Loc Ho | 0d44293 | 2015-05-22 17:32:59 -0600 | [diff] [blame] | 1 | /* |
| 2 | * APM X-Gene SoC EDAC (error detection and correction) |
| 3 | * |
| 4 | * Copyright (c) 2015, Applied Micro Circuits Corporation |
| 5 | * Author: Feng Kan <fkan@apm.com> |
| 6 | * Loc Ho <lho@apm.com> |
| 7 | * |
| 8 | * This program is free software; you can redistribute it and/or modify it |
| 9 | * under the terms of the GNU General Public License as published by the |
| 10 | * Free Software Foundation; either version 2 of the License, or (at your |
| 11 | * option) any later version. |
| 12 | * |
| 13 | * This program is distributed in the hope that it will be useful, |
| 14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 16 | * GNU General Public License for more details. |
| 17 | * |
| 18 | * You should have received a copy of the GNU General Public License |
| 19 | * along with this program. If not, see <http://www.gnu.org/licenses/>. |
| 20 | */ |
| 21 | |
| 22 | #include <linux/ctype.h> |
| 23 | #include <linux/edac.h> |
| 24 | #include <linux/interrupt.h> |
| 25 | #include <linux/mfd/syscon.h> |
| 26 | #include <linux/module.h> |
| 27 | #include <linux/of.h> |
| 28 | #include <linux/of_address.h> |
| 29 | #include <linux/regmap.h> |
| 30 | |
| 31 | #include "edac_core.h" |
| 32 | |
| 33 | #define EDAC_MOD_STR "xgene_edac" |
| 34 | |
| 35 | /* Global error configuration status registers (CSR) */ |
| 36 | #define PCPHPERRINTSTS 0x0000 |
| 37 | #define PCPHPERRINTMSK 0x0004 |
| 38 | #define MCU_CTL_ERR_MASK BIT(12) |
| 39 | #define IOB_PA_ERR_MASK BIT(11) |
| 40 | #define IOB_BA_ERR_MASK BIT(10) |
| 41 | #define IOB_XGIC_ERR_MASK BIT(9) |
| 42 | #define IOB_RB_ERR_MASK BIT(8) |
| 43 | #define L3C_UNCORR_ERR_MASK BIT(5) |
| 44 | #define MCU_UNCORR_ERR_MASK BIT(4) |
| 45 | #define PMD3_MERR_MASK BIT(3) |
| 46 | #define PMD2_MERR_MASK BIT(2) |
| 47 | #define PMD1_MERR_MASK BIT(1) |
| 48 | #define PMD0_MERR_MASK BIT(0) |
| 49 | #define PCPLPERRINTSTS 0x0008 |
| 50 | #define PCPLPERRINTMSK 0x000C |
| 51 | #define CSW_SWITCH_TRACE_ERR_MASK BIT(2) |
| 52 | #define L3C_CORR_ERR_MASK BIT(1) |
| 53 | #define MCU_CORR_ERR_MASK BIT(0) |
| 54 | #define MEMERRINTSTS 0x0010 |
| 55 | #define MEMERRINTMSK 0x0014 |
| 56 | |
| 57 | struct xgene_edac { |
| 58 | struct device *dev; |
| 59 | struct regmap *csw_map; |
| 60 | struct regmap *mcba_map; |
| 61 | struct regmap *mcbb_map; |
| 62 | struct regmap *efuse_map; |
| 63 | void __iomem *pcp_csr; |
| 64 | spinlock_t lock; |
| 65 | struct dentry *dfs; |
| 66 | |
| 67 | struct list_head mcus; |
| 68 | struct list_head pmds; |
| 69 | |
| 70 | struct mutex mc_lock; |
| 71 | int mc_active_mask; |
| 72 | int mc_registered_mask; |
| 73 | }; |
| 74 | |
| 75 | static void xgene_edac_pcp_rd(struct xgene_edac *edac, u32 reg, u32 *val) |
| 76 | { |
| 77 | *val = readl(edac->pcp_csr + reg); |
| 78 | } |
| 79 | |
| 80 | static void xgene_edac_pcp_clrbits(struct xgene_edac *edac, u32 reg, |
| 81 | u32 bits_mask) |
| 82 | { |
| 83 | u32 val; |
| 84 | |
| 85 | spin_lock(&edac->lock); |
| 86 | val = readl(edac->pcp_csr + reg); |
| 87 | val &= ~bits_mask; |
| 88 | writel(val, edac->pcp_csr + reg); |
| 89 | spin_unlock(&edac->lock); |
| 90 | } |
| 91 | |
| 92 | static void xgene_edac_pcp_setbits(struct xgene_edac *edac, u32 reg, |
| 93 | u32 bits_mask) |
| 94 | { |
| 95 | u32 val; |
| 96 | |
| 97 | spin_lock(&edac->lock); |
| 98 | val = readl(edac->pcp_csr + reg); |
| 99 | val |= bits_mask; |
| 100 | writel(val, edac->pcp_csr + reg); |
| 101 | spin_unlock(&edac->lock); |
| 102 | } |
| 103 | |
| 104 | /* Memory controller error CSR */ |
| 105 | #define MCU_MAX_RANK 8 |
| 106 | #define MCU_RANK_STRIDE 0x40 |
| 107 | |
| 108 | #define MCUGECR 0x0110 |
| 109 | #define MCU_GECR_DEMANDUCINTREN_MASK BIT(0) |
| 110 | #define MCU_GECR_BACKUCINTREN_MASK BIT(1) |
| 111 | #define MCU_GECR_CINTREN_MASK BIT(2) |
| 112 | #define MUC_GECR_MCUADDRERREN_MASK BIT(9) |
| 113 | #define MCUGESR 0x0114 |
| 114 | #define MCU_GESR_ADDRNOMATCH_ERR_MASK BIT(7) |
| 115 | #define MCU_GESR_ADDRMULTIMATCH_ERR_MASK BIT(6) |
| 116 | #define MCU_GESR_PHYP_ERR_MASK BIT(3) |
| 117 | #define MCUESRR0 0x0314 |
| 118 | #define MCU_ESRR_MULTUCERR_MASK BIT(3) |
| 119 | #define MCU_ESRR_BACKUCERR_MASK BIT(2) |
| 120 | #define MCU_ESRR_DEMANDUCERR_MASK BIT(1) |
| 121 | #define MCU_ESRR_CERR_MASK BIT(0) |
| 122 | #define MCUESRRA0 0x0318 |
| 123 | #define MCUEBLRR0 0x031c |
| 124 | #define MCU_EBLRR_ERRBANK_RD(src) (((src) & 0x00000007) >> 0) |
| 125 | #define MCUERCRR0 0x0320 |
| 126 | #define MCU_ERCRR_ERRROW_RD(src) (((src) & 0xFFFF0000) >> 16) |
| 127 | #define MCU_ERCRR_ERRCOL_RD(src) ((src) & 0x00000FFF) |
| 128 | #define MCUSBECNT0 0x0324 |
| 129 | #define MCU_SBECNT_COUNT(src) ((src) & 0xFFFF) |
| 130 | |
| 131 | #define CSW_CSWCR 0x0000 |
| 132 | #define CSW_CSWCR_DUALMCB_MASK BIT(0) |
| 133 | |
| 134 | #define MCBADDRMR 0x0000 |
| 135 | #define MCBADDRMR_MCU_INTLV_MODE_MASK BIT(3) |
| 136 | #define MCBADDRMR_DUALMCU_MODE_MASK BIT(2) |
| 137 | #define MCBADDRMR_MCB_INTLV_MODE_MASK BIT(1) |
| 138 | #define MCBADDRMR_ADDRESS_MODE_MASK BIT(0) |
| 139 | |
| 140 | struct xgene_edac_mc_ctx { |
| 141 | struct list_head next; |
| 142 | char *name; |
| 143 | struct mem_ctl_info *mci; |
| 144 | struct xgene_edac *edac; |
| 145 | void __iomem *mcu_csr; |
| 146 | u32 mcu_id; |
| 147 | }; |
| 148 | |
| 149 | static ssize_t xgene_edac_mc_err_inject_write(struct file *file, |
| 150 | const char __user *data, |
| 151 | size_t count, loff_t *ppos) |
| 152 | { |
| 153 | struct mem_ctl_info *mci = file->private_data; |
| 154 | struct xgene_edac_mc_ctx *ctx = mci->pvt_info; |
| 155 | int i; |
| 156 | |
| 157 | for (i = 0; i < MCU_MAX_RANK; i++) { |
| 158 | writel(MCU_ESRR_MULTUCERR_MASK | MCU_ESRR_BACKUCERR_MASK | |
