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 #include "edac_module.h" 33 34 #define EDAC_MOD_STR "xgene_edac" 35 36 /* Global error configuration status registers (CSR) */ 37 #define PCPHPERRINTSTS 0x0000 38 #define PCPHPERRINTMSK 0x0004 39 #define MCU_CTL_ERR_MASK BIT(12) 40 #define IOB_PA_ERR_MASK BIT(11) 41 #define IOB_BA_ERR_MASK BIT(10) 42 #define IOB_XGIC_ERR_MASK BIT(9) 43 #define IOB_RB_ERR_MASK BIT(8) 44 #define L3C_UNCORR_ERR_MASK BIT(5) 45 #define MCU_UNCORR_ERR_MASK BIT(4) 46 #define PMD3_MERR_MASK BIT(3) 47 #define PMD2_MERR_MASK BIT(2) 48 #define PMD1_MERR_MASK BIT(1) 49 #define PMD0_MERR_MASK BIT(0) 50 #define PCPLPERRINTSTS 0x0008 51 #define PCPLPERRINTMSK 0x000C 52 #define CSW_SWITCH_TRACE_ERR_MASK BIT(2) 53 #define L3C_CORR_ERR_MASK BIT(1) 54 #define MCU_CORR_ERR_MASK BIT(0) 55 #define MEMERRINTSTS 0x0010 56 #define MEMERRINTMSK 0x0014 57 58 struct xgene_edac { 59 struct device *dev; 60 struct regmap *csw_map; 61 struct regmap *mcba_map; 62 struct regmap *mcbb_map; 63 struct regmap *efuse_map; 64 void __iomem *pcp_csr; 65 spinlock_t lock; 66 struct dentry *dfs; 67 68 struct list_head mcus; 69 struct list_head pmds; 70 struct list_head l3s; 71 struct list_head socs; 72 73 struct mutex mc_lock; 74 int mc_active_mask; 75 int mc_registered_mask; 76 }; 77 78 static void xgene_edac_pcp_rd(struct xgene_edac *edac, u32 reg, u32 *val) 79 { 80 *val = readl(edac->pcp_csr + reg); 81 } 82 83 static void xgene_edac_pcp_clrbits(struct xgene_edac *edac, u32 reg, 84 u32 bits_mask) 85 { 86 u32 val; 87 88 spin_lock(&edac->lock); 89 val = readl(edac->pcp_csr + reg); 90 val &= ~bits_mask; 91 writel(val, edac->pcp_csr + reg); 92 spin_unlock(&edac->lock); 93 } 94 95 static void xgene_edac_pcp_setbits(struct xgene_edac *edac, u32 reg, 96 u32 bits_mask) 97 { 98 u32 val; 99 100 spin_lock(&edac->lock); 101 val = readl(edac->pcp_csr + reg); 102 val |= bits_mask; 103 writel(val, edac->pcp_csr + reg); 104 spin_unlock(&edac->lock); 105 } 106 107 /* Memory controller error CSR */ 108 #define MCU_MAX_RANK 8 109 #define MCU_RANK_STRIDE 0x40 110 111 #define MCUGECR 0x0110 112 #define MCU_GECR_DEMANDUCINTREN_MASK BIT(0) 113 #define MCU_GECR_BACKUCINTREN_MASK BIT(1) 114 #define MCU_GECR_CINTREN_MASK BIT(2) 115 #define MUC_GECR_MCUADDRERREN_MASK BIT(9) 116 #define MCUGESR 0x0114 117 #define MCU_GESR_ADDRNOMATCH_ERR_MASK BIT(7) 118 #define MCU_GESR_ADDRMULTIMATCH_ERR_MASK BIT(6) 119 #define MCU_GESR_PHYP_ERR_MASK BIT(3) 120 #define MCUESRR0 0x0314 121 #define MCU_ESRR_MULTUCERR_MASK BIT(3) 122 #define MCU_ESRR_BACKUCERR_MASK BIT(2) 123 #define MCU_ESRR_DEMANDUCERR_MASK BIT(1) 124 #define MCU_ESRR_CERR_MASK BIT(0) 125 #define MCUESRRA0 0x0318 126 #define MCUEBLRR0 0x031c 127 #define MCU_EBLRR_ERRBANK_RD(src) (((src) & 0x00000007) >> 0) 128 #define MCUERCRR0 0x0320 129 #define MCU_ERCRR_ERRROW_RD(src) (((src) & 0xFFFF0000) >> 16) 130 #define MCU_ERCRR_ERRCOL_RD(src) ((src) & 0x00000FFF) 131 #define MCUSBECNT0 0x0324 132 #define MCU_SBECNT_COUNT(src) ((src) & 0xFFFF) 133 134 #define CSW_CSWCR 0x0000 135 #define CSW_CSWCR_DUALMCB_MASK BIT(0) 136 137 #define MCBADDRMR 0x0000 138 #define MCBADDRMR_MCU_INTLV_MODE_MASK BIT(3) 139 #define MCBADDRMR_DUALMCU_MODE_MASK BIT(2) 140 #define MCBADDRMR_MCB_INTLV_MODE_MASK BIT(1) 141 #define MCBADDRMR_ADDRESS_MODE_MASK BIT(0) 142 143 struct xgene_edac_mc_ctx { 144 struct list_head next; 145 char *name; 146 struct mem_ctl_info *mci; 147 struct xgene_edac *edac; 148 void __iomem *mcu_csr; 149 u32 mcu_id; 150 }; 151 152 static ssize_t xgene_edac_mc_err_inject_write(struct file *file, 153 const char __user *data, 154 size_t count, loff_t *ppos) 155 { 156 struct mem_ctl_info *mci = file->private_data; 157 struct xgene_edac_mc_ctx *ctx = mci->pvt_info; 158 int i; 159 160 for (i = 0; i < MCU_MAX_RANK; i++) { 161 writel(MCU_ESRR_MULTUCERR_MASK | MCU_ESRR_BACKUCERR_MASK | 162 MCU_ESRR_DEMANDUCERR_MASK | MCU_ESRR_CERR_MASK, 163 ctx->mcu_csr + MCUESRRA0 + i * MCU_RANK_STRIDE); 164 } 165 return count; 166 } 167 168 static const struct file_operations xgene_edac_mc_debug_inject_fops = { 169 .open = simple_open, 170 .write = xgene_edac_mc_err_inject_write, 171 .llseek = generic_file_llseek, 172 }; 173 174 static void xgene_edac_mc_create_debugfs_node(struct mem_ctl_info *mci) 175 { 176 if (!IS_ENABLED(CONFIG_EDAC_DEBUG)) 177 return; 178 179 if (!mci->debugfs) 180 return; 181 182 edac_debugfs_create_file("inject_ctrl", S_IWUSR, mci->debugfs, mci, 183 &xgene_edac_mc_debug_inject_fops); 184 } 185 186 static void xgene_edac_mc_check(struct mem_ctl_info *mci) 187 { 188 struct xgene_edac_mc_ctx *ctx = mci->pvt_info; 189 unsigned int pcp_hp_stat; 190 unsigned int pcp_lp_stat; 191 u32 reg; 192 u32 rank; 193 u32 bank; 194 u32 count; 195 u32 col_row; 196 197 xgene_edac_pcp_rd(ctx->edac, PCPHPERRINTSTS, &pcp_hp_stat); 198 xgene_edac_pcp_rd(ctx->edac, PCPLPERRINTSTS, &pcp_lp_stat); 199 if (!((MCU_UNCORR_ERR_MASK & pcp_hp_stat) || 200 (MCU_CTL_ERR_MASK & pcp_hp_stat) || 201 (MCU_CORR_ERR_MASK & pcp_lp_stat))) 202 return; 203 204 for (rank = 0; rank < MCU_MAX_RANK; rank++) { 205 reg = readl(ctx->mcu_csr + MCUESRR0 + rank * MCU_RANK_STRIDE); 206 207 /* Detect uncorrectable memory error */ 208 if (reg & (MCU_ESRR_DEMANDUCERR_MASK | 209 MCU_ESRR_BACKUCERR_MASK)) { 210 /* Detected uncorrectable memory error */ 211 edac_mc_chipset_printk(mci, KERN_ERR, "X-Gene", 212 "MCU uncorrectable error at rank %d\n", rank); 213 214 edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 215 1, 0, 0, 0, 0, 0, -1, mci->ctl_name, ""); 216 } 217 218 /* Detect correctable memory error */ 219 if (reg & MCU_ESRR_CERR_MASK) { 220 bank = readl(ctx->mcu_csr + MCUEBLRR0 + 221 rank * MCU_RANK_STRIDE); 222 col_row = readl(ctx->mcu_csr + MCUERCRR0 + 223 rank * MCU_RANK_STRIDE); 224 count = readl(ctx->mcu_csr + MCUSBECNT0 + 225 rank * MCU_RANK_STRIDE); 226 edac_mc_chipset_printk(mci, KERN_WARNING, "X-Gene", 227 "MCU correctable error at rank %d bank %d column %d row %d count %d\n", 228 rank, MCU_EBLRR_ERRBANK_RD(bank), 229 MCU_ERCRR_ERRCOL_RD(col_row), 230 MCU_ERCRR_ERRROW_RD(col_row), 231 MCU_SBECNT_COUNT(count)); 232 233 edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 234 1, 0, 0, 0, 0, 0, -1, mci->ctl_name, ""); 235 } 236 237 /* Clear all error registers */ 238 writel(0x0, ctx->mcu_csr + MCUEBLRR0 + rank * MCU_RANK_STRIDE); 239 writel(0x0, ctx->mcu_csr + MCUERCRR0 + rank * MCU_RANK_STRIDE); 240 writel(0x0, ctx->mcu_csr + MCUSBECNT0 + 241 rank * MCU_RANK_STRIDE); 242 writel(reg, ctx->mcu_csr + MCUESRR0 + rank * MCU_RANK_STRIDE); 243 } 244 245 /* Detect memory controller error */ 246 reg = readl(ctx->mcu_csr + MCUGESR); 247 if (reg) { 248 if (reg & MCU_GESR_ADDRNOMATCH_ERR_MASK) 249 edac_mc_chipset_printk(mci, KERN_WARNING, "X-Gene", 250 "MCU address miss-match error\n"); 251 if (reg & MCU_GESR_ADDRMULTIMATCH_ERR_MASK) 252 edac_mc_chipset_printk(mci, KERN_WARNING, "X-Gene", 253 "MCU address multi-match error\n"); 254 255 writel(reg, ctx->mcu_csr + MCUGESR); 256 } 257 } 258 259 static void xgene_edac_mc_irq_ctl(struct mem_ctl_info *mci, bool enable) 260 { 261 struct xgene_edac_mc_ctx *ctx = mci->pvt_info; 262 unsigned int val; 263 264 if (edac_op_state != EDAC_OPSTATE_INT) 265 return; 266 267 mutex_lock(&ctx->edac->mc_lock); 268 269 /* 270 * As there is only single bit for enable error and interrupt mask, 271 * we must only enable top level interrupt after all MCUs are 272 * registered. Otherwise, if there is an error and the corresponding 273 * MCU has not registered, the interrupt will never get cleared. To 274 * determine all MCU have registered, we will keep track of active 275 * MCUs and registered MCUs. 276 */ 277 if (enable) { 278 /* Set registered MCU bit */ 279 ctx->edac->mc_registered_mask |= 1 << ctx->mcu_id; 280 281 /* Enable interrupt after all active MCU registered */ 282 if (ctx->edac->mc_registered_mask == 283 ctx->edac->mc_active_mask) { 284 /* Enable memory controller top level interrupt */ 