(b) ((u64)GET_BITFIELD((b), 4, 23) << 26) +#define SKX_TAD_SKTWAYS(b) (1 << GET_BITFIELD((b), 10, 11)) +#define SKX_TAD_CHNWAYS(b) (GET_BITFIELD((b), 8, 9) + 1) + +/* which bit used for both socket and channel interleave */ +static int skx_granularity[] = { 6, 8, 12, 30 }; + +static u64 skx_do_interleave(u64 addr, int shift, int ways, u64 lowbits) +{ + addr >>= shift; + addr /= ways; + addr <<= shift; + + return addr | (lowbits & ((1ull << shift) - 1)); +} + +static bool skx_tad_decode(struct decoded_addr *res) +{ + int i; + u32 base, wayness, chnilvoffset; + int skt_interleave_bit, chn_interleave_bit; + u64 channel_addr; + + for (i = 0; i < SKX_MAX_TAD; i++) { + SKX_GET_TADBASE(res->dev, res->imc, i, base); + SKX_GET_TADWAYNESS(res->dev, res->imc, i, wayness); + if (SKX_TAD_BASE(base) <= res->addr && res->addr <= SKX_TAD_LIMIT(wayness)) + goto tad_found; + } + edac_dbg(0, "No TAD entry for %llx\n", res->addr); + return false; + +tad_found: + res->sktways = SKX_TAD_SKTWAYS(wayness); + res->chanways = SKX_TAD_CHNWAYS(wayness); + skt_interleave_bit = skx_granularity[SKX_TAD_SKT_GRAN(base)]; + chn_interleave_bit = skx_granularity[SKX_TAD_CHN_GRAN(base)]; + + SKX_GET_TADCHNILVOFFSET(res->dev, res->imc, res->channel, i, chnilvoffset); + channel_addr = res->addr - SKX_TAD_OFFSET(chnilvoffset); + + if (res->chanways == 3 && skt_interleave_bit > chn_interleave_bit) { + /* Must handle channel first, then socket */ + channel_addr = skx_do_interleave(channel_addr, chn_interleave_bit, + res->chanways, channel_addr); + channel_addr = skx_do_interleave(channel_addr, skt_interleave_bit, + res->sktways, channel_addr); + } else { + /* Handle socket then channel. Preserve low bits from original address */ + channel_addr = skx_do_interleave(channel_addr, skt_interleave_bit, + res->sktways, res->addr); + channel_addr = skx_do_interleave(channel_addr, chn_interleave_bit, + res->chanways, res->addr); + } + + res->chan_addr = channel_addr; + + edac_dbg(2, "%llx: chan_addr=%llx sktways=%d chanways=%d\n", + res->addr, res->chan_addr, res->sktways, res->chanways); + return true; +} + +#define SKX_MAX_RIR 4 + +#define SKX_GET_RIRWAYNESS(d, mc, ch, i, reg) \ + pci_read_config_dword((d)->imc[mc].chan[ch].cdev, \ + 0x108 + 4 * (i), ®) +#define SKX_GET_RIRILV(d, mc, ch, idx, i, reg) \ + pci_read_config_dword((d)->imc[mc].chan[ch].cdev, \ + 0x120 + 16 * idx + 4 * (i), ®) + +#define SKX_RIR_VALID(b) GET_BITFIELD((b), 31, 31) +#define SKX_RIR_LIMIT(b) (((u64)GET_BITFIELD((b), 1, 11) << 29) | MASK29) +#define SKX_RIR_WAYS(b) (1 << GET_BITFIELD((b), 28, 29)) +#define SKX_RIR_CHAN_RANK(b) GET_BITFIELD((b), 16, 19) +#define SKX_RIR_OFFSET(b) ((u64)(GET_BITFIELD((b), 2, 15) << 26)) + +static bool skx_rir_decode(struct decoded_addr *res) +{ + int i, idx, chan_rank; + int shift; + u32 rirway, rirlv; + u64 rank_addr, prev_limit = 0, limit; + + if (res->dev->imc[res->imc].chan[res->channel].dimms[0].close_pg) + shift = 6; + else + shift = 13; + + for (i = 0; i < SKX_MAX_RIR; i++) { + SKX_GET_RIRWAYNESS(res->dev, res->imc, res->channel, i, rirway); + limit = SKX_RIR_LIMIT(rirway); + if (SKX_RIR_VALID(rirway)) { + if (prev_limit <= res->chan_addr && + res->chan_addr <= limit) + goto rir_found; + } + prev_limit = limit; + } + edac_dbg(0, "No RIR entry for %llx\n", res->addr); + return false; + +rir_found: + rank_addr = res->chan_addr >> shift; + rank_addr /= SKX_RIR_WAYS(rirway); + rank_addr <<= shift; + rank_addr |= res->chan_addr & GENMASK_ULL(shift - 1, 0); + + res->rank_address = rank_addr; + idx = (res->chan_addr >> shift) % SKX_RIR_WAYS(rirway); + + SKX_GET_RIRILV(res->dev, res->imc, res->channel, idx, i, rirlv); + res->rank_address = rank_addr - SKX_RIR_OFFSET(rirlv); + chan_rank = SKX_RIR_CHAN_RANK(rirlv); + res->channel_rank = chan_rank; + res->dimm = chan_rank / 4; + res->rank = chan_rank % 4; + + edac_dbg(2, "%llx: dimm=%d rank=%d chan_rank=%d rank_addr=%llx\n", + res->addr, res->dimm, res->rank, + res->channel_rank, res->rank_address); + return true; +} + +static u8 skx_close_row[] = { + 15, 16, 17, 18, 20, 21, 22, 28, 10, 11, 12, 13, 29, 30, 31, 32, 33 +}; +static u8 skx_close_column[] = { + 3, 4, 5, 14, 19, 23, 24, 25, 26, 27 +}; +static u8 skx_open_row[] = { + 14, 15, 16, 20, 28, 21, 22, 23, 24, 25, 26, 27, 29, 30, 31, 32, 33 +}; +static u8 skx_open_column[] = { + 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 +}; +static u8 skx_open_fine_column[] = { + 3, 4, 5, 7, 8, 9, 10, 11, 12, 13 +}; + +static int skx_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 skx_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 skx_mad_decode(struct decoded_addr *r) +{ + struct skx_dimm *dimm = &r->dev->imc[r->imc].chan[r->channel].dimms[r->dimm]; + int bg0 = dimm->fine_grain_bank ? 6 : 13; + + if (dimm->close_pg) { + r->row = skx_bits(r->rank_address, dimm->rowbits, skx_close_row); + r->column = skx_bits(r->rank_address, dimm->colbits, skx_close_column); + r->column |= 0x400; /* C10 is autoprecharge, always set */ + r->bank_address = skx_bank_bits(r->rank_address, 8, 9, dimm->bank_xor_enable, 22, 28); + r->bank_group = skx_bank_bits(r->rank_address, 6, 7, dimm->bank_xor_enable, 20, 21); + } else { + r->row = skx_bits(r->rank_address, dimm->rowbits, skx_open_row); + if (dimm->fine_grain_bank) + r->column = skx_bits(r->rank_address, dimm->colbits, skx_open_fine_column); + else + r->column = skx_bits(r->rank_address, dimm->colbits, skx_open_column); + r->bank_address = skx_bank_bits(r->rank_address, 18, 19, dimm->bank_xor_enable, 22, 23); + r->bank_group = skx_bank_bits(r->rank_address, bg0, 17, dimm->bank_xor_enable, 20, 21); + } + r->row &= (1u << dimm->rowbits) - 1; + + edac_dbg(2, "%llx: row=%x col=%x bank_addr=%d bank_group=%d\n", + r->addr, r->row, r->column, r->bank_address, + r->bank_group); + return true; +} + +static bool skx_decode(struct decoded_addr *res) +{ + + return skx_sad_decode(res) && skx_tad_decode(res) && + skx_rir_decode(res) && skx_mad_decode(res); +} + +#ifdef CONFIG_EDAC_DEBUG +/* + * Debug feature. Make /sys/kernel/debug/skx_edac_test/addr. + * Write an address to this file to exercise the address decode + * logic in this driver. + */ +static struct dentry *skx_test; +static u64 skx_fake_addr; + +static int debugfs_u64_set(void *data, u64 val) +{ + struct decoded_addr res; + + res.addr = val; + skx_decode(&res); + + return 