Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
[linux/fpc-iii.git] / drivers / edac / xgene_edac.c
blobe8b81d7ef61fa16b80858518352cab2a0a8daf8f
1 /*
2 * APM X-Gene SoC EDAC (error detection and correction)
4 * Copyright (c) 2015, Applied Micro Circuits Corporation
5 * Author: Feng Kan <fkan@apm.com>
6 * Loc Ho <lho@apm.com>
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.
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.
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/>.
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>
31 #include "edac_module.h"
33 #define EDAC_MOD_STR "xgene_edac"
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
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 struct regmap *rb_map;
64 void __iomem *pcp_csr;
65 spinlock_t lock;
66 struct dentry *dfs;
68 struct list_head mcus;
69 struct list_head pmds;
70 struct list_head l3s;
71 struct list_head socs;
73 struct mutex mc_lock;
74 int mc_active_mask;
75 int mc_registered_mask;
78 static void xgene_edac_pcp_rd(struct xgene_edac *edac, u32 reg, u32 *val)
80 *val = readl(edac->pcp_csr + reg);
83 static void xgene_edac_pcp_clrbits(struct xgene_edac *edac, u32 reg,
84 u32 bits_mask)
86 u32 val;
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);
95 static void xgene_edac_pcp_setbits(struct xgene_edac *edac, u32 reg,
96 u32 bits_mask)
98 u32 val;
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);
107 /* Memory controller error CSR */
108 #define MCU_MAX_RANK 8
109 #define MCU_RANK_STRIDE 0x40
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)
134 #define CSW_CSWCR 0x0000
135 #define CSW_CSWCR_DUALMCB_MASK BIT(0)
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)
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;
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)
156 struct mem_ctl_info *mci = file->private_data;
157 struct xgene_edac_mc_ctx *ctx = mci->pvt_info;
158 int i;
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);
165 return count;
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,
174 static void xgene_edac_mc_create_debugfs_node(struct mem_ctl_info *mci)
176 if (!IS_ENABLED(CONFIG_EDAC_DEBUG))
177 return;
179 if (!mci->debugfs)
180 return;
182 edac_debugfs_create_file("inject_ctrl", S_IWUSR, mci->debugfs, mci,
183 &xgene_edac_mc_debug_inject_fops);
186 static void xgene_edac_mc_check(struct mem_ctl_info *mci)
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;
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;
204 for (rank = 0; rank < MCU_MAX_RANK; rank++) {
205 reg = readl(ctx->mcu_csr + MCUESRR0 + rank * MCU_RANK_STRIDE);
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);
214 edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci,
215 1, 0, 0, 0, 0, 0, -1, mci->ctl_name, "");
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));
233 edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci,
234 1, 0, 0, 0, 0, 0, -1, mci->ctl_name, "");
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);
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");
255 writel(reg, ctx->mcu_csr + MCUGESR);
259 static void xgene_edac_mc_irq_ctl(struct mem_ctl_info *mci, bool enable)
261 struct xgene_edac_mc_ctx *ctx = mci->pvt_info;
262 unsigned int val;
264 if (edac_op_state != EDAC_OPSTATE_INT)
265 return;
267 mutex_lock(&ctx->edac->mc_lock);
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.