| 159 | MCU_ESRR_DEMANDUCERR_MASK | MCU_ESRR_CERR_MASK, |
| 160 | ctx->mcu_csr + MCUESRRA0 + i * MCU_RANK_STRIDE); |
| 161 | } |
| 162 | return count; |
| 163 | } |
| 164 | |
| 165 | static const struct file_operations xgene_edac_mc_debug_inject_fops = { |
| 166 | .open = simple_open, |
| 167 | .write = xgene_edac_mc_err_inject_write, |
| 168 | .llseek = generic_file_llseek, |
| 169 | }; |
| 170 | |
| 171 | static void xgene_edac_mc_create_debugfs_node(struct mem_ctl_info *mci) |
| 172 | { |
| 173 | if (!IS_ENABLED(CONFIG_EDAC_DEBUG)) |
| 174 | return; |
| 175 | #ifdef CONFIG_EDAC_DEBUG |
| 176 | if (!mci->debugfs) |
| 177 | return; |
| 178 | debugfs_create_file("inject_ctrl", S_IWUSR, mci->debugfs, mci, |
| 179 | &xgene_edac_mc_debug_inject_fops); |
| 180 | #endif |
| 181 | } |
| 182 | |
| 183 | static void xgene_edac_mc_check(struct mem_ctl_info *mci) |
| 184 | { |
| 185 | struct xgene_edac_mc_ctx *ctx = mci->pvt_info; |
| 186 | unsigned int pcp_hp_stat; |
| 187 | unsigned int pcp_lp_stat; |
| 188 | u32 reg; |
| 189 | u32 rank; |
| 190 | u32 bank; |
| 191 | u32 count; |
| 192 | u32 col_row; |
| 193 | |
| 194 | xgene_edac_pcp_rd(ctx->edac, PCPHPERRINTSTS, &pcp_hp_stat); |
| 195 | xgene_edac_pcp_rd(ctx->edac, PCPLPERRINTSTS, &pcp_lp_stat); |
| 196 | if (!((MCU_UNCORR_ERR_MASK & pcp_hp_stat) || |
| 197 | (MCU_CTL_ERR_MASK & pcp_hp_stat) || |
| 198 | (MCU_CORR_ERR_MASK & pcp_lp_stat))) |
| 199 | return; |
| 200 | |
| 201 | for (rank = 0; rank < MCU_MAX_RANK; rank++) { |
| 202 | reg = readl(ctx->mcu_csr + MCUESRR0 + rank * MCU_RANK_STRIDE); |
| 203 | |
| 204 | /* Detect uncorrectable memory error */ |
| 205 | if (reg & (MCU_ESRR_DEMANDUCERR_MASK | |
| 206 | MCU_ESRR_BACKUCERR_MASK)) { |
| 207 | /* Detected uncorrectable memory error */ |
| 208 | edac_mc_chipset_printk(mci, KERN_ERR, "X-Gene", |
| 209 | "MCU uncorrectable error at rank %d\n", rank); |
| 210 | |
| 211 | edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, |
| 212 | 1, 0, 0, 0, 0, 0, -1, mci->ctl_name, ""); |
| 213 | } |
| 214 | |
| 215 | /* Detect correctable memory error */ |
| 216 | if (reg & MCU_ESRR_CERR_MASK) { |
| 217 | bank = readl(ctx->mcu_csr + MCUEBLRR0 + |
| 218 | rank * MCU_RANK_STRIDE); |
| 219 | col_row = readl(ctx->mcu_csr + MCUERCRR0 + |
| 220 | rank * MCU_RANK_STRIDE); |
| 221 | count = readl(ctx->mcu_csr + MCUSBECNT0 + |
| 222 | rank * MCU_RANK_STRIDE); |
| 223 | edac_mc_chipset_printk(mci, KERN_WARNING, "X-Gene", |
| 224 | "MCU correctable error at rank %d bank %d column %d row %d count %d\n", |
| 225 | rank, MCU_EBLRR_ERRBANK_RD(bank), |
| 226 | MCU_ERCRR_ERRCOL_RD(col_row), |
| 227 | MCU_ERCRR_ERRROW_RD(col_row), |
| 228 | MCU_SBECNT_COUNT(count)); |
| 229 | |
| 230 | edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, |
| 231 | 1, 0, 0, 0, 0, 0, -1, mci->ctl_name, ""); |
| 232 | } |
| 233 | |
| 234 | /* Clear all error registers */ |
| 235 | writel(0x0, ctx->mcu_csr + MCUEBLRR0 + rank * MCU_RANK_STRIDE); |
| 236 | writel(0x0, ctx->mcu_csr + MCUERCRR0 + rank * MCU_RANK_STRIDE); |
| 237 | writel(0x0, ctx->mcu_csr + MCUSBECNT0 + |
| 238 | rank * MCU_RANK_STRIDE); |
| 239 | writel(reg, ctx->mcu_csr + MCUESRR0 + rank * MCU_RANK_STRIDE); |
| 240 | } |
| 241 | |
| 242 | /* Detect memory controller error */ |
| 243 | reg = readl(ctx->mcu_csr + MCUGESR); |
| 244 | if (reg) { |
| 245 | if (reg & MCU_GESR_ADDRNOMATCH_ERR_MASK) |
| 246 | edac_mc_chipset_printk(mci, KERN_WARNING, "X-Gene", |
| 247 | "MCU address miss-match error\n"); |
| 248 | if (reg & MCU_GESR_ADDRMULTIMATCH_ERR_MASK) |
| 249 | edac_mc_chipset_printk(mci, KERN_WARNING, "X-Gene", |
| 250 | "MCU address multi-match error\n"); |
| 251 | |
| 252 | writel(reg, ctx->mcu_csr + MCUGESR); |
| 253 | } |
| 254 | } |
| 255 | |
| 256 | static void xgene_edac_mc_irq_ctl(struct mem_ctl_info *mci, bool enable) |
| 257 | { |
| 258 | struct xgene_edac_mc_ctx *ctx = mci->pvt_info; |
| 259 | unsigned int val; |
| 260 | |
| 261 | if (edac_op_state != EDAC_OPSTATE_INT) |
| 262 | return; |
| 263 | |
| 264 | mutex_lock(&ctx->edac->mc_lock); |
| 265 | |
| 266 | /* |
| 267 | * As there is only single bit for enable error and interrupt mask, |
| 268 | * we must only enable top level interrupt after all MCUs are |
| 269 | * registered. Otherwise, if there is an error and the corresponding |
| 270 | * MCU has not registered, the interrupt will never get cleared. To |
| 271 | * determine all MCU have registered, we will keep track of active |
| 272 | * MCUs and registered MCUs. |
| 273 | */ |
| 274 | if (enable) { |
| 275 | /* Set registered MCU bit */ |
| 276 | ctx->edac->mc_registered_mask |= 1 << ctx->mcu_id; |
| 277 | |
| 278 | /* Enable interrupt after all active MCU registered */ |
| 279 | if (ctx->edac->mc_registered_mask == |
| 280 | ctx->edac->mc_active_mask) { |
| 281 | /* Enable memory controller top level interrupt */ |
| 282 | xgene_edac_pcp_clrbits(ctx->edac, PCPHPERRINTMSK, |
| 283 | MCU_UNCORR_ERR_MASK | |
| 284 | MCU_CTL_ERR_MASK); |
| 285 | xgene_edac_pcp_clrbits(ctx->edac, PCPLPERRINTMSK, |
| 286 | MCU_CORR_ERR_MASK); |
| 287 | } |
| 288 | |
| 289 | /* Enable MCU interrupt and error reporting */ |
| 290 | val = readl(ctx->mcu_csr + MCUGECR); |
| 291 | val |= MCU_GECR_DEMANDUCINTREN_MASK | |
| 292 | MCU_GECR_BACKUCINTREN_MASK | |
| 293 | MCU_GECR_CINTREN_MASK | |
| 294 | MUC_GECR_MCUADDRERREN_MASK; |
| 295 | writel(val, ctx->mcu_csr + MCUGECR); |
| 296 | } else { |
| 297 | /* Disable MCU interrupt */ |
| 298 | val = readl(ctx->mcu_csr + MCUGECR); |
| 299 | val &= ~(MCU_GECR_DEMANDUCINTREN_MASK | |
| 300 | MCU_GECR_BACKUCINTREN_MASK | |
| 301 | MCU_GECR_CINTREN_MASK | |
| 302 | MUC_GECR_MCUADDRERREN_MASK); |
| 303 | writel(val, ctx->mcu_csr + MCUGECR); |
| 304 | |