0; +} + +DEFINE_SIMPLE_ATTRIBUTE(fops_u64_wo, NULL, debugfs_u64_set, "%llu\n"); + +static struct dentry *mydebugfs_create(const char *name, umode_t mode, + struct dentry *parent, u64 *value) +{ + return debugfs_create_file(name, mode, parent, value, &fops_u64_wo); +} + +static void setup_skx_debug(void) +{ + skx_test = debugfs_create_dir("skx_edac_test", NULL); + mydebugfs_create("addr", S_IWUSR, skx_test, &skx_fake_addr); +} + +static void teardown_skx_debug(void) +{ + debugfs_remove_recursive(skx_test); +} +#else +static void setup_skx_debug(void) +{ +} + +static void teardown_skx_debug(void) +{ +} +#endif /*CONFIG_EDAC_DEBUG*/ + +static void skx_mce_output_error(struct mem_ctl_info *mci, + const struct mce *m, + struct decoded_addr *res) +{ + enum hw_event_mc_err_type tp_event; + char *type, *optype, msg[256]; + 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); + bool recoverable; + u32 core_err_cnt = GET_BITFIELD(m->status, 38, 52); + u32 mscod = GET_BITFIELD(m->status, 16, 31); + u32 errcode = GET_BITFIELD(m->status, 0, 15); + u32 optypenum = GET_BITFIELD(m->status, 4, 6); + + recoverable = GET_BITFIELD(m->status, 56, 56); + + if (uncorrected_error) { + if (ripv) { + type = "FATAL"; + tp_event = HW_EVENT_ERR_FATAL; + } else { + type = "NON_FATAL"; + tp_event = HW_EVENT_ERR_UNCORRECTED; + } + } else { + type = "CORRECTED"; + tp_event = HW_EVENT_ERR_CORRECTED; + } + + /* + * According with Table 15-9 of the Intel Architecture spec vol 3A, + * memory errors should fit in this mask: + * 000f 0000 1mmm cccc (binary) + * where: + * f = Correction Report Filtering Bit. If 1, subsequent errors + * won't be shown + * mmm = error type + * cccc = channel + * If the mask doesn't match, report an error to the parsing logic + */ + if (!((errcode & 0xef80) == 0x80)) { + optype = "Can't parse: it is not a mem"; + } else { + switch (optypenum) { + case 0: + optype = "generic undef request error"; + break; + case 1: + optype = "memory read error"; + break; + case 2: + optype = "memory write error"; + break; + case 3: + optype = "addr/cmd error"; + break; + case 4: + optype = "memory scrubbing error"; + break; + default: + optype = "reserved"; + break; + } + } + + snprintf(msg, sizeof(msg), + "%s%s err_code:%04x:%04x socket:%d imc:%d rank:%d bg:%d ba:%d row:%x col:%x", + overflow ? " OVERFLOW" : "", + (uncorrected_error && recoverable) ? " recoverable" : "", + mscod, errcode, + res->socket, res->imc, res->rank, + res->bank_group, res->bank_address, res->row, res->column); + + edac_dbg(0, "%s\n", msg); + + /* Call the helper to output message */ + edac_mc_handle_error(tp_event, mci, core_err_cnt, + m->addr >> PAGE_SHIFT, m->addr & ~PAGE_MASK, 0, + res->channel, res->dimm, -1, + optype, msg); +} + +static int skx_mce_check_error(struct notifier_block *nb, unsigned long val, + void *data) +{ + struct mce *mce = (struct mce *)data; + struct decoded_addr res; + struct mem_ctl_info *mci; + char *type; + + if (get_edac_report_status() == EDAC_REPORTING_DISABLED) + return NOTIFY_DONE; + + /* ignore unless this is memory related with an address */ + if ((mce->status & 0xefff) >> 7 != 1 || !