277 if (enable) {
278 /* Set registered MCU bit */
279 ctx->edac->mc_registered_mask |= 1 << ctx->mcu_id;
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);
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);
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);
314 /* Clear registered MCU bit */
315 ctx->edac->mc_registered_mask &= ~(1 << ctx->mcu_id);
318 mutex_unlock(&ctx->edac->mc_lock);
321 static int xgene_edac_mc_is_active(struct xgene_edac_mc_ctx *ctx, int mc_idx)
323 unsigned int reg;
324 u32 mcu_mask;
326 if (regmap_read(ctx->edac->csw_map, CSW_CSWCR, &reg))
327 return 0;
329 if (reg & CSW_CSWCR_DUALMCB_MASK) {
331 * Dual MCB active - Determine if all 4 active or just MCU0
332 * and MCU2 active
334 if (regmap_read(ctx->edac->mcbb_map, MCBADDRMR, &reg))
335 return 0;
336 mcu_mask = (reg & MCBADDRMR_DUALMCU_MODE_MASK) ? 0xF : 0x5;
337 } else {
339 * Single MCB active - Determine if MCU0/MCU1 or just MCU0
340 * active
342 if (regmap_read(ctx->edac->mcba_map, MCBADDRMR, &reg))
343 return 0;
344 mcu_mask = (reg & MCBADDRMR_DUALMCU_MODE_MASK) ? 0x3 : 0x1;
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;
351 return (mcu_mask & (1 << mc_idx)) ? 1 : 0;
354 static int xgene_edac_mc_add(struct xgene_edac *edac, struct device_node *np)
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;
363 memset(&tmp_ctx, 0, sizeof(tmp_ctx));
364 tmp_ctx.edac = edac;
366 if (!devres_open_group(edac->dev, xgene_edac_mc_add, GFP_KERNEL))
367 return -ENOMEM;
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;
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;
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;
387 if (!xgene_edac_mc_is_active(&tmp_ctx, tmp_ctx.mcu_id)) {
388 rc = -ENODEV;
389 goto err_group;
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;
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;
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->ctl_page_to_phys = NULL;
419 mci->scrub_cap = SCRUB_FLAG_HW_SRC;
420 mci->scrub_mode = SCRUB_HW_SRC;
422 if (edac_op_state == EDAC_OPSTATE_POLL)
423 mci->edac_check = xgene_edac_mc_check;
425 if (edac_mc_add_mc(mci)) {
426 dev_err(edac->dev, "edac_mc_add_mc failed\n");
427 rc = -EINVAL;
428 goto err_free;
431 xgene_edac_mc_create_debugfs_node(mci);
433 list_add(&ctx->next, &edac->mcus);
435 xgene_edac_mc_irq_ctl(mci, true);
437 devres_remove_group(edac->dev, xgene_edac_mc_add);
439 dev_info(edac->dev, "X-Gene EDAC MC registered\n");
440 return 0;
442 err_free:
443 edac_mc_free(mci);
444 err_group:
445 devres_release_group(edac->dev, xgene_edac_mc_add);
446 return rc;
449 static int xgene_edac_mc_remove(struct xgene_edac_mc_ctx *mcu)
451 xgene_edac_mc_irq_ctl(mcu->mci, false);
452 edac_mc_del_mc(&mcu->mci->dev);
453 edac_mc_free(mcu->mci);
454 return 0;
457 /* CPU L1/L2 error CSR */
458 #define MAX_CPU_PER_PMD 2
459 #define CPU_CSR_STRIDE 0x00100000
460 #define CPU_L2C_PAGE 0x000D0000
461 #define CPU_MEMERR_L2C_PAGE 0x000E0000
462 #define CPU_MEMERR_CPU_PAGE 0x000F0000