| 305 | /* Disable memory controller top level interrupt */ |
| 306 | xgene_edac_pcp_setbits(ctx->edac, PCPHPERRINTMSK, |
| 307 | MCU_UNCORR_ERR_MASK | MCU_CTL_ERR_MASK); |
| 308 | xgene_edac_pcp_setbits(ctx->edac, PCPLPERRINTMSK, |
| 309 | MCU_CORR_ERR_MASK); |
| 310 | |
| 311 | /* Clear registered MCU bit */ |
| 312 | ctx->edac->mc_registered_mask &= ~(1 << ctx->mcu_id); |
| 313 | } |
| 314 | |
| 315 | mutex_unlock(&ctx->edac->mc_lock); |
| 316 | } |
| 317 | |
| 318 | static int xgene_edac_mc_is_active(struct xgene_edac_mc_ctx *ctx, int mc_idx) |
| 319 | { |
| 320 | unsigned int reg; |
| 321 | u32 mcu_mask; |
| 322 | |
| 323 | if (regmap_read(ctx->edac->csw_map, CSW_CSWCR, ®)) |
| 324 | return 0; |
| 325 | |
| 326 | if (reg & CSW_CSWCR_DUALMCB_MASK) { |
| 327 | /* |
| 328 | * Dual MCB active - Determine if all 4 active or just MCU0 |
| 329 | * and MCU2 active |
| 330 | */ |
| 331 | if (regmap_read(ctx->edac->mcbb_map, MCBADDRMR, ®)) |
| 332 | return 0; |
| 333 | mcu_mask = (reg & MCBADDRMR_DUALMCU_MODE_MASK) ? 0xF : 0x5; |
| 334 | } else { |
| 335 | /* |
| 336 | * Single MCB active - Determine if MCU0/MCU1 or just MCU0 |
| 337 | * active |
| 338 | */ |
| 339 | if (regmap_read(ctx->edac->mcba_map, MCBADDRMR, ®)) |
| 340 | return 0; |
| 341 | mcu_mask = (reg & MCBADDRMR_DUALMCU_MODE_MASK) ? 0x3 : 0x1; |
| 342 | } |
| 343 | |
| 344 | /* Save active MC mask if hasn't set already */ |
| 345 | if (!ctx->edac->mc_active_mask) |
| 346 | ctx->edac->mc_active_mask = mcu_mask; |
| 347 | |
| 348 | return (mcu_mask & (1 << mc_idx)) ? 1 : 0; |
| 349 | } |
| 350 | |
| 351 | static int xgene_edac_mc_add(struct xgene_edac *edac, struct device_node *np) |
| 352 | { |
| 353 | struct mem_ctl_info *mci; |
| 354 | struct edac_mc_layer layers[2]; |
| 355 | struct xgene_edac_mc_ctx tmp_ctx; |
| 356 | struct xgene_edac_mc_ctx *ctx; |
| 357 | struct resource res; |
| 358 | int rc; |
| 359 | |
| 360 | memset(&tmp_ctx, 0, sizeof(tmp_ctx)); |
| 361 | tmp_ctx.edac = edac; |
| 362 | |
| 363 | if (!devres_open_group(edac->dev, xgene_edac_mc_add, GFP_KERNEL)) |
| 364 | return -ENOMEM; |
| 365 | |
| 366 | rc = of_address_to_resource(np, 0, &res); |
| 367 | if (rc < 0) { |
| 368 | dev_err(edac->dev, "no MCU resource address\n"); |
| 369 | goto err_group; |
| 370 | } |
| 371 | tmp_ctx.mcu_csr = devm_ioremap_resource(edac->dev, &res); |
| 372 | if (IS_ERR(tmp_ctx.mcu_csr)) { |
| 373 | dev_err(edac->dev, "unable to map MCU resource\n"); |
| 374 | rc = PTR_ERR(tmp_ctx.mcu_csr); |
| 375 | goto err_group; |
| 376 | } |
| 377 | |
| 378 | /* Ignore non-active MCU */ |
| 379 | if (of_property_read_u32(np, "memory-controller", &tmp_ctx.mcu_id)) { |
| 380 | dev_err(edac->dev, "no memory-controller property\n"); |
| 381 | rc = -ENODEV; |
| 382 | goto err_group; |
| 383 | } |
| 384 | if (!xgene_edac_mc_is_active(&tmp_ctx, tmp_ctx.mcu_id)) { |
| 385 | rc = -ENODEV; |
| 386 | goto err_group; |
| 387 | } |
| 388 | |
| 389 | layers[0].type = EDAC_MC_LAYER_CHIP_SELECT; |
| 390 | layers[0].size = 4; |
| 391 | layers[0].is_virt_csrow = true; |
| 392 | layers[1].type = EDAC_MC_LAYER_CHANNEL; |
| 393 | layers[1].size = 2; |
| 394 | layers[1].is_virt_csrow = false; |
| 395 | mci = edac_mc_alloc(tmp_ctx.mcu_id, ARRAY_SIZE(layers), layers, |
| 396 | sizeof(*ctx)); |
| 397 | if (!mci) { |
| 398 | rc = -ENOMEM; |
| 399 | goto err_group; |
| 400 | } |
| 401 | |
| 402 | ctx = mci->pvt_info; |
| 403 | *ctx = tmp_ctx; /* Copy over resource value */ |
| 404 | ctx->name = "xgene_edac_mc_err"; |
| 405 | ctx->mci = mci; |
| 406 | mci->pdev = &mci->dev; |
| 407 | mci->ctl_name = ctx->name; |
| 408 | mci->dev_name = ctx->name; |
| 409 | |
| 410 | mci->mtype_cap = MEM_FLAG_RDDR | MEM_FLAG_RDDR2 | MEM_FLAG_RDDR3 | |
| 411 | MEM_FLAG_DDR | MEM_FLAG_DDR2 | MEM_FLAG_DDR3; |
| 412 | mci->edac_ctl_cap = EDAC_FLAG_SECDED; |
| 413 | mci->edac_cap = EDAC_FLAG_SECDED; |
| 414 | mci->mod_name = EDAC_MOD_STR; |
| 415 | mci->mod_ver = "0.1"; |
| 416 | mci->ctl_page_to_phys = NULL; |
| 417 | mci->scrub_cap = SCRUB_FLAG_HW_SRC; |
| 418 | mci->scrub_mode = SCRUB_HW_SRC; |
| 419 | |
| 420 | if (edac_op_state == EDAC_OPSTATE_POLL) |
| 421 | mci->edac_check = xgene_edac_mc_check; |
| 422 | |
| 423 | if (edac_mc_add_mc(mci)) { |
| 424 | dev_err(edac->dev, "edac_mc_add_mc failed\n"); |
| 425 | rc = -EINVAL; |
| 426 | goto err_free; |
| 427 | } |
| 428 | |
| 429 | xgene_edac_mc_create_debugfs_node(mci); |
| 430 | |
| 431 | list_add(&ctx->next, &edac->mcus); |
| 432 | |
| 433 | xgene_edac_mc_irq_ctl(mci, true); |
| 434 | |
| 435 | devres_remove_group(edac->dev, xgene_edac_mc_add); |
| 436 | |
| 437 | dev_info(edac->dev, "X-Gene EDAC MC registered\n"); |
| 438 | return 0; |
| 439 | |
| 440 | err_free: |
| 441 | edac_mc_free(mci); |
| 442 | err_group: |
| 443 | devres_release_group(edac->dev, xgene_edac_mc_add); |
| 444 | return rc; |
| 445 | } |
| 446 | |
| 447 | static int xgene_edac_mc_remove(struct xgene_edac_mc_ctx *mcu) |
| 448 | { |
| 449 | xgene_edac_mc_irq_ctl(mcu->mci, false); |
| 450 | edac_mc_del_mc(&mcu->mci->dev); |
| 451 | edac_mc_free(mcu->mci); |
| 452 | return 0; |
| 453 | } |
| 454 | |
| 455 | /* CPU L1/L2 error CSR */ |
| 456 | #define MAX_CPU_PER_PMD 2 |
| 457 | #define CPU_CSR_STRIDE 0x00100000 |
| 458 | #define CPU_L2C_PAGE 0x000D0000 |
| 459 | #define CPU_MEMERR_L2C_PAGE 0x000E0000 |
| 460 | #define CPU_MEMERR_CPU_PAGE 0x000F0000 |
| 461 | |
| 462 | #define MEMERR_CPU_ICFECR_PAGE_OFFSET 0x0000 |
| 463 | #define MEMERR_CPU_ICFESR_PAGE_OFFSET 0x0004 |
| 464 | #define MEMERR_CPU_ICFESR_ERRWAY_RD(src) (((src) & 0xFF000000) >> 24) |
| 465 | #define MEMERR_CPU_ICFESR_ERRINDEX_RD(src) (((src) & 0x003F0000) >> 16) |
| 466 | #define MEMERR_CPU_ICFESR_ERRINFO_RD(src) (((src) & 0x0000FF00) >> 8) |