(mce->status & MCI_STATUS_ADDRV)) + return NOTIFY_DONE; + + res.addr = mce->addr; + if (!skx_decode(&res)) + return NOTIFY_DONE; + mci = res.dev->imc[res.imc].mci; + + if (mce->mcgstatus & MCG_STATUS_MCIP) + type = "Exception"; + else + type = "Event"; + + skx_mc_printk(mci, KERN_DEBUG, "HANDLING MCE MEMORY ERROR\n"); + + skx_mc_printk(mci, KERN_DEBUG, "CPU %d: Machine Check %s: %Lx " + "Bank %d: %016Lx\n", mce->extcpu, type, + mce->mcgstatus, mce->bank, mce->status); + skx_mc_printk(mci, KERN_DEBUG, "TSC %llx ", mce->tsc); + skx_mc_printk(mci, KERN_DEBUG, "ADDR %llx ", mce->addr); + skx_mc_printk(mci, KERN_DEBUG, "MISC %llx ", mce->misc); + + skx_mc_printk(mci, KERN_DEBUG, "PROCESSOR %u:%x TIME %llu SOCKET " + "%u APIC %x\n", mce->cpuvendor, mce->cpuid, + mce->time, mce->socketid, mce->apicid); + + skx_mce_output_error(mci, mce, &res); + + return NOTIFY_DONE; +} + +static struct notifier_block skx_mce_dec = { + .notifier_call = skx_mce_check_error, +}; + +static void skx_remove(void) +{ + int i, j; + struct skx_dev *d, *tmp; + + edac_dbg(0, "\n"); + + list_for_each_entry_safe(d, tmp, &skx_edac_list, list) { + list_del(&d->list); + for (i = 0; i < NUM_IMC; i++) { + skx_unregister_mci(&d->imc[i]); + for (j = 0; j < NUM_CHANNELS; j++) + pci_dev_put(d->imc[i].chan[j].cdev); + } + pci_dev_put(d->util_all); + pci_dev_put(d->sad_all); + + kfree(d); + } +} + +/* + * skx_init: + * make sure we are running on the correct cpu model + * search for all the devices we need + * check which DIMMs are present. + */ +int __init skx_init(void) +{ + const struct x86_cpu_id *id; + const struct munit *m; + int rc = 0, i; + u8 mc = 0, src_id, node_id; + struct skx_dev *d; + + edac_dbg(2, "\n"); + + id = x86_match_cpu(skx_cpuids); + if (!id) + return -ENODEV; + + rc = skx_get_hi_lo(); + if (rc) + return rc; + + rc = get_all_bus_mappings(); + if (rc < 0) + goto fail; + if (rc == 0) { + edac_dbg(2, "No memory controllers found\n"); + return -ENODEV; + } + + for (m = skx_all_munits; m->did; m++) { + rc = get_all_munits(m); + if (rc < 0) + goto fail; + if (rc != m->per_socket * skx_num_sockets) { + edac_dbg(2, "Expected %d, got %d of %x\n", + m->per_socket * skx_num_sockets, rc, m->did); + rc = -ENODEV; + goto fail; + } + } + + list_for_each_entry(d, &skx_edac_list, list) { + src_id = get_src_id(d); + node_id = skx_get_node_id(d); + edac_dbg(2, "src_id=%d node_id=%d\n", src_id, node_id); + for (i = 0; i < NUM_IMC; i++) { + d->imc[i].mc = mc++; + d->imc[i].lmc = i; + d->imc[i].src_id = src_id; + d->imc[i].node_id = node_id; + rc = skx_register_mci(&d->imc[i]); + if (rc < 0) + goto fail; + } + } + + /* Ensure that the OPSTATE is set correctly for POLL or NMI */ + opstate_init(); + + setup_skx_debug(); + + mce_register_decode_chain(&skx_mce_dec); + + return 0; +fail: + skx_remove(); + return rc; +} + +static void __exit skx_exit(void) +{ + edac_dbg(2, "\n"); + mce_unregister_decode_chain(&skx_mce_dec); + skx_remove(); + teardown_skx_debug(); +} + +module_init(skx_init); +module_exit(skx_exit); + +module_param(edac_op_state, int, 0444); +MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI"); + +MODULE_LICENSE("GPL v2"); +MODULE_AUTHOR("Tony Luck"); +MODULE_DESCRIPTION("MC Driver for Intel Skylake server processors"); -- 2.20.1