464 #define MEMERR_CPU_ICFECR_PAGE_OFFSET 0x0000
465 #define MEMERR_CPU_ICFESR_PAGE_OFFSET 0x0004
466 #define MEMERR_CPU_ICFESR_ERRWAY_RD(src) (((src) & 0xFF000000) >> 24)
467 #define MEMERR_CPU_ICFESR_ERRINDEX_RD(src) (((src) & 0x003F0000) >> 16)
468 #define MEMERR_CPU_ICFESR_ERRINFO_RD(src) (((src) & 0x0000FF00) >> 8)
469 #define MEMERR_CPU_ICFESR_ERRTYPE_RD(src) (((src) & 0x00000070) >> 4)
470 #define MEMERR_CPU_ICFESR_MULTCERR_MASK BIT(2)
471 #define MEMERR_CPU_ICFESR_CERR_MASK BIT(0)
472 #define MEMERR_CPU_LSUESR_PAGE_OFFSET 0x000c
473 #define MEMERR_CPU_LSUESR_ERRWAY_RD(src) (((src) & 0xFF000000) >> 24)
474 #define MEMERR_CPU_LSUESR_ERRINDEX_RD(src) (((src) & 0x003F0000) >> 16)
475 #define MEMERR_CPU_LSUESR_ERRINFO_RD(src) (((src) & 0x0000FF00) >> 8)
476 #define MEMERR_CPU_LSUESR_ERRTYPE_RD(src) (((src) & 0x00000070) >> 4)
477 #define MEMERR_CPU_LSUESR_MULTCERR_MASK BIT(2)
478 #define MEMERR_CPU_LSUESR_CERR_MASK BIT(0)
479 #define MEMERR_CPU_LSUECR_PAGE_OFFSET 0x0008
480 #define MEMERR_CPU_MMUECR_PAGE_OFFSET 0x0010
481 #define MEMERR_CPU_MMUESR_PAGE_OFFSET 0x0014
482 #define MEMERR_CPU_MMUESR_ERRWAY_RD(src) (((src) & 0xFF000000) >> 24)
483 #define MEMERR_CPU_MMUESR_ERRINDEX_RD(src) (((src) & 0x007F0000) >> 16)
484 #define MEMERR_CPU_MMUESR_ERRINFO_RD(src) (((src) & 0x0000FF00) >> 8)
485 #define MEMERR_CPU_MMUESR_ERRREQSTR_LSU_MASK BIT(7)
486 #define MEMERR_CPU_MMUESR_ERRTYPE_RD(src) (((src) & 0x00000070) >> 4)
487 #define MEMERR_CPU_MMUESR_MULTCERR_MASK BIT(2)
488 #define MEMERR_CPU_MMUESR_CERR_MASK BIT(0)
489 #define MEMERR_CPU_ICFESRA_PAGE_OFFSET 0x0804
490 #define MEMERR_CPU_LSUESRA_PAGE_OFFSET 0x080c
491 #define MEMERR_CPU_MMUESRA_PAGE_OFFSET 0x0814
493 #define MEMERR_L2C_L2ECR_PAGE_OFFSET 0x0000
494 #define MEMERR_L2C_L2ESR_PAGE_OFFSET 0x0004
495 #define MEMERR_L2C_L2ESR_ERRSYN_RD(src) (((src) & 0xFF000000) >> 24)
496 #define MEMERR_L2C_L2ESR_ERRWAY_RD(src) (((src) & 0x00FC0000) >> 18)
497 #define MEMERR_L2C_L2ESR_ERRCPU_RD(src) (((src) & 0x00020000) >> 17)
498 #define MEMERR_L2C_L2ESR_ERRGROUP_RD(src) (((src) & 0x0000E000) >> 13)
499 #define MEMERR_L2C_L2ESR_ERRACTION_RD(src) (((src) & 0x00001C00) >> 10)
500 #define MEMERR_L2C_L2ESR_ERRTYPE_RD(src) (((src) & 0x00000300) >> 8)
501 #define MEMERR_L2C_L2ESR_MULTUCERR_MASK BIT(3)
502 #define MEMERR_L2C_L2ESR_MULTICERR_MASK BIT(2)
503 #define MEMERR_L2C_L2ESR_UCERR_MASK BIT(1)
504 #define MEMERR_L2C_L2ESR_ERR_MASK BIT(0)
505 #define MEMERR_L2C_L2EALR_PAGE_OFFSET 0x0008
506 #define CPUX_L2C_L2RTOCR_PAGE_OFFSET 0x0010
507 #define MEMERR_L2C_L2EAHR_PAGE_OFFSET 0x000c
508 #define CPUX_L2C_L2RTOSR_PAGE_OFFSET 0x0014
509 #define MEMERR_L2C_L2RTOSR_MULTERR_MASK BIT(1)
510 #define MEMERR_L2C_L2RTOSR_ERR_MASK BIT(0)
511 #define CPUX_L2C_L2RTOALR_PAGE_OFFSET 0x0018
512 #define CPUX_L2C_L2RTOAHR_PAGE_OFFSET 0x001c
513 #define MEMERR_L2C_L2ESRA_PAGE_OFFSET 0x0804
516 * Processor Module Domain (PMD) context - Context for a pair of processsors.
517 * Each PMD consists of 2 CPUs and a shared L2 cache. Each CPU consists of
518 * its own L1 cache.