285 xgene_edac_pcp_clrbits(ctx->edac, PCPHPERRINTMSK, 286 MCU_UNCORR_ERR_MASK | 287 MCU_CTL_ERR_MASK); 288 xgene_edac_pcp_clrbits(ctx->edac, PCPLPERRINTMSK, 289 MCU_CORR_ERR_MASK); 290 } 291 292 /* Enable MCU interrupt and error reporting */ 293 val = readl(ctx->mcu_csr + MCUGECR); 294 val |= MCU_GECR_DEMANDUCINTREN_MASK | 295 MCU_GECR_BACKUCINTREN_MASK | 296 MCU_GECR_CINTREN_MASK | 297 MUC_GECR_MCUADDRERREN_MASK; 298 writel(val, ctx->mcu_csr + MCUGECR); 299 } else { 300 /* Disable MCU interrupt */ 301 val = readl(ctx->mcu_csr + MCUGECR); 302 val &= ~(MCU_GECR_DEMANDUCINTREN_MASK | 303 MCU_GECR_BACKUCINTREN_MASK | 304 MCU_GECR_CINTREN_MASK | 305 MUC_GECR_MCUADDRERREN_MASK); 306 writel(val, ctx->mcu_csr + MCUGECR); 307 308 /* Disable memory controller top level interrupt */ 309 xgene_edac_pcp_setbits(ctx->edac, PCPHPERRINTMSK, 310 MCU_UNCORR_ERR_MASK | MCU_CTL_ERR_MASK); 311 xgene_edac_pcp_setbits(ctx->edac, PCPLPERRINTMSK, 312 MCU_CORR_ERR_MASK); 313 314 /* Clear registered MCU bit */ 315 ctx->edac->mc_registered_mask &= ~(1 << ctx->mcu_id); 316 } 317 318 mutex_unlock(&ctx->edac->mc_lock); 319 } 320 321 static int xgene_edac_mc_is_active(struct xgene_edac_mc_ctx *ctx, int mc_idx) 322 { 323 unsigned int reg; 324 u32 mcu_mask; 325 326 if (regmap_read(ctx->edac->csw_map, CSW_CSWCR, ®)) 327 return 0; 328 329 if (reg & CSW_CSWCR_DUALMCB_MASK) { 330 /* 331 * Dual MCB active - Determine if all 4 active or just MCU0 332 * and MCU2 active 333 */ 334 if (regmap_read(ctx->edac->mcbb_map, MCBADDRMR, ®)) 335 return 0; 336 mcu_mask = (reg & MCBADDRMR_DUALMCU_MODE_MASK) ? 0xF : 0x5; 337 } else { 338 /* 339 * Single MCB active - Determine if MCU0/MCU1 or just MCU0 340 * active 341 */ 342 if (regmap_read(ctx->edac->mcba_map, MCBADDRMR, ®)) 343 return 0; 344 mcu_mask = (reg & MCBADDRMR_DUALMCU_MODE_MASK) ? 0x3 : 0x1; 345 } 346 347 /* Save active MC mask if hasn't set already */ 348 if (!ctx->edac->mc_active_mask) 349 ctx->edac->mc_active_mask = mcu_mask; 350 351 return (mcu_mask & (1 << mc_idx)) ? 1 : 0; 352 } 353 354 static int xgene_edac_mc_add(struct xgene_edac *edac, struct device_node *np) 355 { 356 struct mem_ctl_info *mci; 357 struct edac_mc_layer layers[2]; 358 struct xgene_edac_mc_ctx tmp_ctx; 359 struct xgene_edac_mc_ctx *ctx; 360 struct resource res; 361 int rc; 362 363 memset(&tmp_ctx, 0, sizeof(tmp_ctx)); 364 tmp_ctx.edac = edac; 365 366 if (!devres_open_group(edac->dev, xgene_edac_mc_add, GFP_KERNEL)) 367 return -ENOMEM; 368 369 rc = of_address_to_resource(np, 0, &res); 370 if (rc < 0) { 371 dev_err(edac->dev, "no MCU resource address\n"); 372 goto err_group; 373 } 374 tmp_ctx.mcu_csr = devm_ioremap_resource(edac->dev, &res); 375 if (IS_ERR(tmp_ctx.mcu_csr)) { 376 dev_err(edac->dev, "unable to map MCU resource\n"); 377 rc = PTR_ERR(tmp_ctx.mcu_csr); 378 goto err_group; 379 } 380 381 /* Ignore non-active MCU */ 382 if (of_property_read_u32(np, "memory-controller", &tmp_ctx.mcu_id)) { 383 dev_err(edac->dev, "no memory-controller property\n"); 384 rc = -ENODEV; 385 goto err_group; 386 } 387 if (!xgene_edac_mc_is_active(&tmp_ctx, tmp_ctx.mcu_id)) { 388 rc = -ENODEV; 389 goto err_group; 390 } 391 392 layers[0].type = EDAC_MC_LAYER_CHIP_SELECT; 393 layers[0].size = 4; 394 layers[0].is_virt_csrow = true; 395 layers[1].type = EDAC_MC_LAYER_CHANNEL; 396 layers[1].size = 2; 397 layers[1].is_virt_csrow = false; 398 mci = edac_mc_alloc(tmp_ctx.mcu_id, ARRAY_SIZE(layers), layers, 399 sizeof(*ctx)); 400 if (!mci) { 401 rc = -ENOMEM; 402 goto err_group; 403 } 404 405 ctx = mci->pvt_info; 406 *ctx = tmp_ctx; /* Copy over resource value */ 407 ctx->name = "xgene_edac_mc_err"; 408 ctx->mci = mci; 409 mci->pdev = &mci->dev; 410 mci->ctl_name = ctx->name; 411 mci->dev_name = ctx->name; 412 413 mci->mtype_cap = MEM_FLAG_RDDR | MEM_FLAG_RDDR2 | MEM_FLAG_RDDR3 | 414 MEM_FLAG_DDR | MEM_FLAG_DDR2 | MEM_FLAG_DDR3; 415 mci->edac_ctl_cap = EDAC_FLAG_SECDED; 416 mci->edac_cap = EDAC_FLAG_SECDED; 417 mci->mod_name = EDAC_MOD_STR; 418 mci->mod_ver = "0.1"; 419 mci->ctl_page_to_phys = NULL; 420 mci->scrub_cap = SCRUB_FLAG_HW_SRC; 421 mci->scrub_mode = SCRUB_HW_SRC; 422 423 if (edac_op_state == EDAC_OPSTATE_POLL) 424 mci->edac_check = xgene_edac_mc_check; 425 426 if (edac_mc_add_mc(mci)) { 427 dev_err(edac->dev, "edac_mc_add_mc failed\n"); 428 rc = -EINVAL; 429 goto err_free; 430 } 431 432 xgene_edac_mc_create_debugfs_node(mci); 433 434 list_add(&ctx->next, &edac->mcus); 435 436 xgene_edac_mc_irq_ctl(mci, true); 437 438 devres_remove_group(edac->dev, xgene_edac_mc_add); 439 440 dev_info(edac->dev, "X-Gene EDAC MC registered\n"); 441 return 0; 442 443 err_free: 444 edac_mc_free(mci); 445 err_group: 446 devres_release_group(edac->dev, xgene_edac_mc_add); 447 return rc; 448 } 449 450 static int xgene_edac_mc_remove(struct xgene_edac_mc_ctx *mcu) 451 { 452 xgene_edac_mc_irq_ctl(mcu->mci, false); 453 edac_mc_del_mc(&mcu->mci->dev); 454 edac_mc_free(mcu->mci); 455 return 0; 456 } 457 458 /* CPU L1/L2 error CSR */ 459 #define MAX_CPU_PER_PMD 2 460 #define CPU_CSR_STRIDE 0x00100000 461 #define CPU_L2C_PAGE 0x000D0000 462 #define CPU_MEMERR_L2C_PAGE 0x000E0000 463 #define CPU_MEMERR_CPU_PAGE 0x000F0000 464 465 #define MEMERR_CPU_ICFECR_PAGE_OFFSET 0x0000 466 #define MEMERR_CPU_ICFESR_PAGE_OFFSET 0x0004 467 #define MEMERR_CPU_ICFESR_ERRWAY_RD(src) (((src) & 0xFF000000) >> 24) 468 #define MEMERR_CPU_ICFESR_ERRINDEX_RD(src) (((src) & 0x003F0000) >> 16) 469 #define MEMERR_CPU_ICFESR_ERRINFO_RD(src) (((src) & 0x0000FF00) >> 8) 470 #define MEMERR_CPU_ICFESR_ERRTYPE_RD(src) (((src) & 0x00000070) >> 4) 471 #define MEMERR_CPU_ICFESR_MULTCERR_MASK BIT(2) 472 #define MEMERR_CPU_ICFESR_CERR_MASK BIT(0) 473 #define MEMERR_CPU_LSUESR_PAGE_OFFSET 0x000c 474 #define MEMERR_CPU_LSUESR_ERRWAY_RD(src) (((src) & 0xFF000000) >> 24) 475 #define MEMERR_CPU_LSUESR_ERRINDEX_RD(src) (((src) & 0x003F0000) >> 16) 476 #define MEMERR_CPU_LSUESR_ERRINFO_RD(src) (((src) & 0x0000FF00) >> 8) 477 #define MEMERR_CPU_LSUESR_ERRTYPE_RD(src) (((src) & 0x00000070) >> 4) 478 #define MEMERR_CPU_LSUESR_MULTCERR_MASK BIT(2) 479 #define MEMERR_CPU_LSUESR_CERR_MASK BIT(0) 480 #define MEMERR_CPU_LSUECR_PAGE_OFFSET 0x0008 481 #define MEMERR_CPU_MMUECR_PAGE_OFFSET 0x0010 482 #define MEMERR_CPU_MMUESR_PAGE_OFFSET 0x0014 483 #define MEMERR_CPU_MMUESR_ERRWAY_RD(src) (((src) & 0xFF000000) >> 24) 484 #define MEMERR_CPU_MMUESR_ERRINDEX_RD(src) (((src) & 0x007F0000) >> 16) 485 #define MEMERR_CPU_MMUESR_ERRINFO_RD(src) (((src) & 0x0000FF00) >> 8) 486 #define MEMERR_CPU_MMUESR_ERRREQSTR_LSU_MASK BIT(7) 487 #define MEMERR_CPU_MMUESR_ERRTYPE_RD(src) (((src) & 0x00000070) >> 4) 488 #define MEMERR_CPU_MMUESR_MULTCERR_MASK BIT(2) 489 #define MEMERR_CPU_MMUESR_CERR_MASK BIT(0) 490 #define MEMERR_CPU_ICFESRA_PAGE_OFFSET 0x0804 491 #define MEMERR_CPU_LSUESRA_PAGE_OFFSET 0x080c 492 #define MEMERR_CPU_MMUESRA_PAGE_OFFSET 0x0814 493 494 #define MEMERR_L2C_L2ECR_PAGE_OFFSET 0x0000 495 #define MEMERR_L2C_L2ESR_PAGE_OFFSET 0x0004 496 #define MEMERR_L2C_L2ESR_ERRSYN_RD(src) (((src) & 0xFF000000) >> 24) 497 #define MEMERR_L2C_L2ESR_ERRWAY_RD(src) (((src) & 0x00FC0000) >> 18) 498 #define MEMERR_L2C_L2ESR_ERRCPU_RD(src) (((src) & 0x00020000) >> 17) 499 #define MEMERR_L2C_L2ESR_ERRGROUP_RD(src) (((src) & 0x0000E000) >> 13) 500 #define MEMERR_L2C_L2ESR_ERRACTION_RD(src) (((src) & 0x00001C00) >> 10) 501 #define MEMERR_L2C_L2ESR_ERRTYPE_RD(src) (((src) & 0x00000300) >> 8) 502 #define MEMERR_L2C_L2ESR_MULTUCERR_MASK BIT(3) 503 #define MEMERR_L2C_L2ESR_MULTICERR_MASK BIT(2) 504 #define MEMERR_L2C_L2ESR_UCERR_MASK BIT(1) 505 #define MEMERR_L2C_L2ESR_ERR_MASK BIT(0) 506 #define MEMERR_L2C_L2EALR_PAGE_OFFSET 0x0008 507 #define CPUX_L2C_L2RTOCR_PAGE_OFFSET 0x0010 508 #define MEMERR_L2C_L2EAHR_PAGE_OFFSET 0x000c 509 #define CPUX_L2C_L2RTOSR_PAGE_OFFSET 0x0014 510 #define MEMERR_L2C_L2RTOSR_MULTERR_MASK BIT(1) 511 #define MEMERR_L2C_L2RTOSR_ERR_MASK BIT(0) 512 #define CPUX_L2C_L2RTOALR_PAGE_OFFSET 0x0018 513 #define CPUX_L2C_L2RTOAHR_PAGE_OFFSET 0x001c 514 #define MEMERR_L2C_L2ESRA_PAGE_OFFSET 0x0804 515 516 /* 517 * Processor Module Domain (PMD) context - Context for a pair of processsors. 