| 467 | #define MEMERR_CPU_ICFESR_ERRTYPE_RD(src) (((src) & 0x00000070) >> 4) |
| 468 | #define MEMERR_CPU_ICFESR_MULTCERR_MASK BIT(2) |
| 469 | #define MEMERR_CPU_ICFESR_CERR_MASK BIT(0) |
| 470 | #define MEMERR_CPU_LSUESR_PAGE_OFFSET 0x000c |
| 471 | #define MEMERR_CPU_LSUESR_ERRWAY_RD(src) (((src) & 0xFF000000) >> 24) |
| 472 | #define MEMERR_CPU_LSUESR_ERRINDEX_RD(src) (((src) & 0x003F0000) >> 16) |
| 473 | #define MEMERR_CPU_LSUESR_ERRINFO_RD(src) (((src) & 0x0000FF00) >> 8) |
| 474 | #define MEMERR_CPU_LSUESR_ERRTYPE_RD(src) (((src) & 0x00000070) >> 4) |
| 475 | #define MEMERR_CPU_LSUESR_MULTCERR_MASK BIT(2) |
| 476 | #define MEMERR_CPU_LSUESR_CERR_MASK BIT(0) |
| 477 | #define MEMERR_CPU_LSUECR_PAGE_OFFSET 0x0008 |
| 478 | #define MEMERR_CPU_MMUECR_PAGE_OFFSET 0x0010 |
| 479 | #define MEMERR_CPU_MMUESR_PAGE_OFFSET 0x0014 |
| 480 | #define MEMERR_CPU_MMUESR_ERRWAY_RD(src) (((src) & 0xFF000000) >> 24) |
| 481 | #define MEMERR_CPU_MMUESR_ERRINDEX_RD(src) (((src) & 0x007F0000) >> 16) |
| 482 | #define MEMERR_CPU_MMUESR_ERRINFO_RD(src) (((src) & 0x0000FF00) >> 8) |
| 483 | #define MEMERR_CPU_MMUESR_ERRREQSTR_LSU_MASK BIT(7) |
| 484 | #define MEMERR_CPU_MMUESR_ERRTYPE_RD(src) (((src) & 0x00000070) >> 4) |
| 485 | #define MEMERR_CPU_MMUESR_MULTCERR_MASK BIT(2) |
| 486 | #define MEMERR_CPU_MMUESR_CERR_MASK BIT(0) |
| 487 | #define MEMERR_CPU_ICFESRA_PAGE_OFFSET 0x0804 |
| 488 | #define MEMERR_CPU_LSUESRA_PAGE_OFFSET 0x080c |
| 489 | #define MEMERR_CPU_MMUESRA_PAGE_OFFSET 0x0814 |
| 490 | |
| 491 | #define MEMERR_L2C_L2ECR_PAGE_OFFSET 0x0000 |
| 492 | #define MEMERR_L2C_L2ESR_PAGE_OFFSET 0x0004 |
| 493 | #define MEMERR_L2C_L2ESR_ERRSYN_RD(src) (((src) & 0xFF000000) >> 24) |
| 494 | #define MEMERR_L2C_L2ESR_ERRWAY_RD(src) (((src) & 0x00FC0000) >> 18) |
| 495 | #define MEMERR_L2C_L2ESR_ERRCPU_RD(src) (((src) & 0x00020000) >> 17) |
| 496 | #define MEMERR_L2C_L2ESR_ERRGROUP_RD(src) (((src) & 0x0000E000) >> 13) |
| 497 | #define MEMERR_L2C_L2ESR_ERRACTION_RD(src) (((src) & 0x00001C00) >> 10) |
| 498 | #define MEMERR_L2C_L2ESR_ERRTYPE_RD(src) (((src) & 0x00000300) >> 8) |
| 499 | #define MEMERR_L2C_L2ESR_MULTUCERR_MASK BIT(3) |
| 500 | #define MEMERR_L2C_L2ESR_MULTICERR_MASK BIT(2) |
| 501 | #define MEMERR_L2C_L2ESR_UCERR_MASK BIT(1) |
| 502 | #define MEMERR_L2C_L2ESR_ERR_MASK BIT(0) |
| 503 | #define MEMERR_L2C_L2EALR_PAGE_OFFSET 0x0008 |
| 504 | #define CPUX_L2C_L2RTOCR_PAGE_OFFSET 0x0010 |
| 505 | #define MEMERR_L2C_L2EAHR_PAGE_OFFSET 0x000c |
| 506 | #define CPUX_L2C_L2RTOSR_PAGE_OFFSET 0x0014 |
| 507 | #define MEMERR_L2C_L2RTOSR_MULTERR_MASK BIT(1) |
| 508 | #define MEMERR_L2C_L2RTOSR_ERR_MASK BIT(0) |
| 509 | #define CPUX_L2C_L2RTOALR_PAGE_OFFSET 0x0018 |
| 510 | #define CPUX_L2C_L2RTOAHR_PAGE_OFFSET 0x001c |
| 511 | #define MEMERR_L2C_L2ESRA_PAGE_OFFSET 0x0804 |
| 512 | |
| 513 | /* |
| 514 | * Processor Module Domain (PMD) context - Context for a pair of processsors. |
| 515 | * Each PMD consists of 2 CPUs and a shared L2 cache. Each CPU consists of |
| 516 | * its own L1 cache. |
| 517 | */ |
| 518 | struct xgene_edac_pmd_ctx { |
| 519 | struct list_head next; |
| 520 | struct device ddev; |
| 521 | char *name; |
| 522 | struct xgene_edac *edac; |
| 523 | struct edac_device_ctl_info *edac_dev; |
| 524 | void __iomem *pmd_csr; |
| 525 | u32 pmd; |
Arnd Bergmann | 451bb7f | 2015-06-01 16:09:35 -0600 | [diff] [blame] | 526 | int version; |
Loc Ho | 0d44293 | 2015-05-22 17:32:59 -0600 | [diff] [blame] | 527 | }; |
| 528 | |
| 529 | static void xgene_edac_pmd_l1_check(struct edac_device_ctl_info *edac_dev, |
| 530 | int cpu_idx) |
| 531 | { |
| 532 | struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; |
| 533 | void __iomem *pg_f; |
| 534 | u32 val; |
| 535 | |
| 536 | pg_f = ctx->pmd_csr + cpu_idx * CPU_CSR_STRIDE + CPU_MEMERR_CPU_PAGE; |
| 537 | |
| 538 | val = readl(pg_f + MEMERR_CPU_ICFESR_PAGE_OFFSET); |
| 539 | if (val) { |
| 540 | dev_err(edac_dev->dev, |
| 541 | "CPU%d L1 memory error ICF 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X\n", |
| 542 | ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val, |
| 543 | MEMERR_CPU_ICFESR_ERRWAY_RD(val), |
| 544 | MEMERR_CPU_ICFESR_ERRINDEX_RD(val), |
| 545 | MEMERR_CPU_ICFESR_ERRINFO_RD(val)); |
| 546 | if (val & MEMERR_CPU_ICFESR_CERR_MASK) |
| 547 | dev_err(edac_dev->dev, |
| 548 | "One or more correctable error\n"); |
| 549 | if (val & MEMERR_CPU_ICFESR_MULTCERR_MASK) |
| 550 | dev_err(edac_dev->dev, "Multiple correctable error\n"); |
| 551 | switch (MEMERR_CPU_ICFESR_ERRTYPE_RD(val)) { |
| 552 | case 1: |
| 553 | dev_err(edac_dev->dev, "L1 TLB multiple hit\n"); |
| 554 | break; |
| 555 | case 2: |
| 556 | dev_err(edac_dev->dev, "Way select multiple hit\n"); |
| 557 | break; |
| 558 | case 3: |
| 559 | dev_err(edac_dev->dev, "Physical tag parity error\n"); |
| 560 | break; |
| 561 | case 4: |
| 562 | case 5: |
| 563 | dev_err(edac_dev->dev, "L1 data parity error\n"); |
| 564 | break; |
| 565 | case 6: |
| 566 | dev_err(edac_dev->dev, "L1 pre-decode parity error\n"); |
| 567 | break; |
| 568 | } |
| 569 | |
| 570 | /* Clear any HW errors */ |
| 571 | writel(val, pg_f + MEMERR_CPU_ICFESR_PAGE_OFFSET); |
| 572 | |
| 573 | if (val & (MEMERR_CPU_ICFESR_CERR_MASK | |
| 574 | MEMERR_CPU_ICFESR_MULTCERR_MASK)) |
| 575 | edac_device_handle_ce(edac_dev, 0, 0, |
| 576 | edac_dev->ctl_name); |
| 577 | } |
| 578 | |
| 579 | val = readl(pg_f + MEMERR_CPU_LSUESR_PAGE_OFFSET); |
| 580 | if (val) { |
| 581 | dev_err(edac_dev->dev, |
| 582 | "CPU%d memory error LSU 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X\n", |
| 583 | ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val, |
| 584 | MEMERR_CPU_LSUESR_ERRWAY_RD(val), |
| 585 | MEMERR_CPU_LSUESR_ERRINDEX_RD(val), |
| 586 | MEMERR_CPU_LSUESR_ERRINFO_RD(val)); |