520 struct xgene_edac_pmd_ctx {
521 struct list_head next;
522 struct device ddev;
523 char *name;
524 struct xgene_edac *edac;
525 struct edac_device_ctl_info *edac_dev;
526 void __iomem *pmd_csr;
527 u32 pmd;
528 int version;
531 static void xgene_edac_pmd_l1_check(struct edac_device_ctl_info *edac_dev,
532 int cpu_idx)
534 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
535 void __iomem *pg_f;
536 u32 val;
538 pg_f = ctx->pmd_csr + cpu_idx * CPU_CSR_STRIDE + CPU_MEMERR_CPU_PAGE;
540 val = readl(pg_f + MEMERR_CPU_ICFESR_PAGE_OFFSET);
541 if (!val)
542 goto chk_lsu;
543 dev_err(edac_dev->dev,
544 "CPU%d L1 memory error ICF 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X\n",
545 ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val,
546 MEMERR_CPU_ICFESR_ERRWAY_RD(val),
547 MEMERR_CPU_ICFESR_ERRINDEX_RD(val),
548 MEMERR_CPU_ICFESR_ERRINFO_RD(val));
549 if (val & MEMERR_CPU_ICFESR_CERR_MASK)
550 dev_err(edac_dev->dev, "One or more correctable error\n");
551 if (val & MEMERR_CPU_ICFESR_MULTCERR_MASK)
552 dev_err(edac_dev->dev, "Multiple correctable error\n");
553 switch (MEMERR_CPU_ICFESR_ERRTYPE_RD(val)) {
554 case 1:
555 dev_err(edac_dev->dev, "L1 TLB multiple hit\n");
556 break;
557 case 2:
558 dev_err(edac_dev->dev, "Way select multiple hit\n");
559 break;
560 case 3:
561 dev_err(edac_dev->dev, "Physical tag parity error\n");
562 break;
563 case 4:
564 case 5:
565 dev_err(edac_dev->dev, "L1 data parity error\n");
566 break;
567 case 6:
568 dev_err(edac_dev->dev, "L1 pre-decode parity error\n");
569 break;
572 /* Clear any HW errors */
573 writel(val, pg_f + MEMERR_CPU_ICFESR_PAGE_OFFSET);
575 if (val & (MEMERR_CPU_ICFESR_CERR_MASK |
576 MEMERR_CPU_ICFESR_MULTCERR_MASK))
577 edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
579 chk_lsu:
580 val = readl(pg_f + MEMERR_CPU_LSUESR_PAGE_OFFSET);
581 if (!val)
582 goto chk_mmu;
583 dev_err(edac_dev->dev,
584 "CPU%d memory error LSU 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X\n",
585 ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val,
586 MEMERR_CPU_LSUESR_ERRWAY_RD(val),
587 MEMERR_CPU_LSUESR_ERRINDEX_RD(val),
588 MEMERR_CPU_LSUESR_ERRINFO_RD(val));
589 if (val & MEMERR_CPU_LSUESR_CERR_MASK)
590 dev_err(edac_dev->dev, "One or more correctable error\n");
591 if (val & MEMERR_CPU_LSUESR_MULTCERR_MASK)
592 dev_err(edac_dev->dev, "Multiple correctable error\n");
593 switch (MEMERR_CPU_LSUESR_ERRTYPE_RD(val)) {
594 case 0:
595 dev_err(edac_dev->dev, "Load tag error\n");
596 break;
597 case 1:
598 dev_err(edac_dev->dev, "Load data error\n");
599 break;
600 case 2:
601 dev_err(edac_dev->dev, "WSL multihit error\n");
602 break;
603 case 3:
604 dev_err(edac_dev->dev, "Store tag error\n");
605 break;
606 case 4:
607 dev_err(edac_dev->dev,
608 "DTB multihit from load pipeline error\n");
609 break;
610 case 5:
611 dev_err(edac_dev->dev,
612 "DTB multihit from store pipeline error\n");
613 break;
616 /* Clear any HW errors */
617 writel(val, pg_f + MEMERR_CPU_LSUESR_PAGE_OFFSET);
619 if (val & (MEMERR_CPU_LSUESR_CERR_MASK |
620 MEMERR_CPU_LSUESR_MULTCERR_MASK))
621 edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
623 chk_mmu:
624 val = readl(pg_f + MEMERR_CPU_MMUESR_PAGE_OFFSET);