518 * Each PMD consists of 2 CPUs and a shared L2 cache. Each CPU consists of 519 * its own L1 cache. 520 */ 521 struct xgene_edac_pmd_ctx { 522 struct list_head next; 523 struct device ddev; 524 char *name; 525 struct xgene_edac *edac; 526 struct edac_device_ctl_info *edac_dev; 527 void __iomem *pmd_csr; 528 u32 pmd; 529 int version; 530 }; 531 532 static void xgene_edac_pmd_l1_check(struct edac_device_ctl_info *edac_dev, 533 int cpu_idx) 534 { 535 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; 536 void __iomem *pg_f; 537 u32 val; 538 539 pg_f = ctx->pmd_csr + cpu_idx * CPU_CSR_STRIDE + CPU_MEMERR_CPU_PAGE; 540 541 val = readl(pg_f + MEMERR_CPU_ICFESR_PAGE_OFFSET); 542 if (!val) 543 goto chk_lsu; 544 dev_err(edac_dev->dev, 545 "CPU%d L1 memory error ICF 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X\n", 546 ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val, 547 MEMERR_CPU_ICFESR_ERRWAY_RD(val), 548 MEMERR_CPU_ICFESR_ERRINDEX_RD(val), 549 MEMERR_CPU_ICFESR_ERRINFO_RD(val)); 550 if (val & MEMERR_CPU_ICFESR_CERR_MASK) 551 dev_err(edac_dev->dev, "One or more correctable error\n"); 552 if (val & MEMERR_CPU_ICFESR_MULTCERR_MASK) 553 dev_err(edac_dev->dev, "Multiple correctable error\n"); 554 switch (MEMERR_CPU_ICFESR_ERRTYPE_RD(val)) { 555 case 1: 556 dev_err(edac_dev->dev, "L1 TLB multiple hit\n"); 557 break; 558 case 2: 559 dev_err(edac_dev->dev, "Way select multiple hit\n"); 560 break; 561 case 3: 562 dev_err(edac_dev->dev, "Physical tag parity error\n"); 563 break; 564 case 4: 565 case 5: 566 dev_err(edac_dev->dev, "L1 data parity error\n"); 567 break; 568 case 6: 569 dev_err(edac_dev->dev, "L1 pre-decode parity error\n"); 570 break; 571 } 572 573 /* Clear any HW errors */ 574 writel(val, pg_f + MEMERR_CPU_ICFESR_PAGE_OFFSET); 575 576 if (val & (MEMERR_CPU_ICFESR_CERR_MASK | 577 MEMERR_CPU_ICFESR_MULTCERR_MASK)) 578 edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name); 579 580 chk_lsu: 581 val = readl(pg_f + MEMERR_CPU_LSUESR_PAGE_OFFSET); 582 if (!val) 583 goto chk_mmu; 584 dev_err(edac_dev->dev, 585 "CPU%d memory error LSU 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X\n", 586 ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val, 587 MEMERR_CPU_LSUESR_ERRWAY_RD(val), 588 MEMERR_CPU_LSUESR_ERRINDEX_RD(val), 589 MEMERR_CPU_LSUESR_ERRINFO_RD(val)); 590 if (val & MEMERR_CPU_LSUESR_CERR_MASK) 591 dev_err(edac_dev->dev, "One or more correctable error\n"); 592 if (val & MEMERR_CPU_LSUESR_MULTCERR_MASK) 593 dev_err(edac_dev->dev, "Multiple correctable error\n"); 594 switch (MEMERR_CPU_LSUESR_ERRTYPE_RD(val)) { 595 case 0: 596 dev_err(edac_dev->dev, "Load tag error\n"); 597 break; 598 case 1: 599 dev_err(edac_dev->dev, "Load data error\n"); 600 break; 601 case 2: 602 dev_err(edac_dev->dev, "WSL multihit error\n"); 603 break; 604 case 3: 605 dev_err(edac_dev->dev, "Store tag error\n"); 606 break; 607 case 4: 608 dev_err(edac_dev->dev, 609 "DTB multihit from load pipeline error\n"); 610 break; 611 case 5: 612 dev_err(edac_dev->dev, 613 "DTB multihit from store pipeline error\n"); 614 break; 615 } 616 617 /* Clear any HW errors */ 618 writel(val, pg_f + MEMERR_CPU_LSUESR_PAGE_OFFSET); 619 620 if (val & (MEMERR_CPU_LSUESR_CERR_MASK | 621 MEMERR_CPU_LSUESR_MULTCERR_MASK)) 622 edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name); 623 624 chk_mmu: 625 val = readl(pg_f + MEMERR_CPU_MMUESR_PAGE_OFFSET); 626 if (!val) 627 return; 628 dev_err(edac_dev->dev, 629 "CPU%d memory error MMU 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X %s\n", 630 ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val, 631 MEMERR_CPU_MMUESR_ERRWAY_RD(val), 632 MEMERR_CPU_MMUESR_ERRINDEX_RD(val), 633 MEMERR_CPU_MMUESR_ERRINFO_RD(val), 634 val & MEMERR_CPU_MMUESR_ERRREQSTR_LSU_MASK ? "LSU" : "ICF"); 635 if (val & MEMERR_CPU_MMUESR_CERR_MASK) 636 dev_err(edac_dev->dev, "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, "TMO operation single bank error\n"); 651 break; 652 case 4: 653 dev_err(edac_dev->dev, "Stage 2 UTB error\n"); 654 break; 655 case 5: 656 dev_err(edac_dev->dev, "Stage 2 UTB miss error\n"); 657 break; 658 case 6: 659 dev_err(edac_dev->dev, "Stage 2 UTB allocate error\n"); 660 break; 661 case 7: 662 dev_err(edac_dev->dev, "TMO operation multiple bank error\n"); 663 break; 664 } 665 666 /* Clear any HW errors */ 667 writel(val, pg_f + MEMERR_CPU_MMUESR_PAGE_OFFSET); 668 669 edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name); 670 } 671 672 static void xgene_edac_pmd_l2_check(struct edac_device_ctl_info *edac_dev) 673 { 674 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; 675 void __iomem *pg_d; 676 void __iomem *pg_e; 677 u32 val_hi; 678 u32 val_lo; 679 u32 val; 680 681 /* Check L2 */ 682 pg_e = ctx->pmd_csr + CPU_MEMERR_L2C_PAGE; 683 val = readl(pg_e + MEMERR_L2C_L2ESR_PAGE_OFFSET); 684 if (!val) 685 goto chk_l2c; 686 val_lo = readl(pg_e + MEMERR_L2C_L2EALR_PAGE_OFFSET); 687 val_hi = readl(pg_e + MEMERR_L2C_L2EAHR_PAGE_OFFSET); 688 dev_err(edac_dev->dev, 689 "PMD%d memory error L2C L2ESR 0x%08X @ 0x%08X.%08X\n", 690 ctx->pmd, val, val_hi, val_lo); 691 dev_err(edac_dev->dev, 692 "ErrSyndrome 0x%02X ErrWay 0x%02X ErrCpu %d ErrGroup 0x%02X ErrAction 0x%02X\n", 693 MEMERR_L2C_L2ESR_ERRSYN_RD(val), 694 MEMERR_L2C_L2ESR_ERRWAY_RD(val), 695 MEMERR_L2C_L2ESR_ERRCPU_RD(val), 696 MEMERR_L2C_L2ESR_ERRGROUP_RD(val), 697 MEMERR_L2C_L2ESR_ERRACTION_RD(val)); 698 699 if (val & MEMERR_L2C_L2ESR_ERR_MASK) 700 dev_err(edac_dev->dev, "One or more correctable error\n"); 701 if (val & MEMERR_L2C_L2ESR_MULTICERR_MASK) 702 dev_err(edac_dev->dev, "Multiple correctable error\n"); 703 if (val & MEMERR_L2C_L2ESR_UCERR_MASK) 704 dev_err(edac_dev->dev, "One or more uncorrectable error\n"); 705 if (val & MEMERR_L2C_L2ESR_MULTUCERR_MASK) 706 dev_err(edac_dev->dev, "Multiple uncorrectable error\n"); 707 708 switch (MEMERR_L2C_L2ESR_ERRTYPE_RD(val)) { 709 case 0: 710 dev_err(edac_dev->dev, "Outbound SDB parity error\n"); 711 break; 712 case 1: 713 dev_err(edac_dev->dev, "Inbound SDB parity error\n"); 714 break; 715 case 2: 716 dev_err(edac_dev->dev, "Tag ECC error\n"); 717 break; 718 case 3: 719 dev_err(edac_dev->dev, "Data ECC error\n"); 720 break; 721 } 722 723 /* Clear any HW errors */ 724 writel(val, pg_e + MEMERR_L2C_L2ESR_PAGE_OFFSET); 725 726 if (val & (MEMERR_L2C_L2ESR_ERR_MASK | 727 MEMERR_L2C_L2ESR_MULTICERR_MASK)) 728 edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name); 729 if (val & (MEMERR_L2C_L2ESR_UCERR_MASK | 730 MEMERR_L2C_L2ESR_MULTUCERR_MASK)) 731 edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name); 732 733 chk_l2c: 734 /* Check if any memory request timed out on L2 cache */ 735 pg_d = ctx->pmd_csr + CPU_L2C_PAGE; 736 val = readl(pg_d + CPUX_L2C_L2RTOSR_PAGE_OFFSET); 737 if (val) { 738 val_lo = readl(pg_d + CPUX_L2C_L2RTOALR_PAGE_OFFSET); 739 val_hi = readl(pg_d + CPUX_L2C_L2RTOAHR_PAGE_OFFSET); 740 dev_err(edac_dev->dev, 741 "PMD%d L2C error L2C RTOSR 0x%08X @ 0x%08X.%08X\n", 742 ctx->pmd, val, val_hi, val_lo); 743 writel(val, pg_d + CPUX_L2C_L2RTOSR_PAGE_OFFSET); 744 } 745 } 746 747 static void xgene_edac_pmd_check(struct edac_device_ctl_info *edac_dev) 748 { 749 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; 750 unsigned int pcp_hp_stat; 751 int i; 752 753 xgene_edac_pcp_rd(ctx->edac, PCPHPERRINTSTS, &pcp_hp_stat); 754 if (!