1 /* SPDX-License-Identifier: GPL-2.0 */ 2 /* 3 * Common codes for both the skx_edac driver and Intel 10nm server EDAC driver. 4 * Originally split out from the skx_edac driver. 5 * 6 * Copyright (c) 2018, Intel Corporation. 7 */ 8 9 #ifndef _SKX_COMM_EDAC_H 10 #define _SKX_COMM_EDAC_H 11 12 #include <linux/bits.h> 13 #include <asm/mce.h> 14 15 #define MSG_SIZE 1024 16 17 /* 18 * Debug macros 19 */ 20 #define skx_printk(level, fmt, arg...) \ 21 edac_printk(level, "skx", fmt, ##arg) 22 23 #define skx_mc_printk(mci, level, fmt, arg...) \ 24 edac_mc_chipset_printk(mci, level, "skx", fmt, ##arg) 25 26 /* 27 * Get a bit field at register value <v>, from bit <lo> to bit <hi> 28 */ 29 #define GET_BITFIELD(v, lo, hi) \ 30 (((v) & GENMASK_ULL((hi), (lo))) >> (lo)) 31 32 #define SKX_NUM_IMC 2 /* Memory controllers per socket */ 33 #define SKX_NUM_CHANNELS 3 /* Channels per memory controller */ 34 #define SKX_NUM_DIMMS 2 /* Max DIMMS per channel */ 35 36 #define I10NM_NUM_DDR_IMC 12 37 #define I10NM_NUM_DDR_CHANNELS 2 38 #define I10NM_NUM_DDR_DIMMS 2 39 40 #define I10NM_NUM_HBM_IMC 16 41 #define I10NM_NUM_HBM_CHANNELS 2 42 #define I10NM_NUM_HBM_DIMMS 1 43 44 #define I10NM_NUM_IMC (I10NM_NUM_DDR_IMC + I10NM_NUM_HBM_IMC) 45 #define I10NM_NUM_CHANNELS MAX(I10NM_NUM_DDR_CHANNELS, I10NM_NUM_HBM_CHANNELS) 46 #define I10NM_NUM_DIMMS MAX(I10NM_NUM_DDR_DIMMS, I10NM_NUM_HBM_DIMMS) 47 48 #define MAX(a, b) ((a) > (b) ? (a) : (b)) 49 #define NUM_IMC MAX(SKX_NUM_IMC, I10NM_NUM_IMC) 50 #define NUM_CHANNELS MAX(SKX_NUM_CHANNELS, I10NM_NUM_CHANNELS) 51 #define NUM_DIMMS MAX(SKX_NUM_DIMMS, I10NM_NUM_DIMMS) 52 53 #define IS_DIMM_PRESENT(r) GET_BITFIELD(r, 15, 15) 54 #define IS_NVDIMM_PRESENT(r, i) GET_BITFIELD(r, i, i) 55 56 #define MCI_MISC_ECC_MODE(m) (((m) >> 59) & 15) 57 #define MCI_MISC_ECC_DDRT 8 /* read from DDRT */ 58 59 /* 60 * According to Intel Architecture spec vol 3B, 61 * Table 15-10 "IA32_MCi_Status [15:0] Compound Error Code Encoding" 62 * memory errors should fit one of these masks: 63 * 000f 0000 1mmm cccc (binary) 64 * 000f 0010 1mmm cccc (binary) [RAM used as cache] 65 * where: 66 * f = Correction Report Filtering Bit. If 1, subsequent errors 67 * won't be shown 68 * mmm = error type 69 * cccc = channel 70 */ 71 #define MCACOD_MEM_ERR_MASK 0xef80 72 /* 73 * Errors from either the memory of the 1-level memory system or the 74 * 2nd level memory (the slow "far" memory) of the 2-level memory system. 75 */ 76 #define MCACOD_MEM_CTL_ERR 0x80 77 /* 78 * Errors from the 1st level memory (the fast "near" memory as cache) 79 * of the 2-level memory system. 80 */ 81 #define MCACOD_EXT_MEM_ERR 0x280 82 83 /* 84 * Each cpu socket contains some pci devices that provide global 85 * information, and also some that are local to each of the two 86 * memory controllers on the die. 87 */ 88 struct skx_dev { 89 struct list_head list; 90 u8 bus[4]; 91 int seg; 92 struct pci_dev *sad_all; 93 struct pci_dev *util_all; 94 struct pci_dev *uracu; /* for i10nm CPU */ 95 struct pci_dev *pcu_cr3; /* for HBM memory detection */ 96 u32 mcroute; 97 struct skx_imc { 98 struct mem_ctl_info *mci; 99 struct pci_dev *mdev; /* for i10nm CPU */ 100 void __iomem *mbase; /* for i10nm CPU */ 101 int chan_mmio_sz; /* for i10nm CPU */ 102 int num_channels; /* channels per memory controller */ 103 int num_dimms; /* dimms per channel */ 104 bool hbm_mc; 105 u8 mc; /* system wide mc# */ 106 u8 lmc; /* socket relative mc# */ 107 u8 src_id, node_id; 108 struct skx_channel { 109 struct pci_dev *cdev; 110 struct pci_dev *edev; 111 u32 retry_rd_err_log_s; 112 u32 retry_rd_err_log_d; 113 u32 retry_rd_err_log_d2; 114 struct skx_dimm { 115 u8 close_pg; 116 u8 bank_xor_enable; 117 u8 fine_grain_bank; 118 u8 rowbits; 119 u8 colbits; 120 } dimms[NUM_DIMMS]; 121 } chan[NUM_CHANNELS]; 122 } imc[NUM_IMC]; 123 }; 124 125 struct skx_pvt { 126 struct skx_imc *imc; 127 }; 128 129 enum