| 587 | if (val & MEMERR_CPU_LSUESR_CERR_MASK) |
| 588 | dev_err(edac_dev->dev, |
| 589 | "One or more correctable error\n"); |
| 590 | if (val & MEMERR_CPU_LSUESR_MULTCERR_MASK) |
| 591 | dev_err(edac_dev->dev, "Multiple correctable error\n"); |
| 592 | switch (MEMERR_CPU_LSUESR_ERRTYPE_RD(val)) { |
| 593 | case 0: |
| 594 | dev_err(edac_dev->dev, "Load tag error\n"); |
| 595 | break; |
| 596 | case 1: |
| 597 | dev_err(edac_dev->dev, "Load data error\n"); |
| 598 | break; |
| 599 | case 2: |
| 600 | dev_err(edac_dev->dev, "WSL multihit error\n"); |
| 601 | break; |
| 602 | case 3: |
| 603 | dev_err(edac_dev->dev, "Store tag error\n"); |
| 604 | break; |
| 605 | case 4: |
| 606 | dev_err(edac_dev->dev, |
| 607 | "DTB multihit from load pipeline error\n"); |
| 608 | break; |
| 609 | case 5: |
| 610 | dev_err(edac_dev->dev, |
| 611 | "DTB multihit from store pipeline error\n"); |
| 612 | break; |
| 613 | } |
| 614 | |
| 615 | /* Clear any HW errors */ |
| 616 | writel(val, pg_f + MEMERR_CPU_LSUESR_PAGE_OFFSET); |
| 617 | |
| 618 | if (val & (MEMERR_CPU_LSUESR_CERR_MASK | |
| 619 | MEMERR_CPU_LSUESR_MULTCERR_MASK)) |
| 620 | edac_device_handle_ce(edac_dev, 0, 0, |
| 621 | edac_dev->ctl_name); |
| 622 | } |
| 623 | |
| 624 | val = readl(pg_f + MEMERR_CPU_MMUESR_PAGE_OFFSET); |
| 625 | if (val) { |
| 626 | dev_err(edac_dev->dev, |
| 627 | "CPU%d memory error MMU 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X %s\n", |
| 628 | ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val, |
| 629 | MEMERR_CPU_MMUESR_ERRWAY_RD(val), |
| 630 | MEMERR_CPU_MMUESR_ERRINDEX_RD(val), |
| 631 | MEMERR_CPU_MMUESR_ERRINFO_RD(val), |
| 632 | val & MEMERR_CPU_MMUESR_ERRREQSTR_LSU_MASK ? "LSU" : |
| 633 | "ICF"); |
| 634 | if (val & MEMERR_CPU_MMUESR_CERR_MASK) |
| 635 | dev_err(edac_dev->dev, |
| 636 | "One or more correctable error\n"); |
| 637 | if (val & MEMERR_CPU_MMUESR_MULTCERR_MASK) |
| 638 | dev_err(edac_dev->dev, "Multiple correctable error\n"); |
| 639 | switch (MEMERR_CPU_MMUESR_ERRTYPE_RD(val)) { |
| 640 | case 0: |
| 641 | dev_err(edac_dev->dev, "Stage 1 UTB hit error\n"); |
| 642 | break; |
| 643 | case 1: |
| 644 | dev_err(edac_dev->dev, "Stage 1 UTB miss error\n"); |
| 645 | break; |
| 646 | case 2: |
| 647 | dev_err(edac_dev->dev, "Stage 1 UTB allocate error\n"); |
| 648 | break; |
| 649 | case 3: |
| 650 | dev_err(edac_dev->dev, |
| 651 | "TMO operation single bank error\n"); |
| 652 | break; |
| 653 | case 4: |
| 654 | dev_err(edac_dev->dev, "Stage 2 UTB error\n"); |
| 655 | break; |
| 656 | case 5: |
| 657 | dev_err(edac_dev->dev, "Stage 2 UTB miss error\n"); |
| 658 | break; |
| 659 | case 6: |
| 660 | dev_err(edac_dev->dev, "Stage 2 UTB allocate error\n"); |
| 661 | break; |
| 662 | case 7: |
| 663 | dev_err(edac_dev->dev, |
| 664 | "TMO operation multiple bank error\n"); |
| 665 | break; |
| 666 | } |
| 667 | |
| 668 | /* Clear any HW errors */ |
| 669 | writel(val, pg_f + MEMERR_CPU_MMUESR_PAGE_OFFSET); |
| 670 | |
| 671 | edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name); |
| 672 | } |
| 673 | } |
| 674 | |
| 675 | static void xgene_edac_pmd_l2_check(struct edac_device_ctl_info *edac_dev) |
| 676 | { |
| 677 | struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; |
| 678 | void __iomem *pg_d; |
| 679 | void __iomem *pg_e; |
| 680 | u32 val_hi; |
| 681 | u32 val_lo; |
| 682 | u32 val; |
| 683 | |
| 684 | /* Check L2 */ |
| 685 | pg_e = ctx->pmd_csr + CPU_MEMERR_L2C_PAGE; |
| 686 | val = readl(pg_e + MEMERR_L2C_L2ESR_PAGE_OFFSET); |
| 687 | if (val) { |
| 688 | val_lo = readl(pg_e + MEMERR_L2C_L2EALR_PAGE_OFFSET); |
| 689 | val_hi = readl(pg_e + MEMERR_L2C_L2EAHR_PAGE_OFFSET); |
| 690 | dev_err(edac_dev->dev, |
| 691 | "PMD%d memory error L2C L2ESR 0x%08X @ 0x%08X.%08X\n", |
| 692 | ctx->pmd, val, val_hi, val_lo); |
| 693 | dev_err(edac_dev->dev, |
| 694 | "ErrSyndrome 0x%02X ErrWay 0x%02X ErrCpu %d ErrGroup 0x%02X ErrAction 0x%02X\n", |
| 695 | MEMERR_L2C_L2ESR_ERRSYN_RD(val), |
| 696 | MEMERR_L2C_L2ESR_ERRWAY_RD(val), |
| 697 | MEMERR_L2C_L2ESR_ERRCPU_RD(val), |
| 698 | MEMERR_L2C_L2ESR_ERRGROUP_RD(val), |
| 699 | MEMERR_L2C_L2ESR_ERRACTION_RD(val)); |
| 700 | |
| 701 | if (val & MEMERR_L2C_L2ESR_ERR_MASK) |
| 702 | dev_err(edac_dev->dev, |
| 703 | "One or more correctable error\n"); |
| 704 | if (val & MEMERR_L2C_L2ESR_MULTICERR_MASK) |
| 705 | dev_err(edac_dev->dev, "Multiple correctable error\n"); |
| 706 | if (val & MEMERR_L2C_L2ESR_UCERR_MASK) |
| 707 | dev_err(edac_dev->dev, |
| 708 | "One or more uncorrectable error\n"); |
| 709 | if (val & MEMERR_L2C_L2ESR_MULTUCERR_MASK) |
| 710 | dev_err(edac_dev->dev, |
| 711 | "Multiple uncorrectable error\n"); |
| 712 | |
| 713 | switch (MEMERR_L2C_L2ESR_ERRTYPE_RD(val)) { |
| 714 | case 0: |
| 715 | dev_err(edac_dev->dev, "Outbound SDB parity error\n"); |
| 716 | break; |
| 717 | case 1: |
| 718 | dev_err(edac_dev->dev, "Inbound SDB parity error\n"); |
| 719 | break; |
| 720 | case 2: |
| 721 | dev_err(edac_dev->dev, "Tag ECC error\n"); |
| 722 | break; |
| 723 | case 3: |
| 724 | dev_err(edac_dev->dev, "Data ECC error\n"); |
| 725 | break; |
| 726 | } |
| 727 | |
| 728 | /* Clear any HW errors */ |
| 729 | writel(val, pg_e + MEMERR_L2C_L2ESR_PAGE_OFFSET); |
| 730 | |
| 731 | if (val & (MEMERR_L2C_L2ESR_ERR_MASK | |
| 732 | MEMERR_L2C_L2ESR_MULTICERR_MASK)) |
| 733 | edac_device_handle_ce(edac_dev, 0, 0, |
| 734 | edac_dev->ctl_name); |
| 735 | if (val & (MEMERR_L2C_L2ESR_UCERR_MASK | |
| 736 | MEMERR_L2C_L2ESR_MULTUCERR_MASK)) |
| 737 | edac_device_handle_ue(edac_dev, 0, 0, |
| 738 | edac_dev->ctl_name); |
| 739 | } |
| 740 | |
| 741 | /* Check if any memory request timed out on L2 cache */ |
| 742 | pg_d = ctx->pmd_csr + CPU_L2C_PAGE; |
| 743 | val = readl(pg_d + CPUX_L2C_L2RTOSR_PAGE_OFFSET); |