((PMD0_MERR_MASK << ctx->pmd) & pcp_hp_stat)) 755 return; 756 757 /* Check CPU L1 error */ 758 for (i = 0; i < MAX_CPU_PER_PMD; i++) 759 xgene_edac_pmd_l1_check(edac_dev, i); 760 761 /* Check CPU L2 error */ 762 xgene_edac_pmd_l2_check(edac_dev); 763 } 764 765 static void xgene_edac_pmd_cpu_hw_cfg(struct edac_device_ctl_info *edac_dev, 766 int cpu) 767 { 768 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; 769 void __iomem *pg_f = ctx->pmd_csr + cpu * CPU_CSR_STRIDE + 770 CPU_MEMERR_CPU_PAGE; 771 772 /* 773 * Enable CPU memory error: 774 * MEMERR_CPU_ICFESRA, MEMERR_CPU_LSUESRA, and MEMERR_CPU_MMUESRA 775 */ 776 writel(0x00000301, pg_f + MEMERR_CPU_ICFECR_PAGE_OFFSET); 777 writel(0x00000301, pg_f + MEMERR_CPU_LSUECR_PAGE_OFFSET); 778 writel(0x00000101, pg_f + MEMERR_CPU_MMUECR_PAGE_OFFSET); 779 } 780 781 static void xgene_edac_pmd_hw_cfg(struct edac_device_ctl_info *edac_dev) 782 { 783 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; 784 void __iomem *pg_d = ctx->pmd_csr + CPU_L2C_PAGE; 785 void __iomem *pg_e = ctx->pmd_csr + CPU_MEMERR_L2C_PAGE; 786 787 /* Enable PMD memory error - MEMERR_L2C_L2ECR and L2C_L2RTOCR */ 788 writel(0x00000703, pg_e + MEMERR_L2C_L2ECR_PAGE_OFFSET); 789 /* Configure L2C HW request time out feature if supported */ 790 if (ctx->version > 1) 791 writel(0x00000119, pg_d + CPUX_L2C_L2RTOCR_PAGE_OFFSET); 792 } 793 794 static void xgene_edac_pmd_hw_ctl(struct edac_device_ctl_info *edac_dev, 795 bool enable) 796 { 797 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; 798 int i; 799 800 /* Enable PMD error interrupt */ 801 if (edac_dev->op_state == OP_RUNNING_INTERRUPT) { 802 if (enable) 803 xgene_edac_pcp_clrbits(ctx->edac, PCPHPERRINTMSK, 804 PMD0_MERR_MASK << ctx->pmd); 805 else 806 xgene_edac_pcp_setbits(ctx->edac, PCPHPERRINTMSK, 807 PMD0_MERR_MASK << ctx->pmd); 808 } 809 810 if (enable) { 811 xgene_edac_pmd_hw_cfg(edac_dev); 812 813 /* Two CPUs per a PMD */ 814 for (i = 0; i < MAX_CPU_PER_PMD; i++) 815 xgene_edac_pmd_cpu_hw_cfg(edac_dev, i); 816 } 817 } 818 819 static ssize_t xgene_edac_pmd_l1_inject_ctrl_write(struct file *file, 820 const char __user *data, 821 size_t count, loff_t *ppos) 822 { 823 struct edac_device_ctl_info *edac_dev = file->private_data; 824 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; 825 void __iomem *cpux_pg_f; 826 int i; 827 828 for (i = 0; i < MAX_CPU_PER_PMD; i++) { 829 cpux_pg_f = ctx->pmd_csr + i * CPU_CSR_STRIDE + 830 CPU_MEMERR_CPU_PAGE; 831 832 writel(MEMERR_CPU_ICFESR_MULTCERR_MASK | 833 MEMERR_CPU_ICFESR_CERR_MASK, 834 cpux_pg_f + MEMERR_CPU_ICFESRA_PAGE_OFFSET); 835 writel(MEMERR_CPU_LSUESR_MULTCERR_MASK | 836 MEMERR_CPU_LSUESR_CERR_MASK, 837 cpux_pg_f + MEMERR_CPU_LSUESRA_PAGE_OFFSET); 838 writel(MEMERR_CPU_MMUESR_MULTCERR_MASK | 839 MEMERR_CPU_MMUESR_CERR_MASK, 840 cpux_pg_f + MEMERR_CPU_MMUESRA_PAGE_OFFSET); 841 } 842 return count; 843 } 844 845 static ssize_t xgene_edac_pmd_l2_inject_ctrl_write(struct file *file, 846 const char __user *data, 847 size_t count, loff_t *ppos) 848 { 849 struct edac_device_ctl_info *edac_dev = file->private_data; 850 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; 851 void __iomem *pg_e = ctx->pmd_csr + CPU_MEMERR_L2C_PAGE; 852 853 writel(MEMERR_L2C_L2ESR_MULTUCERR_MASK | 854 MEMERR_L2C_L2ESR_MULTICERR_MASK | 855 MEMERR_L2C_L2ESR_UCERR_MASK | 856 MEMERR_L2C_L2ESR_ERR_MASK, 857 pg_e + MEMERR_L2C_L2ESRA_PAGE_OFFSET); 858 return count; 859 } 860 861 static const struct file_operations xgene_edac_pmd_debug_inject_fops[] = { 862 { 863 .open = simple_open, 864 .write = xgene_edac_pmd_l1_inject_ctrl_write, 865 .llseek = generic_file_llseek, }, 866 { 867 .open = simple_open, 868 .write = xgene_edac_pmd_l2_inject_ctrl_write, 869 .llseek = generic_file_llseek, }, 870 { } 871 }; 872 873 static void 874 xgene_edac_pmd_create_debugfs_nodes(struct edac_device_ctl_info *edac_dev) 875 { 876 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info; 877 struct dentry *dbgfs_dir; 878 char name[10]; 879 880 if (!IS_ENABLED(CONFIG_EDAC_DEBUG) || !ctx->edac->dfs) 881 return; 882 883 snprintf(name, sizeof(name), "PMD%d", ctx->pmd); 884 dbgfs_dir = edac_debugfs_create_dir_at(name, ctx->edac->dfs); 885 if (!dbgfs_dir) 886 return; 887 888 edac_debugfs_create_file("l1_inject_ctrl", S_IWUSR, dbgfs_dir, edac_dev, 889 &xgene_edac_pmd_debug_inject_fops[0]); 890 edac_debugfs_create_file("l2_inject_ctrl", S_IWUSR, dbgfs_dir, edac_dev, 891 &xgene_edac_pmd_debug_inject_fops[1]); 892 } 893 894 static int xgene_edac_pmd_available(u32 efuse, int pmd) 895 { 896 return (efuse & (1 << pmd)) ? 0 : 1; 897 } 898 899 static int xgene_edac_pmd_add(struct xgene_edac *edac, struct device_node *np, 900 int version) 901 { 902 struct edac_device_ctl_info *edac_dev; 903 struct xgene_edac_pmd_ctx *ctx; 904 struct resource res; 905 char edac_name[10]; 906 u32 pmd; 907 int rc; 908 u32 val; 909 910 if (!devres_open_group(edac->dev, xgene_edac_pmd_add, GFP_KERNEL)) 911 return -ENOMEM; 912 913 /* Determine if this PMD is disabled */ 914 if (of_property_read_u32(np, "pmd-controller", &pmd)) { 915 dev_err(edac->dev, "no pmd-controller property\n"); 916 rc = -ENODEV; 917 goto err_group; 918 } 919 rc = regmap_read(edac->efuse_map, 0, &val); 920 if (rc) 921 goto err_group; 922 if (!xgene_edac_pmd_available(val, pmd)) { 923 rc = -ENODEV; 924 goto err_group; 925 } 926 927 snprintf(edac_name, sizeof(edac_name), "l2c%d", pmd); 928 edac_dev = edac_device_alloc_ctl_info(sizeof(*ctx), 929 edac_name, 1, "l2c", 1, 2, NULL, 930 0, edac_device_alloc_index()); 931 if (!edac_dev) { 932 rc = -ENOMEM; 933 goto err_group; 934 } 935 936 ctx = edac_dev->pvt_info; 937 ctx->name = "xgene_pmd_err"; 938 ctx->pmd = pmd; 939 ctx->edac = edac; 940 ctx->edac_dev = edac_dev; 941 ctx->ddev = *edac->dev; 942 ctx->version = version; 943 edac_dev->dev = &ctx->ddev; 944 edac_dev->ctl_name = ctx->name; 945 edac_dev->dev_name = ctx->name; 946 edac_dev->mod_name = EDAC_MOD_STR; 947 948 rc = of_address_to_resource(np, 0, &res); 949 if (rc < 0) { 950 dev_err(edac->dev, "no PMD resource address\n"); 951 goto err_free; 952 } 953 ctx->pmd_csr = devm_ioremap_resource(edac->dev, &res); 954 if (IS_ERR(ctx->pmd_csr)) { 955 dev_err(edac->dev, 956 "devm_ioremap_resource failed for PMD resource address\n"); 957 rc = PTR_ERR(ctx->pmd_csr); 958 goto err_free; 959 } 960 961 if (edac_op_state == EDAC_OPSTATE_POLL) 962 edac_dev->edac_check = xgene_edac_pmd_check; 963 964 xgene_edac_pmd_create_debugfs_nodes(edac_dev); 965 966 rc = edac_device_add_device(edac_dev); 967 if (rc > 0) { 968 dev_err(edac->dev, "edac_device_add_device failed\n"); 969 rc = -ENOMEM; 970 goto err_free; 971 } 972 973 if (edac_op_state == EDAC_OPSTATE_INT) 974 edac_dev->op_state = OP_RUNNING_INTERRUPT; 975 976 list_add(&ctx->next, &edac->pmds); 977 978 xgene_edac_pmd_hw_ctl(edac_dev, 1); 979 980 devres_remove_group(edac->dev, xgene_edac_pmd_add); 981 982 dev_info(edac->dev, "X-Gene EDAC PMD%d registered\n", ctx->pmd); 983 return 0; 984 985 err_free: 986 edac_device_free_ctl_info(edac_dev); 987 err_group: 988 devres_release_group(edac->dev, xgene_edac_pmd_add); 989 return rc; 990 } 991 992 static int xgene_edac_pmd_remove(struct xgene_edac_pmd_ctx *pmd) 993 { 994 struct edac_device_ctl_info *edac_dev = pmd->edac_dev; 995 996 xgene_edac_pmd_hw_ctl(edac_dev, 0); 997 edac_device_del_device(edac_dev->dev); 998 edac_device_free_ctl_info(edac_dev); 999 return 0; 1000 } 1001 1002 /* L3 Error device */ 1003 #define L3C_ESR (0x0A * 4) 1004 #define L3C_ESR_DATATAG_MASK BIT(9) 1005 #define L3C_ESR_MULTIHIT_MASK BIT(8) 1006 #define L3C_ESR_UCEVICT_MASK BIT(6) 1007 #define L3C_ESR_MULTIUCERR_MASK BIT(5) 1008 #define L3C_ESR_MULTICERR_MASK BIT(4) 1009 #define L3C_ESR_UCERR_MASK BIT(3) 1010 #define L3C_ESR_CERR_MASK BIT(2) 1011 #define L3C_ESR_UCERRINTR_MASK BIT(1) 1012 #define L3C_ESR_CERRINTR_MASK BIT(0) 1013 #define L3C_ECR (0x0B * 4) 1014 #define L3C_ECR_UCINTREN BIT(3) 1015 #define L3C_ECR_CINTREN BIT(2) 1016 #define L3C_UCERREN BIT(1) 1017 #define L3C_CERREN BIT(0) 1018 #define L3C_ELR (0x0C * 4) 1019 #define L3C_ELR_ERRSYN(src) ((src & 0xFF800000) >> 23) 1020 #define L3C_ELR_ERRWAY(src) ((src & 0x007E0000) >> 17) 1021 #define L3C_ELR_AGENTID(src) ((src & 0x0001E000) >> 13) 1022 #define L3C_ELR_ERRGRP(src) ((src & 0x00000F00) >> 8) 1023 #define L3C_ELR_OPTYPE(src) ((src & 0x000000F0) >> 4) 1024 #define