625 if (!val)
626 return;
627 dev_err(edac_dev->dev,
628 "CPU%d memory error MMU 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X %s\n",
629 ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val,
630 MEMERR_CPU_MMUESR_ERRWAY_RD(val),
631 MEMERR_CPU_MMUESR_ERRINDEX_RD(val),
632 MEMERR_CPU_MMUESR_ERRINFO_RD(val),
633 val & MEMERR_CPU_MMUESR_ERRREQSTR_LSU_MASK ? "LSU" : "ICF");
634 if (val & MEMERR_CPU_MMUESR_CERR_MASK)
635 dev_err(edac_dev->dev, "One or more correctable error\n");
636 if (val & MEMERR_CPU_MMUESR_MULTCERR_MASK)
637 dev_err(edac_dev->dev, "Multiple correctable error\n");
638 switch (MEMERR_CPU_MMUESR_ERRTYPE_RD(val)) {
639 case 0:
640 dev_err(edac_dev->dev, "Stage 1 UTB hit error\n");
641 break;
642 case 1:
643 dev_err(edac_dev->dev, "Stage 1 UTB miss error\n");
644 break;
645 case 2:
646 dev_err(edac_dev->dev, "Stage 1 UTB allocate error\n");
647 break;
648 case 3:
649 dev_err(edac_dev->dev, "TMO operation single bank error\n");
650 break;
651 case 4:
652 dev_err(edac_dev->dev, "Stage 2 UTB error\n");
653 break;
654 case 5:
655 dev_err(edac_dev->dev, "Stage 2 UTB miss error\n");
656 break;
657 case 6:
658 dev_err(edac_dev->dev, "Stage 2 UTB allocate error\n");
659 break;
660 case 7:
661 dev_err(edac_dev->dev, "TMO operation multiple bank error\n");
662 break;
665 /* Clear any HW errors */
666 writel(val, pg_f + MEMERR_CPU_MMUESR_PAGE_OFFSET);
668 edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
671 static void xgene_edac_pmd_l2_check(struct edac_device_ctl_info *edac_dev)
673 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
674 void __iomem *pg_d;
675 void __iomem *pg_e;
676 u32 val_hi;
677 u32 val_lo;
678 u32 val;
680 /* Check L2 */
681 pg_e = ctx->pmd_csr + CPU_MEMERR_L2C_PAGE;
682 val = readl(pg_e + MEMERR_L2C_L2ESR_PAGE_OFFSET);
683 if (!val)
684 goto chk_l2c;
685 val_lo = readl(pg_e + MEMERR_L2C_L2EALR_PAGE_OFFSET);
686 val_hi = readl(pg_e + MEMERR_L2C_L2EAHR_PAGE_OFFSET);
687 dev_err(edac_dev->dev,
688 "PMD%d memory error L2C L2ESR 0x%08X @ 0x%08X.%08X\n",
689 ctx->pmd, val, val_hi, val_lo);
690 dev_err(edac_dev->dev,
691 "ErrSyndrome 0x%02X ErrWay 0x%02X ErrCpu %d ErrGroup 0x%02X ErrAction 0x%02X\n",
692 MEMERR_L2C_L2ESR_ERRSYN_RD(val),
693 MEMERR_L2C_L2ESR_ERRWAY_RD(val),
694 MEMERR_L2C_L2ESR_ERRCPU_RD(val),
695 MEMERR_L2C_L2ESR_ERRGROUP_RD(val),
696 MEMERR_L2C_L2ESR_ERRACTION_RD(val));
698 if (val & MEMERR_L2C_L2ESR_ERR_MASK)
699 dev_err(edac_dev->dev, "One or more correctable error\n");
700 if (val & MEMERR_L2C_L2ESR_MULTICERR_MASK)
701 dev_err(edac_dev->dev, "Multiple correctable error\n");
702 if (val & MEMERR_L2C_L2ESR_UCERR_MASK)
703 dev_err(edac_dev->dev, "One or more uncorrectable error\n");
704 if (val & MEMERR_L2C_L2ESR_MULTUCERR_MASK)
705 dev_err(edac_dev->dev, "Multiple uncorrectable error\n");
707 switch (MEMERR_L2C_L2ESR_ERRTYPE_RD(val)) {
708 case 0:
709 dev_err(edac_dev->dev, "Outbound SDB parity error\n");
710 break;
711 case 1:
712 dev_err(edac_dev->dev, "Inbound SDB parity error\n");
713 break;
714 case 2:
715 dev_err(edac_dev->dev, "Tag ECC error\n");
716 break;
717 case 3:
718 dev_err(edac_dev->dev, "Data ECC error\n");
719 break;
722 /* Clear any HW errors */