| 744 | if (val) { |
| 745 | val_lo = readl(pg_d + CPUX_L2C_L2RTOALR_PAGE_OFFSET); |
| 746 | val_hi = readl(pg_d + CPUX_L2C_L2RTOAHR_PAGE_OFFSET); |
| 747 | dev_err(edac_dev->dev, |
| 748 | "PMD%d L2C error L2C RTOSR 0x%08X @ 0x%08X.%08X\n", |
| 749 | ctx->pmd, val, val_hi, val_lo); |
| 750 | writel(val, pg_d + CPUX_L2C_L2RTOSR_PAGE_OFFSET); |
| 751 | } |
| 752 | } |
| 753 | |
| 754 | static void xgene_edac_pmd_check(struct edac_device_ctl_info *edac_dev) |
| 755 | { |
| 756 | struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; |
| 757 | unsigned int pcp_hp_stat; |
| 758 | int i; |
| 759 | |
| 760 | xgene_edac_pcp_rd(ctx->edac, PCPHPERRINTSTS, &pcp_hp_stat); |
| 761 | if (!((PMD0_MERR_MASK << ctx->pmd) & pcp_hp_stat)) |
| 762 | return; |
| 763 | |
| 764 | /* Check CPU L1 error */ |
| 765 | for (i = 0; i < MAX_CPU_PER_PMD; i++) |
| 766 | xgene_edac_pmd_l1_check(edac_dev, i); |
| 767 | |
| 768 | /* Check CPU L2 error */ |
| 769 | xgene_edac_pmd_l2_check(edac_dev); |
| 770 | } |
| 771 | |
| 772 | static void xgene_edac_pmd_cpu_hw_cfg(struct edac_device_ctl_info *edac_dev, |
| 773 | int cpu) |
| 774 | { |
| 775 | struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; |
| 776 | void __iomem *pg_f = ctx->pmd_csr + cpu * CPU_CSR_STRIDE + |
| 777 | CPU_MEMERR_CPU_PAGE; |
| 778 | |
| 779 | /* |
| 780 | * Enable CPU memory error: |
| 781 | * MEMERR_CPU_ICFESRA, MEMERR_CPU_LSUESRA, and MEMERR_CPU_MMUESRA |
| 782 | */ |
| 783 | writel(0x00000301, pg_f + MEMERR_CPU_ICFECR_PAGE_OFFSET); |
| 784 | writel(0x00000301, pg_f + MEMERR_CPU_LSUECR_PAGE_OFFSET); |
| 785 | writel(0x00000101, pg_f + MEMERR_CPU_MMUECR_PAGE_OFFSET); |
| 786 | } |
| 787 | |
Loc Ho | 0d44293 | 2015-05-22 17:32:59 -0600 | [diff] [blame] | 788 | static void xgene_edac_pmd_hw_cfg(struct edac_device_ctl_info *edac_dev) |
| 789 | { |
| 790 | struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; |
| 791 | void __iomem *pg_d = ctx->pmd_csr + CPU_L2C_PAGE; |
| 792 | void __iomem *pg_e = ctx->pmd_csr + CPU_MEMERR_L2C_PAGE; |
| 793 | |
| 794 | /* Enable PMD memory error - MEMERR_L2C_L2ECR and L2C_L2RTOCR */ |
| 795 | writel(0x00000703, pg_e + MEMERR_L2C_L2ECR_PAGE_OFFSET); |
| 796 | /* Configure L2C HW request time out feature if supported */ |
Arnd Bergmann | 451bb7f | 2015-06-01 16:09:35 -0600 | [diff] [blame] | 797 | if (ctx->version > 1) |
Loc Ho | 0d44293 | 2015-05-22 17:32:59 -0600 | [diff] [blame] | 798 | writel(0x00000119, pg_d + CPUX_L2C_L2RTOCR_PAGE_OFFSET); |
| 799 | } |
| 800 | |
| 801 | static void xgene_edac_pmd_hw_ctl(struct edac_device_ctl_info *edac_dev, |
| 802 | bool enable) |
| 803 | { |
| 804 | struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; |
| 805 | int i; |
| 806 | |
| 807 | /* Enable PMD error interrupt */ |
| 808 | if (edac_dev->op_state == OP_RUNNING_INTERRUPT) { |
| 809 | if (enable) |
| 810 | xgene_edac_pcp_clrbits(ctx->edac, PCPHPERRINTMSK, |
| 811 | PMD0_MERR_MASK << ctx->pmd); |
| 812 | else |
| 813 | xgene_edac_pcp_setbits(ctx->edac, PCPHPERRINTMSK, |
| 814 | PMD0_MERR_MASK << ctx->pmd); |
| 815 | } |
| 816 | |
| 817 | if (enable) { |
| 818 | xgene_edac_pmd_hw_cfg(edac_dev); |
| 819 | |
| 820 | /* Two CPUs per a PMD */ |
| 821 | for (i = 0; i < MAX_CPU_PER_PMD; i++) |
| 822 | xgene_edac_pmd_cpu_hw_cfg(edac_dev, i); |
| 823 | } |
| 824 | } |
| 825 | |
| 826 | static ssize_t xgene_edac_pmd_l1_inject_ctrl_write(struct file *file, |
| 827 | const char __user *data, |
| 828 | size_t count, loff_t *ppos) |
| 829 | { |
| 830 | struct edac_device_ctl_info *edac_dev = file->private_data; |
| 831 | struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; |
| 832 | void __iomem *cpux_pg_f; |
| 833 | int i; |
| 834 | |
| 835 | for (i = 0; i < MAX_CPU_PER_PMD; i++) { |
| 836 | cpux_pg_f = ctx->pmd_csr + i * CPU_CSR_STRIDE + |
| 837 | CPU_MEMERR_CPU_PAGE; |
| 838 | |
| 839 | writel(MEMERR_CPU_ICFESR_MULTCERR_MASK | |
| 840 | MEMERR_CPU_ICFESR_CERR_MASK, |
| 841 | cpux_pg_f + MEMERR_CPU_ICFESRA_PAGE_OFFSET); |
| 842 | writel(MEMERR_CPU_LSUESR_MULTCERR_MASK | |
| 843 | MEMERR_CPU_LSUESR_CERR_MASK, |
| 844 | cpux_pg_f + MEMERR_CPU_LSUESRA_PAGE_OFFSET); |
| 845 | writel(MEMERR_CPU_MMUESR_MULTCERR_MASK | |
| 846 | MEMERR_CPU_MMUESR_CERR_MASK, |
| 847 | cpux_pg_f + MEMERR_CPU_MMUESRA_PAGE_OFFSET); |
| 848 | } |
| 849 | return count; |
| 850 | } |
| 851 | |
| 852 | static ssize_t xgene_edac_pmd_l2_inject_ctrl_write(struct file *file, |
| 853 | const char __user *data, |
| 854 | size_t count, loff_t *ppos) |
| 855 | { |
| 856 | struct edac_device_ctl_info *edac_dev = file->private_data; |
| 857 | struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; |
| 858 | void __iomem *pg_e = ctx->pmd_csr + CPU_MEMERR_L2C_PAGE; |
| 859 | |
| 860 | writel(MEMERR_L2C_L2ESR_MULTUCERR_MASK | |
| 861 | MEMERR_L2C_L2ESR_MULTICERR_MASK | |
| 862 | MEMERR_L2C_L2ESR_UCERR_MASK | |
| 863 | MEMERR_L2C_L2ESR_ERR_MASK, |
| 864 | pg_e + MEMERR_L2C_L2ESRA_PAGE_OFFSET); |
| 865 | return count; |
| 866 | } |
| 867 | |
| 868 | static const struct file_operations xgene_edac_pmd_debug_inject_fops[] = { |
| 869 | { |
| 870 | .open = simple_open, |
| 871 | .write = xgene_edac_pmd_l1_inject_ctrl_write, |
| 872 | .llseek = generic_file_llseek, }, |
| 873 | { |
| 874 | .open = simple_open, |
| 875 | .write = xgene_edac_pmd_l2_inject_ctrl_write, |
| 876 | .llseek = generic_file_llseek, }, |
| 877 | { } |
| 878 | }; |
| 879 | |
| 880 | static void xgene_edac_pmd_create_debugfs_nodes( |
| 881 | struct edac_device_ctl_info *edac_dev) |
| 882 | { |
| 883 | struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; |
| 884 | struct dentry *edac_debugfs; |
| 885 | char name[30]; |
| 886 | |
| 887 | if (!IS_ENABLED(CONFIG_EDAC_DEBUG)) |
| 888 | return; |
| 889 | |