L3C_ELR_PADDRHIGH(src) (src & 0x0000000F) 1025 #define L3C_AELR (0x0D * 4) 1026 #define L3C_BELR (0x0E * 4) 1027 #define L3C_BELR_BANK(src) (src & 0x0000000F) 1028 1029 struct xgene_edac_dev_ctx { 1030 struct list_head next; 1031 struct device ddev; 1032 char *name; 1033 struct xgene_edac *edac; 1034 struct edac_device_ctl_info *edac_dev; 1035 int edac_idx; 1036 void __iomem *dev_csr; 1037 int version; 1038 }; 1039 1040 /* 1041 * Version 1 of the L3 controller has broken single bit correctable logic for 1042 * certain error syndromes. Log them as uncorrectable in that case. 1043 */ 1044 static bool xgene_edac_l3_promote_to_uc_err(u32 l3cesr, u32 l3celr) 1045 { 1046 if (l3cesr & L3C_ESR_DATATAG_MASK) { 1047 switch (L3C_ELR_ERRSYN(l3celr)) { 1048 case 0x13C: 1049 case 0x0B4: 1050 case 0x007: 1051 case 0x00D: 1052 case 0x00E: 1053 case 0x019: 1054 case 0x01A: 1055 case 0x01C: 1056 case 0x04E: 1057 case 0x041: 1058 return true; 1059 } 1060 } else if (L3C_ELR_ERRSYN(l3celr) == 9) 1061 return true; 1062 1063 return false; 1064 } 1065 1066 static void xgene_edac_l3_check(struct edac_device_ctl_info *edac_dev) 1067 { 1068 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info; 1069 u32 l3cesr; 1070 u32 l3celr; 1071 u32 l3caelr; 1072 u32 l3cbelr; 1073 1074 l3cesr = readl(ctx->dev_csr + L3C_ESR); 1075 if (!(l3cesr & (L3C_ESR_UCERR_MASK | L3C_ESR_CERR_MASK))) 1076 return; 1077 1078 if (l3cesr & L3C_ESR_UCERR_MASK) 1079 dev_err(edac_dev->dev, "L3C uncorrectable error\n"); 1080 if (l3cesr & L3C_ESR_CERR_MASK) 1081 dev_warn(edac_dev->dev, "L3C correctable error\n"); 1082 1083 l3celr = readl(ctx->dev_csr + L3C_ELR); 1084 l3caelr = readl(ctx->dev_csr + L3C_AELR); 1085 l3cbelr = readl(ctx->dev_csr + L3C_BELR); 1086 if (l3cesr & L3C_ESR_MULTIHIT_MASK) 1087 dev_err(edac_dev->dev, "L3C multiple hit error\n"); 1088 if (l3cesr & L3C_ESR_UCEVICT_MASK) 1089 dev_err(edac_dev->dev, 1090 "L3C dropped eviction of line with error\n"); 1091 if (l3cesr & L3C_ESR_MULTIUCERR_MASK) 1092 dev_err(edac_dev->dev, "L3C multiple uncorrectable error\n"); 1093 if (l3cesr & L3C_ESR_DATATAG_MASK) 1094 dev_err(edac_dev->dev, 1095 "L3C data error syndrome 0x%X group 0x%X\n", 1096 L3C_ELR_ERRSYN(l3celr), L3C_ELR_ERRGRP(l3celr)); 1097 else 1098 dev_err(edac_dev->dev, 1099 "L3C tag error syndrome 0x%X Way of Tag 0x%X Agent ID 0x%X Operation type 0x%X\n", 1100 L3C_ELR_ERRSYN(l3celr), L3C_ELR_ERRWAY(l3celr), 1101 L3C_ELR_AGENTID(l3celr), L3C_ELR_OPTYPE(l3celr)); 1102 /* 1103 * NOTE: Address [41:38] in L3C_ELR_PADDRHIGH(l3celr). 1104 * Address [37:6] in l3caelr. Lower 6 bits are zero. 1105 */ 1106 dev_err(edac_dev->dev, "L3C error address 0x%08X.%08X bank %d\n", 1107 L3C_ELR_PADDRHIGH(l3celr) << 6 | (l3caelr >> 26), 1108 (l3caelr & 0x3FFFFFFF) << 6, L3C_BELR_BANK(l3cbelr)); 1109 dev_err(edac_dev->dev, 1110 "L3C error status register value 0x%X\n", l3cesr); 1111 1112 /* Clear L3C error interrupt */ 1113 writel(0, ctx->dev_csr + L3C_ESR); 1114 1115 if (ctx->version <= 1 && 1116 xgene_edac_l3_promote_to_uc_err(l3cesr, l3celr)) { 1117 edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name); 1118 return; 1119 } 1120 if (l3cesr & L3C_ESR_CERR_MASK) 1121 edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name); 1122 if (l3cesr & L3C_ESR_UCERR_MASK) 1123 edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name); 1124 } 1125 1126 static void xgene_edac_l3_hw_init(struct edac_device_ctl_info *edac_dev, 1127 bool enable) 1128 { 1129 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info; 1130 u32 val; 1131 1132 val = readl(ctx->dev_csr + L3C_ECR); 1133 val |= L3C_UCERREN | L3C_CERREN; 1134 /* On disable, we just disable interrupt but keep error enabled */ 1135 if (edac_dev->op_state == OP_RUNNING_INTERRUPT) { 1136 if (enable) 1137 val |= L3C_ECR_UCINTREN | L3C_ECR_CINTREN; 1138 else 1139 val &= ~(L3C_ECR_UCINTREN | L3C_ECR_CINTREN); 1140 } 1141 writel(val, ctx->dev_csr + L3C_ECR); 1142 1143 if (edac_dev->op_state == OP_RUNNING_INTERRUPT) { 1144 /* Enable/disable L3 error top level interrupt */ 1145 if (enable) { 1146 xgene_edac_pcp_clrbits(ctx->edac, PCPHPERRINTMSK, 1147 L3C_UNCORR_ERR_MASK); 1148 xgene_edac_pcp_clrbits(ctx->edac, PCPLPERRINTMSK, 1149 L3C_CORR_ERR_MASK); 1150 } else { 1151 xgene_edac_pcp_setbits(ctx->edac, PCPHPERRINTMSK, 1152 L3C_UNCORR_ERR_MASK); 1153 xgene_edac_pcp_setbits(ctx->edac, PCPLPERRINTMSK, 1154 L3C_CORR_ERR_MASK); 1155 } 1156 } 1157 } 1158 1159 static ssize_t xgene_edac_l3_inject_ctrl_write(struct file *file, 1160 const char __user *data, 1161 size_t count, loff_t *ppos) 1162 { 1163 struct edac_device_ctl_info *edac_dev = file->private_data; 1164 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info; 1165 1166 /* Generate all errors */ 1167 writel(0xFFFFFFFF, ctx->dev_csr + L3C_ESR); 1168 return count; 1169 } 1170 1171 static const struct file_operations xgene_edac_l3_debug_inject_fops = { 1172 .open = simple_open, 1173 .write = xgene_edac_l3_inject_ctrl_write, 1174 .llseek = generic_file_llseek 1175 }; 1176 1177 static void 1178 xgene_edac_l3_create_debugfs_nodes(struct edac_device_ctl_info *edac_dev) 1179 { 1180 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info; 1181 struct dentry *dbgfs_dir; 1182 char name[10]; 1183 1184 if (!IS_ENABLED(CONFIG_EDAC_DEBUG) || !ctx->edac->dfs) 1185 return; 1186 1187 snprintf(name, sizeof(name), "l3c%d", ctx->edac_idx); 1188 dbgfs_dir = edac_debugfs_create_dir_at(name, ctx->edac->dfs); 1189 if (!dbgfs_dir) 1190 return; 1191 1192 debugfs_create_file("l3_inject_ctrl", S_IWUSR, dbgfs_dir, edac_dev, 1193 &xgene_edac_l3_debug_inject_fops); 1194 } 1195 1196 static int xgene_edac_l3_add(struct xgene_edac *edac, struct device_node *np, 1197 int version) 1198 { 1199 struct edac_device_ctl_info *edac_dev; 1200 struct xgene_edac_dev_ctx *ctx; 1201 struct resource res; 1202 void __iomem *dev_csr; 1203 int edac_idx; 1204 int rc = 0; 1205 1206 if (!devres_open_group(edac->dev, xgene_edac_l3_add, GFP_KERNEL)) 1207 return -ENOMEM; 1208 1209 rc = of_address_to_resource(np, 0, &res); 1210 if (rc < 0) { 1211 dev_err(edac->dev, "no L3 resource address\n"); 1212 goto err_release_group; 1213 } 1214 dev_csr = devm_ioremap_resource(edac->dev, &res); 1215 if (IS_ERR(dev_csr)) { 1216 dev_err(edac->dev, 1217 "devm_ioremap_resource failed for L3 resource address\n"); 1218 rc = PTR_ERR(dev_csr); 1219 goto err_release_group; 1220 } 1221 1222 edac_idx = edac_device_alloc_index(); 1223 edac_dev = edac_device_alloc_ctl_info(sizeof(*ctx), 1224 "l3c", 1, "l3c", 1, 0, NULL, 0, 1225 edac_idx); 1226 if (!edac_dev) { 1227 rc = -ENOMEM; 1228 goto err_release_group; 1229 } 1230 1231 ctx = edac_dev->pvt_info; 1232 ctx->dev_csr = dev_csr; 1233 ctx->name = "xgene_l3_err"; 1234 ctx->edac_idx = edac_idx; 1235 ctx->edac = edac; 1236 ctx->edac_dev = edac_dev; 1237 ctx->ddev = *edac->dev; 1238 ctx->version = version; 1239 edac_dev->dev = &ctx->ddev; 1240 edac_dev->ctl_name = ctx->name; 1241 edac_dev->dev_name = ctx->name; 1242 edac_dev->mod_name = EDAC_MOD_STR; 1243 1244 if (edac_op_state == EDAC_OPSTATE_POLL) 1245 edac_dev->edac_check = xgene_edac_l3_check; 1246 1247 xgene_edac_l3_create_debugfs_nodes(edac_dev); 1248 1249 rc = edac_device_add_device(edac_dev); 1250 if (rc > 0) { 1251 dev_err(edac->dev, "failed edac_device_add_device()\n"); 1252 rc = -ENOMEM; 1253 goto err_ctl_free; 1254 } 1255 1256 if (edac_op_state == EDAC_OPSTATE_INT) 1257 edac_dev->op_state = OP_RUNNING_INTERRUPT; 1258 1259 list_add(&ctx->next, &edac->l3s); 1260 1261 xgene_edac_l3_hw_init(edac_dev, 1); 1262 1263 devres_remove_group(edac->dev, xgene_edac_l3_add); 1264 1265 dev_info(edac->dev, "X-Gene EDAC L3 registered\n"); 1266 return 0; 1267 1268 err_ctl_free: 1269 edac_device_free_ctl_info(edac_dev); 1270 err_release_group: 1271 devres_release_group(edac->dev, xgene_edac_l3_add); 1272 return rc; 1273 } 1274 1275 static int xgene_edac_l3_remove(struct xgene_edac_dev_ctx *l3) 1276 { 1277 struct edac_device_ctl_info *edac_dev = l3->edac_dev; 1278 1279 xgene_edac_l3_hw_init(edac_dev, 0); 1280 edac_device_del_device(l3->edac->dev); 1281 edac_device_free_ctl_info(edac_dev); 1282 return 0; 1283 } 1284 1285 /* SoC error device */ 1286 #define IOBAXIS0TRANSERRINTSTS 0x0000 1287 #define IOBAXIS0_M_ILLEGAL_ACCESS_MASK BIT(1) 1288 #define IOBAXIS0_ILLEGAL_ACCESS_MASK BIT(0) 1289 #define IOBAXIS0TRANSERRINTMSK 0x0004 1290 #define IOBAXIS0TRANSERRREQINFOL 0x0008 1291 #define IOBAXIS0TRANSERRREQINFOH 0x000c 1292 #define REQTYPE_RD(src) (((src) & BIT(0))) 1293 #define ERRADDRH_RD(src) (((src) & 0xffc00000) >> 22) 1294 #define IOBAXIS1TRANSERRINTSTS 0x0010 1295 #define IOBAXIS1TRANSERRINTMSK 0x0014 1296 #define IOBAXIS1TRANSERRREQINFOL 0x0018 1297 #define IOBAXIS1TRANSERRREQINFOH 0x001c 1298 #define IOBPATRANSERRINTSTS 0x0020 1299 #define IOBPA_M_REQIDRAM_CORRUPT_MASK BIT(7) 1300 #define IOBPA_REQIDRAM_CORRUPT_MASK BIT(6) 1301 #define IOBPA_M_TRANS_CORRUPT_MASK BIT(5) 1302 #define IOBPA_TRANS_CORRUPT_MASK BIT(4) 1303 #define IOBPA_M_WDATA_CORRUPT_MASK BIT(3) 1304 #define IOBPA_WDATA_CORRUPT_MASK BIT(2) 1305 #define IOBPA_M_RDATA_CORRUPT_MASK BIT(1) 1306 #define IOBPA_RDATA_CORRUPT_MASK BIT(0) 1307 #define IOBBATRANSERRINTSTS 0x0030 1308 #define M_ILLEGAL_ACCESS_MASK BIT(15) 1309 #define ILLEGAL_ACCESS_MASK BIT(14) 1310 #define M_WIDRAM_CORRUPT_MASK BIT(13) 1311 #define WIDRAM_CORRUPT_MASK BIT(12) 1312 #define M_RIDRAM_CORRUPT_MASK BIT(11) 1313 #define RIDRAM_CORRUPT_MASK BIT(10) 1314 #define M_TRANS_CORRUPT_MASK BIT(9) 1315 #define TRANS_CORRUPT_MASK BIT(8) 1316 #define M_WDATA_CORRUPT_MASK BIT(7) 1317 #define WDATA_CORRUPT_MASK BIT(6) 1318 #define M_RBM_POISONED_REQ_MASK BIT(5) 1319 #define RBM_POISONED_REQ_MASK BIT(4) 1320 #define M_XGIC_POISONED_REQ_MASK BIT(3) 1321 #define XGIC_POISONED_REQ_MASK BIT(2) 1322 #define M_WRERR_RESP_MASK BIT(1) 1323 #define WRERR_RESP_MASK BIT(0) 1324 #define IOBBATRANSERRREQINFOL 0x0038 1325 #define IOBBATRANSERRREQINFOH 0x003c 1326 #define REQTYPE_F2_RD(src) ((src) & BIT(0)) 1327 #define ERRADDRH_F2_RD(src) (((src) & 0xffc00000) >> 22) 1328 #define IOBBATRANSERRCSWREQID 0x0040 1329 #define XGICTRANSERRINTSTS 0x0050 1330 #define M_WR_ACCESS_ERR_MASK BIT(3) 1331 #define WR_ACCESS_ERR_MASK BIT(2) 1332 #define M_RD_ACCESS_ERR_MASK BIT(1) 1333 #define RD_ACCESS_ERR_MASK BIT(0) 1334 #define XGICTRANSERRINTMSK 0x0054 1335 #define XGICTRANSERRREQINFO 0x0058 1336 #define REQTYPE_MASK BIT(26) 1337 #define ERRADDR_RD(src) ((src) & 0x03ffffff) 1338 #define GLBL_ERR_STS 0x0800 1339 #define MDED_ERR_MASK BIT(3) 1340 #define DED_ERR_MASK BIT(2) 1341 #define MSEC_ERR_MASK BIT(1) 1342 #define SEC_ERR_MASK BIT(0) 1343 #define GLBL_SEC_ERRL 0x0810 1344 #define GLBL_SEC_ERRH 0x0818 1345 #define GLBL_MSEC_ERRL 0x0820 1346 #define GLBL_MSEC_ERRH 0x0828 1347 #define GLBL_DED_ERRL 0x0830 1348 #define GLBL_DED_ERRLMASK 0x0834 1349 #define GLBL_DED_ERRH 0x0838 1350 #define GLBL_DED_ERRHMASK 0x083c 1351 #define GLBL_MDED_ERRL 0x0840 1352 #define GLBL_MDED_ERRLMASK 0x0844 1353 #define GLBL_MDED_ERRH 0x0848 1354 #define GLBL_MDED_ERRHMASK 0x084c 1355 1356 static const char * const soc_mem_err_v1[] = { 1357 "10GbE0", 1358 "10GbE1", 1359 "Security", 1360 "SATA45", 1361 "SATA23/ETH23", 1362 "SATA01/ETH01", 1363 "USB1", 1364 "USB0", 1365 "QML", 1366 "QM0", 1367 "QM1 (XGbE01)", 1368 "PCIE4", 1369 "PCIE3", 1370 "PCIE2", 1371 "PCIE1", 1372 "PCIE0", 1373 "CTX Manager", 1374 "OCM", 1375 "1GbE", 1376 "CLE", 1377 "AHBC", 1378 "PktDMA", 1379 "GFC", 1380 "MSLIM", 1381 "10GbE2", 1382 "10GbE3", 1383 "QM2 (XGbE23)", 1384 "IOB", 1385 "unknown", 1386 "unknown", 1387 "unknown", 1388 "unknown", 1389 }; 1390 1391 static void xgene_edac_iob_gic_report(struct edac_device_ctl_info *edac_dev) 1392 { 1393 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info; 1394 u32 err_addr_lo; 1395 u32 err_addr_hi; 1396 u32 reg; 1397 u32 info; 1398 1399 /* GIC transaction error interrupt */ 1400 reg = readl(ctx->dev_csr + XGICTRANSERRINTSTS); 1401 if (!reg) 1402 goto chk_iob_err; 1403 dev_err(edac_dev->dev, "XGIC transaction error\n"); 1404 if (reg & RD_ACCESS_ERR_MASK) 1405 dev_err(edac_dev->dev, "XGIC read size error\n"); 1406 if (reg & M_RD_ACCESS_ERR_MASK) 1407 dev_err(edac_dev->dev, "Multiple XGIC read size error\n"); 1408 if (reg & WR_ACCESS_ERR_MASK) 1409 dev_err(edac_dev->dev, "XGIC write size error\n"); 1410 if (reg & M_WR_ACCESS_ERR_MASK) 1411 dev_err(edac_dev->dev, "Multiple XGIC write size error\n"); 1412 info = readl(ctx->dev_csr + XGICTRANSERRREQINFO); 1413 dev_err(edac_dev->dev, "XGIC %s access @ 0x%08X (0x%08X)\n", 1414 info & REQTYPE_MASK ? "read" : "write", ERRADDR_RD(info), 1415 info); 1416 writel(reg, ctx->dev_csr + XGICTRANSERRINTSTS); 1417 1418 chk_iob_err: 1419 /* IOB memory error */ 1420 reg = readl(ctx->dev_csr + GLBL_ERR_STS); 1421 if (!reg) 1422 return; 1423 if (reg & SEC_ERR_MASK) { 1424 err_addr_lo = readl(ctx->dev_csr + GLBL_SEC_ERRL); 1425 err_addr_hi = readl(ctx->dev_csr + GLBL_SEC_ERRH); 1426 dev_err(edac_dev->dev, 1427 "IOB single-bit correctable memory at 0x%08X.%08X error\n", 1428 err_addr_lo, err_addr_hi); 1429 writel(err_addr_lo, ctx->dev_csr + GLBL_SEC_ERRL); 1430 writel(err_addr_hi, ctx->dev_csr + GLBL_SEC_ERRH); 1431 } 1432 if (reg & MSEC_ERR_MASK) { 1433 err_addr_lo = readl(ctx->dev_csr + GLBL_MSEC_ERRL); 1434 err_addr_hi = readl(ctx->dev_csr + GLBL_MSEC_ERRH); 1435 dev_err(edac_dev->dev, 1436 "IOB multiple single-bit correctable memory at 0x%08X.%08X error\n", 1437 err_addr_lo, err_addr_hi); 1438 writel(err_addr_lo, ctx->dev_csr + GLBL_MSEC_ERRL); 1439 writel(err_addr_hi, ctx->dev_csr + GLBL_MSEC_ERRH); 1440 } 1441 if (reg & (SEC_ERR_MASK | MSEC_ERR_MASK)) 1442 edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name); 1443 1444 if (reg & DED_ERR_MASK) { 1445 err_addr_lo = readl(ctx->dev_csr + GLBL_DED_ERRL); 1446 err_addr_hi = readl(ctx->dev_csr + GLBL_DED_ERRH); 1447 dev_err(edac_dev->dev, 1448 "IOB double-bit uncorrectable memory at 0x%08X.%08X error\n", 1449 err_addr_lo, err_addr_hi); 1450 writel(err_addr_lo, ctx->dev_csr + GLBL_DED_ERRL); 1451 writel(err_addr_hi, ctx->dev_csr + GLBL_DED_ERRH); 1452 } 1453 if (reg & MDED_ERR_MASK) { 1454 err_addr_lo = readl(ctx->dev_csr + GLBL_MDED_ERRL); 1455 err_addr_hi = readl(ctx->dev_csr + GLBL_MDED_ERRH); 1456 dev_err(edac_dev->dev, 1457 "Multiple IOB double-bit uncorrectable memory at 0x%08X.%08X error\n", 1458 err_addr_lo, err_addr_hi); 1459 writel(err_addr_lo, ctx->dev_csr + GLBL_MDED_ERRL); 1460 writel(err_addr_hi, ctx->dev_csr + GLBL_MDED_ERRH); 1461 } 1462 if (reg & (DED_ERR_MASK | MDED_ERR_MASK)) 1463 edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name); 1464 } 1465 1466 static void xgene_edac_rb_report(struct edac_device_ctl_info *edac_dev) 1467 { 1468 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info; 1469 u32 err_addr_lo; 1470 u32 err_addr_hi; 1471 u32 reg; 1472 1473 /* IOB Bridge agent transaction error interrupt */ 1474 reg = readl(ctx->dev_csr + IOBBATRANSERRINTSTS); 1475 if (!reg) 1476 return; 1477 1478 dev_err(edac_dev->dev, "IOB bridge agent (BA) transaction error\n"); 1479 if (reg & WRERR_RESP_MASK) 1480 dev_err(edac_dev->dev, "IOB BA write response error\n"); 1481 if (reg & M_WRERR_RESP_MASK) 1482 dev_err(edac_dev->dev, 1483 "Multiple IOB BA write response error\n"); 1484 if (reg & XGIC_POISONED_REQ_MASK) 1485 dev_err(edac_dev->dev, "IOB BA XGIC poisoned write error\n"); 1486 if (reg & M_XGIC_POISONED_REQ_MASK) 1487 dev_err(edac_dev->dev, 1488 "Multiple IOB BA XGIC poisoned write error\n"); 1489 if (reg & RBM_POISONED_REQ_MASK) 1490 dev_err(edac_dev->dev, "IOB BA RBM poisoned write error\n"); 1491 if (reg & M_RBM_POISONED_REQ_MASK) 1492 dev_err(edac_dev->dev, 1493 "Multiple IOB BA RBM poisoned write error\n"); 1494 if (reg & WDATA_CORRUPT_MASK) 1495 dev_err(edac_dev->dev, "IOB BA write error\n"); 1496 if (reg & M_WDATA_CORRUPT_MASK) 1497 dev_err(edac_dev->dev, "Multiple IOB BA write error\n"); 1498 if (reg & TRANS_CORRUPT_MASK) 1499 dev_err(edac_dev->dev, "IOB BA transaction error\n"); 1500 if (reg & M_TRANS_CORRUPT_MASK) 1501 dev_err(edac_dev->dev, "Multiple IOB BA transaction error\n"); 1502 if (reg & RIDRAM_CORRUPT_MASK) 1503 dev_err(edac_dev->dev, 1504 "IOB BA RDIDRAM read transaction ID error\n"); 1505 if (reg & M_RIDRAM_CORRUPT_MASK) 1506 dev_err(edac_dev->dev, 1507 "Multiple IOB BA RDIDRAM read transaction ID error\n"); 1508 if (reg & WIDRAM_CORRUPT_MASK) 1509 dev_err(edac_dev->dev, 1510 "IOB BA RDIDRAM write transaction ID error\n"); 1511 if (reg & M_WIDRAM_CORRUPT_MASK) 1512 dev_err(edac_dev->dev, 1513 "Multiple IOB BA RDIDRAM write transaction ID error\n"); 1514 if (reg & ILLEGAL_ACCESS_MASK) 1515 dev_err(edac_dev->dev, 1516 "IOB BA XGIC/RB illegal access error\n"); 1517 if (reg & M_ILLEGAL_ACCESS_MASK) 1518 dev_err(edac_dev->dev, 1519 "Multiple IOB BA XGIC/RB illegal access error\n"); 1520 1521 err_addr_lo = readl(ctx->dev_csr + IOBBATRANSERRREQINFOL); 1522 err_addr_hi = readl(ctx->dev_csr + IOBBATRANSERRREQINFOH); 1523 dev_err(edac_dev->dev, "IOB BA %s access at 0x%02X.