723 writel(val, pg_e + MEMERR_L2C_L2ESR_PAGE_OFFSET);
725 if (val & (MEMERR_L2C_L2ESR_ERR_MASK |
726 MEMERR_L2C_L2ESR_MULTICERR_MASK))
727 edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
728 if (val & (MEMERR_L2C_L2ESR_UCERR_MASK |
729 MEMERR_L2C_L2ESR_MULTUCERR_MASK))
730 edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
732 chk_l2c:
733 /* Check if any memory request timed out on L2 cache */
734 pg_d = ctx->pmd_csr + CPU_L2C_PAGE;
735 val = readl(pg_d + CPUX_L2C_L2RTOSR_PAGE_OFFSET);
736 if (val) {
737 val_lo = readl(pg_d + CPUX_L2C_L2RTOALR_PAGE_OFFSET);
738 val_hi = readl(pg_d + CPUX_L2C_L2RTOAHR_PAGE_OFFSET);
739 dev_err(edac_dev->dev,
740 "PMD%d L2C error L2C RTOSR 0x%08X @ 0x%08X.%08X\n",
741 ctx->pmd, val, val_hi, val_lo);
742 writel(val, pg_d + CPUX_L2C_L2RTOSR_PAGE_OFFSET);
746 static void xgene_edac_pmd_check(struct edac_device_ctl_info *edac_dev)
748 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
749 unsigned int pcp_hp_stat;
750 int i;
752 xgene_edac_pcp_rd(ctx->edac, PCPHPERRINTSTS, &pcp_hp_stat);
753 if (!((PMD0_MERR_MASK << ctx->pmd) & pcp_hp_stat))
754 return;
756 /* Check CPU L1 error */
757 for (i = 0; i < MAX_CPU_PER_PMD; i++)
758 xgene_edac_pmd_l1_check(edac_dev, i);
760 /* Check CPU L2 error */
761 xgene_edac_pmd_l2_check(edac_dev);
764 static void xgene_edac_pmd_cpu_hw_cfg(struct edac_device_ctl_info *edac_dev,
765 int cpu)
767 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
768 void __iomem *pg_f = ctx->pmd_csr + cpu * CPU_CSR_STRIDE +
769 CPU_MEMERR_CPU_PAGE;
772 * Enable CPU memory error:
773 * MEMERR_CPU_ICFESRA, MEMERR_CPU_LSUESRA, and MEMERR_CPU_MMUESRA
775 writel(0x00000301, pg_f + MEMERR_CPU_ICFECR_PAGE_OFFSET);
776 writel(0x00000301, pg_f + MEMERR_CPU_LSUECR_PAGE_OFFSET);
777 writel(0x00000101, pg_f + MEMERR_CPU_MMUECR_PAGE_OFFSET);
780 static void xgene_edac_pmd_hw_cfg(struct edac_device_ctl_info *edac_dev)
782 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
783 void __iomem *pg_d = ctx->pmd_csr + CPU_L2C_PAGE;
784 void __iomem *pg_e = ctx->pmd_csr + CPU_MEMERR_L2C_PAGE;
786 /* Enable PMD memory error - MEMERR_L2C_L2ECR and L2C_L2RTOCR */
787 writel(0x00000703, pg_e + MEMERR_L2C_L2ECR_PAGE_OFFSET);
788 /* Configure L2C HW request time out feature if supported */
789 if (ctx->version > 1)
790 writel(0x00000119, pg_d + CPUX_L2C_L2RTOCR_PAGE_OFFSET);
793 static void xgene_edac_pmd_hw_ctl(struct edac_device_ctl_info *edac_dev,
794 bool enable)
796 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
797 int i;
799 /* Enable PMD error interrupt */
800 if (edac_dev->op_state == OP_RUNNING_INTERRUPT) {
801 if (enable)
802 xgene_edac_pcp_clrbits(ctx->edac, PCPHPERRINTMSK,
803 PMD0_MERR_MASK << ctx->pmd);
804 else
805 xgene_edac_pcp_setbits(ctx->edac, PCPHPERRINTMSK,
806 PMD0_MERR_MASK << ctx->pmd);
809 if (enable) {
810 xgene_edac_pmd_hw_cfg(edac_dev);
812 /* Two CPUs per a PMD */
813 for (i = 0; i < MAX_CPU_PER_PMD; i++)
814 xgene_edac_pmd_cpu_hw_cfg(edac_dev, i);
818 static ssize_t xgene_edac_pmd_l1_inject_ctrl_write(struct file *file,
819 const char __user *data,
820 size_t count, loff_t *ppos)
822 struct edac_device_ctl_info *edac_dev = file->private_data;
823 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