type { 130 SKX, 131 I10NM, 132 SPR, 133 GNR 134 }; 135 136 enum { 137 INDEX_SOCKET, 138 INDEX_MEMCTRL, 139 INDEX_CHANNEL, 140 INDEX_DIMM, 141 INDEX_CS, 142 INDEX_NM_FIRST, 143 INDEX_NM_MEMCTRL = INDEX_NM_FIRST, 144 INDEX_NM_CHANNEL, 145 INDEX_NM_DIMM, 146 INDEX_NM_CS, 147 INDEX_MAX 148 }; 149 150 enum error_source { 151 ERR_SRC_1LM, 152 ERR_SRC_2LM_NM, 153 ERR_SRC_2LM_FM, 154 ERR_SRC_NOT_MEMORY, 155 }; 156 157 #define BIT_NM_MEMCTRL BIT_ULL(INDEX_NM_MEMCTRL) 158 #define BIT_NM_CHANNEL BIT_ULL(INDEX_NM_CHANNEL) 159 #define BIT_NM_DIMM BIT_ULL(INDEX_NM_DIMM) 160 #define BIT_NM_CS BIT_ULL(INDEX_NM_CS) 161 162 struct decoded_addr { 163 struct mce *mce; 164 struct skx_dev *dev; 165 u64 addr; 166 int socket; 167 int imc; 168 int channel; 169 u64 chan_addr; 170 int sktways; 171 int chanways; 172 int dimm; 173 int cs; 174 int rank; 175 int channel_rank; 176 u64 rank_address; 177 int row; 178 int column; 179 int bank_address; 180 int bank_group; 181 bool decoded_by_adxl; 182 }; 183 184 struct pci_bdf { 185 u32 bus : 8; 186 u32 dev : 5; 187 u32 fun : 3; 188 }; 189 190 struct res_config { 191 enum type type; 192 /* Configuration agent device ID */ 193 unsigned int decs_did; 194 /* Default bus number configuration register offset */ 195 int busno_cfg_offset; 196 /* DDR memory controllers per socket */ 197 int ddr_imc_num; 198 /* DDR channels per DDR memory controller */ 199 int ddr_chan_num; 200 /* DDR DIMMs per DDR memory channel */ 201 int ddr_dimm_num; 202 /* Per DDR channel memory-mapped I/O size */ 203 int ddr_chan_mmio_sz; 204 /* HBM memory controllers per socket */ 205 int hbm_imc_num; 206 /* HBM channels per HBM memory controller */ 207 int hbm_chan_num; 208 /* HBM DIMMs per HBM memory channel */ 209 int hbm_dimm_num; 210 /* Per HBM channel memory-mapped I/O size */ 211 int hbm_chan_mmio_sz; 212 bool support_ddr5; 213 /* SAD device BDF */ 214 struct pci_bdf sad_all_bdf; 215 /* PCU device BDF */ 216 struct pci_bdf pcu_cr3_bdf; 217 /* UTIL device BDF */ 218 struct pci_bdf util_all_bdf; 219 /* URACU device BDF */ 220 struct pci_bdf uracu_bdf; 221 /* DDR mdev device BDF */ 222 struct pci_bdf ddr_mdev_bdf; 223 /* HBM mdev device BDF */ 224 struct pci_bdf hbm_mdev_bdf; 225 int sad_all_offset; 226 /* Offsets of retry_rd_err_log registers */ 227 u32 *offsets_scrub; 228 u32 *offsets_scrub_hbm0; 229 u32 *offsets_scrub_hbm1; 230 u32 *offsets_demand; 231 u32 *offsets_demand2; 232 u32 *offsets_demand_hbm0; 233 u32 *offsets_demand_hbm1; 234 }; 235 236 typedef int (*get_dimm_config_f)(struct mem_ctl_info *mci, 237 struct res_config *cfg); 238 typedef bool (*skx_decode_f)(struct decoded_addr *res); 239 typedef void (*skx_show_retry_log_f)(struct decoded_addr *res, char *msg, int len, bool scrub_err); 240 241 int skx_adxl_get(void); 242 void skx_adxl_put(void); 243 void skx_set_decode(skx_decode_f decode, skx_show_retry_log_f show_retry_log); 244 void skx_set_mem_cfg(bool mem_cfg_2lm); 245 void skx_set_res_cfg(struct res_config *cfg); 246 247 int skx_get_src_id(struct skx_dev *d, int off, u8 *id); 248 int skx_get_node_id(struct skx_dev *d, u8 *id); 249 250 int skx_get_all_bus_mappings(struct res_config *cfg, struct list_head **list); 251 252 int skx_get_hi_lo(unsigned int did, int off[], u64 *tolm, u64 *tohm); 253 254 int skx_get_dimm_info(u32 mtr, u32 mcmtr, u32 amap, struct dimm_info *dimm, 255 struct skx_imc *imc, int chan, int dimmno, 256 struct res_config *cfg); 257 258 int skx_get_nvdimm_info(struct dimm_info *dimm, struct skx_imc *imc, 259 int chan, int dimmno, const char *mod_str); 260 261 int skx_register_mci(struct skx_imc *imc, struct pci_dev *pdev, 262 const char *ctl_name, const char *mod_str, 263 get_dimm_config_f get_dimm_config, 264 struct res_config *cfg); 265 266 int skx_mce_check_error(struct notifier_block *nb, unsigned long val, 267 void *data); 268 269 void skx_remove(void); 270 271 #endif /* _SKX_COMM_EDAC_H */ 272