%08X (0x%08X)\n", 1524 REQTYPE_F2_RD(err_addr_hi) ? "read" : "write", 1525 ERRADDRH_F2_RD(err_addr_hi), err_addr_lo, err_addr_hi); 1526 if (reg & WRERR_RESP_MASK) 1527 dev_err(edac_dev->dev, "IOB BA requestor ID 0x%08X\n", 1528 readl(ctx->dev_csr + IOBBATRANSERRCSWREQID)); 1529 writel(reg, ctx->dev_csr + IOBBATRANSERRINTSTS); 1530 } 1531 1532 static void xgene_edac_pa_report(struct edac_device_ctl_info *edac_dev) 1533 { 1534 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info; 1535 u32 err_addr_lo; 1536 u32 err_addr_hi; 1537 u32 reg; 1538 1539 /* IOB Processing agent transaction error interrupt */ 1540 reg = readl(ctx->dev_csr + IOBPATRANSERRINTSTS); 1541 if (!reg) 1542 goto chk_iob_axi0; 1543 dev_err(edac_dev->dev, "IOB procesing agent (PA) transaction error\n"); 1544 if (reg & IOBPA_RDATA_CORRUPT_MASK) 1545 dev_err(edac_dev->dev, "IOB PA read data RAM error\n"); 1546 if (reg & IOBPA_M_RDATA_CORRUPT_MASK) 1547 dev_err(edac_dev->dev, 1548 "Mutilple IOB PA read data RAM error\n"); 1549 if (reg & IOBPA_WDATA_CORRUPT_MASK) 1550 dev_err(edac_dev->dev, "IOB PA write data RAM error\n"); 1551 if (reg & IOBPA_M_WDATA_CORRUPT_MASK) 1552 dev_err(edac_dev->dev, 1553 "Mutilple IOB PA write data RAM error\n"); 1554 if (reg & IOBPA_TRANS_CORRUPT_MASK) 1555 dev_err(edac_dev->dev, "IOB PA transaction error\n"); 1556 if (reg & IOBPA_M_TRANS_CORRUPT_MASK) 1557 dev_err(edac_dev->dev, "Mutilple IOB PA transaction error\n"); 1558 if (reg & IOBPA_REQIDRAM_CORRUPT_MASK) 1559 dev_err(edac_dev->dev, "IOB PA transaction ID RAM error\n"); 1560 if (reg & IOBPA_M_REQIDRAM_CORRUPT_MASK) 1561 dev_err(edac_dev->dev, 1562 "Multiple IOB PA transaction ID RAM error\n"); 1563 writel(reg, ctx->dev_csr + IOBPATRANSERRINTSTS); 1564 1565 chk_iob_axi0: 1566 /* IOB AXI0 Error */ 1567 reg = readl(ctx->dev_csr + IOBAXIS0TRANSERRINTSTS); 1568 if (!reg) 1569 goto chk_iob_axi1; 1570 err_addr_lo = readl(ctx->dev_csr + IOBAXIS0TRANSERRREQINFOL); 1571 err_addr_hi = readl(ctx->dev_csr + IOBAXIS0TRANSERRREQINFOH); 1572 dev_err(edac_dev->dev, 1573 "%sAXI slave 0 illegal %s access @ 0x%02X.%08X (0x%08X)\n", 1574 reg & IOBAXIS0_M_ILLEGAL_ACCESS_MASK ? "Multiple " : "", 1575 REQTYPE_RD(err_addr_hi) ? "read" : "write", 1576 ERRADDRH_RD(err_addr_hi), err_addr_lo, err_addr_hi); 1577 writel(reg, ctx->dev_csr + IOBAXIS0TRANSERRINTSTS); 1578 1579 chk_iob_axi1: 1580 /* IOB AXI1 Error */ 1581 reg = readl(ctx->dev_csr + IOBAXIS1TRANSERRINTSTS); 1582 if (!reg) 1583 return; 1584 err_addr_lo = readl(ctx->dev_csr + IOBAXIS1TRANSERRREQINFOL); 1585 err_addr_hi = readl(ctx->dev_csr + IOBAXIS1TRANSERRREQINFOH); 1586 dev_err(edac_dev->dev, 1587 "%sAXI slave 1 illegal %s access @ 0x%02X.%08X (0x%08X)\n", 1588 reg & IOBAXIS0_M_ILLEGAL_ACCESS_MASK ? "Multiple " : "", 1589 REQTYPE_RD(err_addr_hi) ? "read" : "write", 1590 ERRADDRH_RD(err_addr_hi), err_addr_lo, err_addr_hi); 1591 writel(reg, ctx->dev_csr + IOBAXIS1TRANSERRINTSTS); 1592 } 1593 1594 static void xgene_edac_soc_check(struct edac_device_ctl_info *edac_dev) 1595 { 1596 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info; 1597 const char * const *soc_mem_err = NULL; 1598 u32 pcp_hp_stat; 1599 u32 pcp_lp_stat; 1600 u32 reg; 1601 int i; 1602 1603 xgene_edac_pcp_rd(ctx->edac, PCPHPERRINTSTS, &pcp_hp_stat); 1604 xgene_edac_pcp_rd(ctx->edac, PCPLPERRINTSTS, &pcp_lp_stat); 1605 xgene_edac_pcp_rd(ctx->edac, MEMERRINTSTS, ®); 1606 if (!((pcp_hp_stat & (IOB_PA_ERR_MASK | IOB_BA_ERR_MASK | 1607 IOB_XGIC_ERR_MASK | IOB_RB_ERR_MASK)) || 1608 (pcp_lp_stat & CSW_SWITCH_TRACE_ERR_MASK) || reg)) 1609 return; 1610 1611 if (pcp_hp_stat & IOB_XGIC_ERR_MASK) 1612 xgene_edac_iob_gic_report(edac_dev); 1613 1614 if (pcp_hp_stat & (IOB_RB_ERR_MASK | IOB_BA_ERR_MASK)) 1615 xgene_edac_rb_report(edac_dev); 1616 1617 if (pcp_hp_stat & IOB_PA_ERR_MASK) 1618 xgene_edac_pa_report(edac_dev); 1619 1620 if (pcp_lp_stat & CSW_SWITCH_TRACE_ERR_MASK) { 1621 dev_info(edac_dev->dev, 1622 "CSW switch trace correctable memory parity error\n"); 1623 edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name); 1624 } 1625 1626 if (!reg) 1627 return; 1628 if (ctx->version == 1) 1629 soc_mem_err = soc_mem_err_v1; 1630 if (!soc_mem_err) { 1631 dev_err(edac_dev->dev, "SoC memory parity error 0x%08X\n", 1632 reg); 1633 edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name); 1634 return; 1635 } 1636 for (i = 0; i < 31; i++) { 1637 if (reg & (1 << i)) { 1638 dev_err(edac_dev->dev, "%s memory parity error\n", 1639 soc_mem_err[i]); 1640 edac_device_handle_ue(edac_dev, 0, 0, 1641 edac_dev->ctl_name); 1642 } 1643 } 1644 } 1645 1646 static void xgene_edac_soc_hw_init(struct edac_device_ctl_info *edac_dev, 1647 bool enable) 1648 { 1649 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info; 1650 1651 /* Enable SoC IP error interrupt */ 1652 if (edac_dev->op_state == OP_RUNNING_INTERRUPT) { 1653 if (enable) { 1654 xgene_edac_pcp_clrbits(ctx->edac, PCPHPERRINTMSK, 1655 IOB_PA_ERR_MASK | 1656 IOB_BA_ERR_MASK | 1657 IOB_XGIC_ERR_MASK | 1658 IOB_RB_ERR_MASK); 1659 xgene_edac_pcp_clrbits(ctx->edac, PCPLPERRINTMSK, 1660 CSW_SWITCH_TRACE_ERR_MASK); 1661 } else { 1662 xgene_edac_pcp_setbits(ctx->edac, PCPHPERRINTMSK, 1663 IOB_PA_ERR_MASK | 1664 IOB_BA_ERR_MASK | 1665 IOB_XGIC_ERR_MASK | 1666 IOB_RB_ERR_MASK); 1667 xgene_edac_pcp_setbits(ctx->edac, PCPLPERRINTMSK, 1668 CSW_SWITCH_TRACE_ERR_MASK); 1669 } 1670 1671 writel(enable ? 0x0 : 0xFFFFFFFF, 1672 ctx->dev_csr + IOBAXIS0TRANSERRINTMSK); 1673 writel(enable ? 0x0 : 0xFFFFFFFF, 1674 ctx->dev_csr + IOBAXIS1TRANSERRINTMSK); 1675 writel(enable ? 0x0 : 0xFFFFFFFF, 1676 ctx->dev_csr + XGICTRANSERRINTMSK); 1677 1678 xgene_edac_pcp_setbits(ctx->edac, MEMERRINTMSK, 1679 enable ? 