824 void __iomem *cpux_pg_f;
825 int i;
827 for (i = 0; i < MAX_CPU_PER_PMD; i++) {
828 cpux_pg_f = ctx->pmd_csr + i * CPU_CSR_STRIDE +
829 CPU_MEMERR_CPU_PAGE;
831 writel(MEMERR_CPU_ICFESR_MULTCERR_MASK |
832 MEMERR_CPU_ICFESR_CERR_MASK,
833 cpux_pg_f + MEMERR_CPU_ICFESRA_PAGE_OFFSET);
834 writel(MEMERR_CPU_LSUESR_MULTCERR_MASK |
835 MEMERR_CPU_LSUESR_CERR_MASK,
836 cpux_pg_f + MEMERR_CPU_LSUESRA_PAGE_OFFSET);
837 writel(MEMERR_CPU_MMUESR_MULTCERR_MASK |
838 MEMERR_CPU_MMUESR_CERR_MASK,
839 cpux_pg_f + MEMERR_CPU_MMUESRA_PAGE_OFFSET);
841 return count;
844 static ssize_t xgene_edac_pmd_l2_inject_ctrl_write(struct file *file,
845 const char __user *data,
846 size_t count, loff_t *ppos)
848 struct edac_device_ctl_info *edac_dev = file->private_data;
849 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
850 void __iomem *pg_e = ctx->pmd_csr + CPU_MEMERR_L2C_PAGE;
852 writel(MEMERR_L2C_L2ESR_MULTUCERR_MASK |
853 MEMERR_L2C_L2ESR_MULTICERR_MASK |
854 MEMERR_L2C_L2ESR_UCERR_MASK |
855 MEMERR_L2C_L2ESR_ERR_MASK,
856 pg_e + MEMERR_L2C_L2ESRA_PAGE_OFFSET);
857 return count;
860 static const struct file_operations xgene_edac_pmd_debug_inject_fops[] = {
862 .open = simple_open,
863 .write = xgene_edac_pmd_l1_inject_ctrl_write,
864 .llseek = generic_file_llseek, },
866 .open = simple_open,
867 .write = xgene_edac_pmd_l2_inject_ctrl_write,
868 .llseek = generic_file_llseek, },
872 static void
873 xgene_edac_pmd_create_debugfs_nodes(struct edac_device_ctl_info *edac_dev)
875 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
876 struct dentry *dbgfs_dir;
877 char name[10];
879 if (!IS_ENABLED(CONFIG_EDAC_DEBUG) || !ctx->edac->dfs)
880 return;
882 snprintf(name, sizeof(name), "PMD%d", ctx->pmd);
883 dbgfs_dir = edac_debugfs_create_dir_at(name, ctx->edac->dfs);
884 if (!dbgfs_dir)
885 return;
887 edac_debugfs_create_file("l1_inject_ctrl", S_IWUSR, dbgfs_dir, edac_dev,
888 &xgene_edac_pmd_debug_inject_fops[0]);
889 edac_debugfs_create_file("l2_inject_ctrl", S_IWUSR, dbgfs_dir, edac_dev,
890 &xgene_edac_pmd_debug_inject_fops[1]);
893 static int xgene_edac_pmd_available(u32 efuse, int pmd)
895 return (efuse & (1 << pmd)) ? 0 : 1;
898 static int xgene_edac_pmd_add(struct xgene_edac *edac, struct device_node *np,
899 int version)
901 struct edac_device_ctl_info *edac_dev;
902 struct xgene_edac_pmd_ctx *ctx;
903 struct resource res;
904 char edac_name[10];
905 u32 pmd;
906 int rc;
907 u32 val;
909 if (!devres_open_group(edac->dev, xgene_edac_pmd_add, GFP_KERNEL))
910 return -ENOMEM;
912 /* Determine if this PMD is disabled */
913 if (of_property_read_u32(np, "pmd-controller", &pmd)) {
914 dev_err(edac->dev, "no pmd-controller property\n");
915 rc = -ENODEV;
916 goto err_group;
918 rc = regmap_read(edac->efuse_map, 0, &val);
919 if (rc)
920 goto err_group;
921 if (!xgene_edac_pmd_available(val, pmd)) {
922 rc = -ENODEV;
923 goto err_group;
926 snprintf(edac_name, sizeof(edac_name), "l2c%d", pmd);
927 edac_dev = edac_device_alloc_ctl_info(sizeof(*ctx),
928 edac_name, 1, "l2c", 1, 2, NULL,
929 0, edac_device_alloc_index());
930 if (!edac_dev) {
931 rc = -ENOMEM;
932 goto err_group;
935 ctx = edac_dev->pvt_info;
936 ctx->name = "xgene_pmd_err";
937 ctx->pmd = pmd;
938 ctx->edac = edac;