| 890 | /* |
| 891 | * Todo: Switch to common EDAC debug file system for edac device |
| 892 | * when available. |
| 893 | */ |
| 894 | if (!ctx->edac->dfs) { |
| 895 | ctx->edac->dfs = debugfs_create_dir(edac_dev->dev->kobj.name, |
| 896 | NULL); |
| 897 | if (!ctx->edac->dfs) |
| 898 | return; |
| 899 | } |
| 900 | sprintf(name, "PMD%d", ctx->pmd); |
| 901 | edac_debugfs = debugfs_create_dir(name, ctx->edac->dfs); |
| 902 | if (!edac_debugfs) |
| 903 | return; |
| 904 | |
| 905 | debugfs_create_file("l1_inject_ctrl", S_IWUSR, edac_debugfs, edac_dev, |
| 906 | &xgene_edac_pmd_debug_inject_fops[0]); |
| 907 | debugfs_create_file("l2_inject_ctrl", S_IWUSR, edac_debugfs, edac_dev, |
| 908 | &xgene_edac_pmd_debug_inject_fops[1]); |
| 909 | } |
| 910 | |
| 911 | static int xgene_edac_pmd_available(u32 efuse, int pmd) |
| 912 | { |
| 913 | return (efuse & (1 << pmd)) ? 0 : 1; |
| 914 | } |
| 915 | |
Arnd Bergmann | 451bb7f | 2015-06-01 16:09:35 -0600 | [diff] [blame] | 916 | static int xgene_edac_pmd_add(struct xgene_edac *edac, struct device_node *np, |
| 917 | int version) |
Loc Ho | 0d44293 | 2015-05-22 17:32:59 -0600 | [diff] [blame] | 918 | { |
| 919 | struct edac_device_ctl_info *edac_dev; |
| 920 | struct xgene_edac_pmd_ctx *ctx; |
| 921 | struct resource res; |
| 922 | char edac_name[10]; |
| 923 | u32 pmd; |
| 924 | int rc; |
| 925 | u32 val; |
| 926 | |
| 927 | if (!devres_open_group(edac->dev, xgene_edac_pmd_add, GFP_KERNEL)) |
| 928 | return -ENOMEM; |
| 929 | |
| 930 | /* Determine if this PMD is disabled */ |
| 931 | if (of_property_read_u32(np, "pmd-controller", &pmd)) { |
| 932 | dev_err(edac->dev, "no pmd-controller property\n"); |
| 933 | rc = -ENODEV; |
| 934 | goto err_group; |
| 935 | } |
| 936 | rc = regmap_read(edac->efuse_map, 0, &val); |
| 937 | if (rc) |
| 938 | goto err_group; |
| 939 | if (!xgene_edac_pmd_available(val, pmd)) { |
| 940 | rc = -ENODEV; |
| 941 | goto err_group; |
| 942 | } |
| 943 | |
| 944 | sprintf(edac_name, "l2c%d", pmd); |
| 945 | edac_dev = edac_device_alloc_ctl_info(sizeof(*ctx), |
| 946 | edac_name, 1, "l2c", 1, 2, NULL, |
| 947 | 0, edac_device_alloc_index()); |
| 948 | if (!edac_dev) { |
| 949 | rc = -ENOMEM; |
| 950 | goto err_group; |
| 951 | } |
| 952 | |
| 953 | ctx = edac_dev->pvt_info; |
| 954 | ctx->name = "xgene_pmd_err"; |
| 955 | ctx->pmd = pmd; |
| 956 | ctx->edac = edac; |
| 957 | ctx->edac_dev = edac_dev; |
| 958 | ctx->ddev = *edac->dev; |
Arnd Bergmann | 451bb7f | 2015-06-01 16:09:35 -0600 | [diff] [blame] | 959 | ctx->version = version; |
Loc Ho | 0d44293 | 2015-05-22 17:32:59 -0600 | [diff] [blame] | 960 | edac_dev->dev = &ctx->ddev; |
| 961 | edac_dev->ctl_name = ctx->name; |
| 962 | edac_dev->dev_name = ctx->name; |
| 963 | edac_dev->mod_name = EDAC_MOD_STR; |
| 964 | |
| 965 | rc = of_address_to_resource(np, 0, &res); |
| 966 | if (rc < 0) { |
| 967 | dev_err(edac->dev, "no PMD resource address\n"); |
| 968 | goto err_free; |
| 969 | } |
| 970 | ctx->pmd_csr = devm_ioremap_resource(edac->dev, &res); |
| 971 | if (IS_ERR(ctx->pmd_csr)) { |
| 972 | dev_err(edac->dev, |
| 973 | "devm_ioremap_resource failed for PMD resource address\n"); |
| 974 | rc = PTR_ERR(ctx->pmd_csr); |
| 975 | goto err_free; |
| 976 | } |
| 977 | |
| 978 | if (edac_op_state == EDAC_OPSTATE_POLL) |
| 979 | edac_dev->edac_check = xgene_edac_pmd_check; |
| 980 | |
| 981 | xgene_edac_pmd_create_debugfs_nodes(edac_dev); |
| 982 | |
| 983 | rc = edac_device_add_device(edac_dev); |
| 984 | if (rc > 0) { |
| 985 | dev_err(edac->dev, "edac_device_add_device failed\n"); |
| 986 | rc = -ENOMEM; |
| 987 | goto err_free; |
| 988 | } |
| 989 | |
| 990 | if (edac_op_state == EDAC_OPSTATE_INT) |
| 991 | edac_dev->op_state = OP_RUNNING_INTERRUPT; |
| 992 | |
| 993 | list_add(&ctx->next, &edac->pmds); |
| 994 | |
| 995 | xgene_edac_pmd_hw_ctl(edac_dev, 1); |
| 996 | |
| 997 | devres_remove_group(edac->dev, xgene_edac_pmd_add); |
| 998 | |
| 999 | dev_info(edac->dev, "X-Gene EDAC PMD%d registered\n", ctx->pmd); |
| 1000 | return 0; |
| 1001 | |
| 1002 | err_free: |
| 1003 | edac_device_free_ctl_info(edac_dev); |
| 1004 | err_group: |
| 1005 | devres_release_group(edac->dev, xgene_edac_pmd_add); |
| 1006 | return rc; |
| 1007 | } |
| 1008 | |
| 1009 | static int xgene_edac_pmd_remove(struct xgene_edac_pmd_ctx *pmd) |
| 1010 | { |
| 1011 | struct edac_device_ctl_info *edac_dev = pmd->edac_dev; |
| 1012 | |
| 1013 | xgene_edac_pmd_hw_ctl(edac_dev, 0); |
| 1014 | edac_device_del_device(edac_dev->dev); |
| 1015 | edac_device_free_ctl_info(edac_dev); |
| 1016 | return 0; |
| 1017 | } |
| 1018 | |
| 1019 | static irqreturn_t xgene_edac_isr(int irq, void *dev_id) |
| 1020 | { |
| 1021 | struct xgene_edac *ctx = dev_id; |
| 1022 | struct xgene_edac_pmd_ctx *pmd; |
| 1023 | unsigned int pcp_hp_stat; |
| 1024 | unsigned int pcp_lp_stat; |
| 1025 | |
| 1026 | xgene_edac_pcp_rd(ctx, PCPHPERRINTSTS, &pcp_hp_stat); |
| 1027 | xgene_edac_pcp_rd(ctx, PCPLPERRINTSTS, &pcp_lp_stat); |
| 1028 | if ((MCU_UNCORR_ERR_MASK & pcp_hp_stat) || |
| 1029 | (MCU_CTL_ERR_MASK & pcp_hp_stat) || |
| 1030 | (MCU_CORR_ERR_MASK & pcp_lp_stat)) { |
| 1031 | struct xgene_edac_mc_ctx *mcu; |
| 1032 | |
| 1033 | list_for_each_entry(mcu, &ctx->mcus, next) { |
| 1034 | xgene_edac_mc_check(mcu->mci); |
| 1035 | } |
| 1036 | } |
| 1037 | |
| 1038 | list_for_each_entry(pmd, &ctx->pmds, next) { |
| 1039 | if ((PMD0_MERR_MASK << pmd->pmd) & pcp_hp_stat) |
| 1040 | xgene_edac_pmd_check(pmd->edac_dev); |
| 1041 | } |
| 1042 | |
| 1043 | return IRQ_HANDLED; |
| 1044 | } |
| 1045 | |
| 1046 | static int xgene_edac_probe(struct platform_device *pdev) |
| 1047 | { |
| 1048 | struct xgene_edac *edac; |
| 1049 | struct device_node *child; |
| 1050 | struct resource *res; |