0x0 : 0xFFFFFFFF); 1680 } 1681 } 1682 1683 static int xgene_edac_soc_add(struct xgene_edac *edac, struct device_node *np, 1684 int version) 1685 { 1686 struct edac_device_ctl_info *edac_dev; 1687 struct xgene_edac_dev_ctx *ctx; 1688 void __iomem *dev_csr; 1689 struct resource res; 1690 int edac_idx; 1691 int rc; 1692 1693 if (!devres_open_group(edac->dev, xgene_edac_soc_add, GFP_KERNEL)) 1694 return -ENOMEM; 1695 1696 rc = of_address_to_resource(np, 0, &res); 1697 if (rc < 0) { 1698 dev_err(edac->dev, "no SoC resource address\n"); 1699 goto err_release_group; 1700 } 1701 dev_csr = devm_ioremap_resource(edac->dev, &res); 1702 if (IS_ERR(dev_csr)) { 1703 dev_err(edac->dev, 1704 "devm_ioremap_resource failed for soc resource address\n"); 1705 rc = PTR_ERR(dev_csr); 1706 goto err_release_group; 1707 } 1708 1709 edac_idx = edac_device_alloc_index(); 1710 edac_dev = edac_device_alloc_ctl_info(sizeof(*ctx), 1711 "SOC", 1, "SOC", 1, 2, NULL, 0, 1712 edac_idx); 1713 if (!edac_dev) { 1714 rc = -ENOMEM; 1715 goto err_release_group; 1716 } 1717 1718 ctx = edac_dev->pvt_info; 1719 ctx->dev_csr = dev_csr; 1720 ctx->name = "xgene_soc_err"; 1721 ctx->edac_idx = edac_idx; 1722 ctx->edac = edac; 1723 ctx->edac_dev = edac_dev; 1724 ctx->ddev = *edac->dev; 1725 ctx->version = version; 1726 edac_dev->dev = &ctx->ddev; 1727 edac_dev->ctl_name = ctx->name; 1728 edac_dev->dev_name = ctx->name; 1729 edac_dev->mod_name = EDAC_MOD_STR; 1730 1731 if (edac_op_state == EDAC_OPSTATE_POLL) 1732 edac_dev->edac_check = xgene_edac_soc_check; 1733 1734 rc = edac_device_add_device(edac_dev); 1735 if (rc > 0) { 1736 dev_err(edac->dev, "failed edac_device_add_device()\n"); 1737 rc = -ENOMEM; 1738 goto err_ctl_free; 1739 } 1740 1741 if (edac_op_state == EDAC_OPSTATE_INT) 1742 edac_dev->op_state = OP_RUNNING_INTERRUPT; 1743 1744 list_add(&ctx->next, &edac->socs); 1745 1746 xgene_edac_soc_hw_init(edac_dev, 1); 1747 1748 devres_remove_group(edac->dev, xgene_edac_soc_add); 1749 1750 dev_info(edac->dev, "X-Gene EDAC SoC registered\n"); 1751 1752 return 0; 1753 1754 err_ctl_free: 1755 edac_device_free_ctl_info(edac_dev); 1756 err_release_group: 1757 devres_release_group(edac->dev, xgene_edac_soc_add); 1758 return rc; 1759 } 1760 1761 static int xgene_edac_soc_remove(struct xgene_edac_dev_ctx *soc) 1762 { 1763 struct edac_device_ctl_info *edac_dev = soc->edac_dev; 1764 1765 xgene_edac_soc_hw_init(edac_dev, 0); 1766 edac_device_del_device(soc->edac->dev); 1767 edac_device_free_ctl_info(edac_dev); 1768 return 0; 1769 } 1770 1771 static irqreturn_t xgene_edac_isr(int irq, void *dev_id) 1772 { 1773 struct xgene_edac *ctx = dev_id; 1774 struct xgene_edac_pmd_ctx *pmd; 1775 struct xgene_edac_dev_ctx *node; 1776 unsigned int pcp_hp_stat; 1777 unsigned int pcp_lp_stat; 1778 1779 xgene_edac_pcp_rd(ctx, PCPHPERRINTSTS, &pcp_hp_stat); 1780 xgene_edac_pcp_rd(ctx, PCPLPERRINTSTS, &pcp_lp_stat); 1781 if ((MCU_UNCORR_ERR_MASK & pcp_hp_stat) || 1782 (MCU_CTL_ERR_MASK & pcp_hp_stat) || 1783 (MCU_CORR_ERR_MASK & pcp_lp_stat)) { 1784 struct xgene_edac_mc_ctx *mcu; 1785 1786 list_for_each_entry(mcu, &ctx->mcus, next) 1787 xgene_edac_mc_check(mcu->mci); 1788 } 1789 1790 list_for_each_entry(pmd, &ctx->pmds, next) { 1791 if ((PMD0_MERR_MASK << pmd->pmd) & pcp_hp_stat) 1792 xgene_edac_pmd_check(pmd->edac_dev); 1793 } 1794 1795 list_for_each_entry(node, &ctx->l3s, next) 1796 xgene_edac_l3_check(node->edac_dev); 1797 1798 list_for_each_entry(node, &ctx->socs, next) 1799 xgene_edac_soc_check(node->edac_dev); 1800 1801 return IRQ_HANDLED; 1802 } 1803 1804 static int xgene_edac_probe(struct platform_device *pdev) 1805 { 1806 struct xgene_edac *edac; 1807 struct device_node *child; 1808 struct resource *res; 1809 int rc; 1810 1811 edac = devm_kzalloc(&pdev->dev, sizeof(*edac), GFP_KERNEL); 1812 if (!edac) 1813 return -ENOMEM; 1814 1815 edac->dev = &pdev->dev; 1816 platform_set_drvdata(pdev, edac); 1817 INIT_LIST_HEAD(&edac->mcus); 1818 INIT_LIST_HEAD(&edac->pmds); 1819 INIT_LIST_HEAD(&edac->l3s); 1820 INIT_LIST_HEAD(&edac->socs); 1821 spin_lock_init(&edac->lock); 1822 mutex_init(&edac->mc_lock); 1823 1824 edac->csw_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node, 1825 "regmap-csw"); 1826 if (IS_ERR(edac->csw_map)) { 1827 dev_err(edac->dev, "unable to get syscon regmap csw\n"); 1828 rc = PTR_ERR(edac->csw_map); 1829 goto out_err; 1830 } 1831 1832 edac->mcba_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node, 1833 "regmap-mcba"); 1834 if (IS_ERR(edac->mcba_map)) { 1835 dev_err(edac->dev, "unable to get syscon regmap mcba\n"); 1836 rc = PTR_ERR(edac->mcba_map); 1837 goto out_err; 1838 } 1839 1840 edac->mcbb_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node, 1841 "regmap-mcbb"); 1842 if (IS_ERR(edac->mcbb_map)) { 1843 dev_err(edac->dev, "unable to get syscon regmap mcbb\n"); 1844 rc = PTR_ERR(edac->mcbb_map); 1845 goto out_err; 1846 } 1847 edac->efuse_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node, 1848 "regmap-efuse"); 1849 if (IS_ERR(edac->efuse_map)) { 1850 dev_err(edac->dev, "unable to get syscon regmap efuse\n"); 1851 rc = PTR_ERR(edac->efuse_map); 1852 goto out_err; 1853 } 1854 1855 res = platform_get_resource(pdev, IORESOURCE_MEM, 0); 1856 edac->pcp_csr = devm_ioremap_resource(&pdev->dev, res); 1857 if (IS_ERR(edac->pcp_csr)) { 1858 dev_err(&pdev->dev, "no PCP resource address\n"); 1859 rc = PTR_ERR(edac->pcp_csr); 1860 goto out_err; 1861 } 1862 1863 if (edac_op_state == EDAC_OPSTATE_INT) { 1864 int irq; 1865 int i; 1866 1867 for (i = 0; i < 3; i++) { 1868 irq = platform_get_irq(pdev, i); 1869 if (irq < 0) { 1870 dev_err(&pdev->dev, "No IRQ resource\n"); 1871 rc = -EINVAL; 1872 goto out_err; 1873 } 1874 rc = devm_request_irq(&pdev->dev, irq, 1875 xgene_edac_isr, IRQF_SHARED, 1876 dev_name(&pdev->dev), edac); 1877 if (rc) { 1878 dev_err(&pdev->dev, 1879 "Could not request IRQ %d\n", irq); 1880 goto out_err; 1881 } 1882 } 1883 } 1884 1885 edac->dfs = edac_debugfs_create_dir(pdev->dev.kobj.name); 1886 1887 for_each_child_of_node(pdev->dev.of_node, child) { 1888 if (!of_device_is_available(child)) 1889 continue; 1890 if (of_device_is_compatible(child, "apm,xgene-edac-mc")) 1891 xgene_edac_mc_add(edac, child); 1892 if (of_device_is_compatible(child, "apm,xgene-edac-pmd")) 1893 xgene_edac_pmd_add(edac, child, 1); 1894 if (of_device_is_compatible(child, "apm,xgene-edac-pmd-v2")) 1895 xgene_edac_pmd_add(edac, child, 2); 1896 if (of_device_is_compatible(child, "apm,xgene-edac-l3")) 1897 xgene_edac_l3_add(edac, child, 1); 1898 if (of_device_is_compatible(child, "apm,xgene-edac-l3-v2")) 1899 xgene_edac_l3_add(edac, child, 2); 1900 if (of_device_is_compatible(child, "apm,xgene-edac-soc")) 1901 xgene_edac_soc_add(edac, child, 0); 1902 if (of_device_is_compatible(child, "apm,xgene-edac-soc-v1")) 1903 xgene_edac_soc_add(edac, child, 1); 1904 } 1905 1906 return 0; 1907 1908 out_err: 1909 return rc; 1910 } 1911 1912 static int xgene_edac_remove(struct platform_device *pdev) 1913 { 1914 struct xgene_edac *edac = dev_get_drvdata(&pdev->dev); 1915 struct xgene_edac_mc_ctx *mcu; 1916 struct xgene_edac_mc_ctx *temp_mcu; 1917 struct xgene_edac_pmd_ctx *pmd; 1918 struct xgene_edac_pmd_ctx *temp_pmd; 1919 struct xgene_edac_dev_ctx *node; 1920 struct xgene_edac_dev_ctx *temp_node; 1921 1922 list_for_each_entry_safe(mcu, temp_mcu, &edac->mcus, next) 1923 xgene_edac_mc_remove(mcu); 1924 1925 list_for_each_entry_safe(pmd, temp_pmd, &edac->pmds, next) 1926 xgene_edac_pmd_remove(pmd); 1927 1928 list_for_each_entry_safe(node, temp_node, &edac->l3s, next) 1929 xgene_edac_l3_remove(node); 1930 1931 list_for_each_entry_safe(node, temp_node, &edac->socs, next) 1932 xgene_edac_soc_remove(node); 1933 1934 return 0; 1935 } 1936 1937 static const struct of_device_id xgene_edac_of_match[] = { 1938 { .compatible = "apm,xgene-edac" }, 1939 {}, 1940 }; 1941 MODULE_DEVICE_TABLE(of, xgene_edac_of_match); 1942 1943 static struct platform_driver xgene_edac_driver = { 1944 .probe = xgene_edac_probe, 1945 .remove = xgene_edac_remove, 1946 .driver = { 1947 .name = "xgene-edac", 1948 .of_match_table = xgene_edac_of_match, 1949 }, 1950 }; 1951 1952 static int __init xgene_edac_init(void) 1953 { 1954 int rc; 1955 1956 /* Make sure error reporting method is sane */ 1957 switch (edac_op_state) { 1958 case EDAC_OPSTATE_POLL: 1959 case EDAC_OPSTATE_INT: 1960 break; 1961 default: 1962 edac_op_state = EDAC_OPSTATE_INT; 1963 break; 1964 } 1965 1966 rc = platform_driver_register(&xgene_edac_driver); 1967 if (rc) { 1968 edac_printk(KERN_ERR, EDAC_MOD_STR, 1969 "EDAC fails to register\n"); 1970 goto reg_failed; 1971 } 1972 1973 return 0; 1974 1975 reg_failed: 1976 return rc; 1977 } 1978 module_init(xgene_edac_init); 1979 1980 static void __exit xgene_edac_exit(void) 1981 { 1982 platform_driver_unregister(&xgene_edac_driver); 1983 } 1984 module_exit(xgene_edac_exit); 1985 1986 MODULE_LICENSE("GPL"); 1987 MODULE_AUTHOR("Feng Kan <fkan@apm.com>"); 1988 MODULE_DESCRIPTION("APM X-Gene EDAC driver"); 1989 module_param(edac_op_state, int, 0444); 1990 MODULE_PARM_DESC(edac_op_state, 1991 "EDAC error reporting state: 0=Poll, 2=Interrupt"); 1992