| 1051 | int rc; |
| 1052 | |
| 1053 | edac = devm_kzalloc(&pdev->dev, sizeof(*edac), GFP_KERNEL); |
| 1054 | if (!edac) |
| 1055 | return -ENOMEM; |
| 1056 | |
| 1057 | edac->dev = &pdev->dev; |
| 1058 | platform_set_drvdata(pdev, edac); |
| 1059 | INIT_LIST_HEAD(&edac->mcus); |
| 1060 | INIT_LIST_HEAD(&edac->pmds); |
| 1061 | spin_lock_init(&edac->lock); |
| 1062 | mutex_init(&edac->mc_lock); |
| 1063 | |
| 1064 | edac->csw_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node, |
| 1065 | "regmap-csw"); |
| 1066 | if (IS_ERR(edac->csw_map)) { |
| 1067 | dev_err(edac->dev, "unable to get syscon regmap csw\n"); |
| 1068 | rc = PTR_ERR(edac->csw_map); |
| 1069 | goto out_err; |
| 1070 | } |
| 1071 | |
| 1072 | edac->mcba_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node, |
| 1073 | "regmap-mcba"); |
| 1074 | if (IS_ERR(edac->mcba_map)) { |
| 1075 | dev_err(edac->dev, "unable to get syscon regmap mcba\n"); |
| 1076 | rc = PTR_ERR(edac->mcba_map); |
| 1077 | goto out_err; |
| 1078 | } |
| 1079 | |
| 1080 | edac->mcbb_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node, |
| 1081 | "regmap-mcbb"); |
| 1082 | if (IS_ERR(edac->mcbb_map)) { |
| 1083 | dev_err(edac->dev, "unable to get syscon regmap mcbb\n"); |
| 1084 | rc = PTR_ERR(edac->mcbb_map); |
| 1085 | goto out_err; |
| 1086 | } |
| 1087 | edac->efuse_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node, |
| 1088 | "regmap-efuse"); |
| 1089 | if (IS_ERR(edac->efuse_map)) { |
| 1090 | dev_err(edac->dev, "unable to get syscon regmap efuse\n"); |
| 1091 | rc = PTR_ERR(edac->efuse_map); |
| 1092 | goto out_err; |
| 1093 | } |
| 1094 | |
| 1095 | res = platform_get_resource(pdev, IORESOURCE_MEM, 0); |
| 1096 | edac->pcp_csr = devm_ioremap_resource(&pdev->dev, res); |
| 1097 | if (IS_ERR(edac->pcp_csr)) { |
| 1098 | dev_err(&pdev->dev, "no PCP resource address\n"); |
| 1099 | rc = PTR_ERR(edac->pcp_csr); |
| 1100 | goto out_err; |
| 1101 | } |
| 1102 | |
| 1103 | if (edac_op_state == EDAC_OPSTATE_INT) { |
| 1104 | int irq; |
| 1105 | int i; |
| 1106 | |
| 1107 | for (i = 0; i < 3; i++) { |
| 1108 | irq = platform_get_irq(pdev, i); |
| 1109 | if (irq < 0) { |
| 1110 | dev_err(&pdev->dev, "No IRQ resource\n"); |
| 1111 | rc = -EINVAL; |
| 1112 | goto out_err; |
| 1113 | } |
| 1114 | rc = devm_request_irq(&pdev->dev, irq, |
| 1115 | xgene_edac_isr, IRQF_SHARED, |
| 1116 | dev_name(&pdev->dev), edac); |
| 1117 | if (rc) { |
| 1118 | dev_err(&pdev->dev, |
| 1119 | "Could not request IRQ %d\n", irq); |
| 1120 | goto out_err; |
| 1121 | } |
| 1122 | } |
| 1123 | } |
| 1124 | |
| 1125 | for_each_child_of_node(pdev->dev.of_node, child) { |
| 1126 | if (!of_device_is_available(child)) |
| 1127 | continue; |
| 1128 | if (of_device_is_compatible(child, "apm,xgene-edac-mc")) |
| 1129 | xgene_edac_mc_add(edac, child); |
| 1130 | if (of_device_is_compatible(child, "apm,xgene-edac-pmd")) |
Arnd Bergmann | 451bb7f | 2015-06-01 16:09:35 -0600 | [diff] [blame] | 1131 | xgene_edac_pmd_add(edac, child, 1); |
| 1132 | if (of_device_is_compatible(child, "apm,xgene-edac-pmd-v2")) |
| 1133 | xgene_edac_pmd_add(edac, child, 2); |
Loc Ho | 0d44293 | 2015-05-22 17:32:59 -0600 | [diff] [blame] | 1134 | } |
| 1135 | |
| 1136 | return 0; |
| 1137 | |
| 1138 | out_err: |
| 1139 | return rc; |
| 1140 | } |
| 1141 | |
| 1142 | static int xgene_edac_remove(struct platform_device *pdev) |
| 1143 | { |
| 1144 | struct xgene_edac *edac = dev_get_drvdata(&pdev->dev); |
| 1145 | struct xgene_edac_mc_ctx *mcu; |
| 1146 | struct xgene_edac_mc_ctx *temp_mcu; |
| 1147 | struct xgene_edac_pmd_ctx *pmd; |
| 1148 | struct xgene_edac_pmd_ctx *temp_pmd; |
| 1149 | |
| 1150 | list_for_each_entry_safe(mcu, temp_mcu, &edac->mcus, next) { |
| 1151 | xgene_edac_mc_remove(mcu); |
| 1152 | } |
| 1153 | |
| 1154 | list_for_each_entry_safe(pmd, temp_pmd, &edac->pmds, next) { |
| 1155 | xgene_edac_pmd_remove(pmd); |
| 1156 | } |
| 1157 | return 0; |
| 1158 | } |
| 1159 | |
| 1160 | static const struct of_device_id xgene_edac_of_match[] = { |
| 1161 | { .compatible = "apm,xgene-edac" }, |
| 1162 | {}, |
| 1163 | }; |
| 1164 | MODULE_DEVICE_TABLE(of, xgene_edac_of_match); |
| 1165 | |
| 1166 | static struct platform_driver xgene_edac_driver = { |
| 1167 | .probe = xgene_edac_probe, |
| 1168 | .remove = xgene_edac_remove, |
| 1169 | .driver = { |
| 1170 | .name = "xgene-edac", |
Loc Ho | 0d44293 | 2015-05-22 17:32:59 -0600 | [diff] [blame] | 1171 | .of_match_table = xgene_edac_of_match, |
| 1172 | }, |
| 1173 | }; |
| 1174 | |
| 1175 | static int __init xgene_edac_init(void) |
| 1176 | { |
| 1177 | int rc; |
| 1178 | |
| 1179 | /* Make sure error reporting method is sane */ |
| 1180 | switch (edac_op_state) { |
| 1181 | case EDAC_OPSTATE_POLL: |
| 1182 | case EDAC_OPSTATE_INT: |
| 1183 | break; |
| 1184 | default: |
| 1185 | edac_op_state = EDAC_OPSTATE_INT; |
| 1186 | break; |
| 1187 | } |
| 1188 | |
| 1189 | rc = platform_driver_register(&xgene_edac_driver); |
| 1190 | if (rc) { |
| 1191 | edac_printk(KERN_ERR, EDAC_MOD_STR, |
| 1192 | "EDAC fails to register\n"); |
| 1193 | goto reg_failed; |
| 1194 | } |
| 1195 | |
| 1196 | return 0; |
| 1197 | |
| 1198 | reg_failed: |
| 1199 | return rc; |
| 1200 | } |
| 1201 | module_init(xgene_edac_init); |
| 1202 | |
| 1203 | static void __exit xgene_edac_exit(void) |
| 1204 | { |
| 1205 | platform_driver_unregister(&xgene_edac_driver); |
| 1206 | } |
| 1207 | module_exit(xgene_edac_exit); |
| 1208 | |
| 1209 | MODULE_LICENSE("GPL"); |
| 1210 | MODULE_AUTHOR("Feng Kan <fkan@apm.com>"); |
| 1211 | MODULE_DESCRIPTION("APM X-Gene EDAC driver"); |
| 1212 | module_param(edac_op_state, int, 0444); |
| 1213 | MODULE_PARM_DESC(edac_op_state, |
| 1214 | "EDAC error reporting state: 0=Poll, 2=Interrupt"); |