939 ctx->edac_dev = edac_dev;
940 ctx->ddev = *edac->dev;
941 ctx->version = version;
942 edac_dev->dev = &ctx->ddev;
943 edac_dev->ctl_name = ctx->name;
944 edac_dev->dev_name = ctx->name;
945 edac_dev->mod_name = EDAC_MOD_STR;
947 rc = of_address_to_resource(np, 0, &res);
948 if (rc < 0) {
949 dev_err(edac->dev, "no PMD resource address\n");
950 goto err_free;
952 ctx->pmd_csr = devm_ioremap_resource(edac->dev, &res);
953 if (IS_ERR(ctx->pmd_csr)) {
954 dev_err(edac->dev,
955 "devm_ioremap_resource failed for PMD resource address\n");
956 rc = PTR_ERR(ctx->pmd_csr);
957 goto err_free;
960 if (edac_op_state == EDAC_OPSTATE_POLL)
961 edac_dev->edac_check = xgene_edac_pmd_check;
963 xgene_edac_pmd_create_debugfs_nodes(edac_dev);
965 rc = edac_device_add_device(edac_dev);
966 if (rc > 0) {
967 dev_err(edac->dev, "edac_device_add_device failed\n");
968 rc = -ENOMEM;
969 goto err_free;
972 if (edac_op_state == EDAC_OPSTATE_INT)
973 edac_dev->op_state = OP_RUNNING_INTERRUPT;
975 list_add(&ctx->next, &edac->pmds);
977 xgene_edac_pmd_hw_ctl(edac_dev, 1);
979 devres_remove_group(edac->dev, xgene_edac_pmd_add);
981 dev_info(edac->dev, "X-Gene EDAC PMD%d registered\n", ctx->pmd);
982 return 0;
984 err_free:
985 edac_device_free_ctl_info(edac_dev);
986 err_group:
987 devres_release_group(edac->dev, xgene_edac_pmd_add);
988 return rc;
991 static int xgene_edac_pmd_remove(struct xgene_edac_pmd_ctx *pmd)
993 struct edac_device_ctl_info *edac_dev = pmd->edac_dev;
995 xgene_edac_pmd_hw_ctl(edac_dev, 0);
996 edac_device_del_device(edac_dev->dev);
997 edac_device_free_ctl_info(edac_dev);
998 return 0;
1001 /* L3 Error device */
1002 #define L3C_ESR (0x0A * 4)
1003 #define L3C_ESR_DATATAG_MASK BIT(9)
1004 #define L3C_ESR_MULTIHIT_MASK BIT(8)
1005 #define L3C_ESR_UCEVICT_MASK BIT(6)
1006 #define L3C_ESR_MULTIUCERR_MASK BIT(5)
1007 #define L3C_ESR_MULTICERR_MASK BIT(4)
1008 #define L3C_ESR_UCERR_MASK BIT(3)
1009 #define L3C_ESR_CERR_MASK BIT(2)
1010 #define L3C_ESR_UCERRINTR_MASK BIT(1)
1011 #define L3C_ESR_CERRINTR_MASK BIT(0)
1012 #define L3C_ECR (0x0B * 4)
1013 #define L3C_ECR_UCINTREN BIT(3)
1014 #define L3C_ECR_CINTREN BIT(2)
1015 #define L3C_UCERREN BIT(1)
1016 #define L3C_CERREN BIT(0)
1017 #define L3C_ELR (0x0C * 4)
1018 #define L3C_ELR_ERRSYN(src) ((src & 0xFF800000) >> 23)
1019 #define L3C_ELR_ERRWAY(src) ((src & 0x007E0000) >> 17)
1020 #define L3C_ELR_AGENTID(src) ((src & 0x0001E000) >> 13)
1021 #define L3C_ELR_ERRGRP(src) ((src & 0x00000F00) >> 8)
1022 #define L3C_ELR_OPTYPE(src) ((src & 0x000000F0) >> 4)
1023 #define L3C_ELR_PADDRHIGH(src) (src & 0x0000000F)
1024 #define L3C_AELR (0x0D * 4)
1025 #define L3C_BELR (0x0E * 4)
1026 #define L3C_BELR_BANK(src) (src & 0x0000000F)
1028 struct xgene_edac_dev_ctx {
1029 struct list_head next;
1030 struct device ddev;
1031 char *name;
1032 struct xgene_edac *edac;
1033 struct edac_device_ctl_info *edac_dev;
1034 int edac_idx;
1035 void __iomem *dev_csr;
1036 int version;
1040 * Version 1 of the L3 controller has broken single bit correctable logic for
1041 * certain error syndromes. Log them as uncorrectable in that case.
1043 static bool xgene_edac_l3_promote_to_uc_err(u32 l3cesr, u32 l3celr)
1045 if (l3cesr & L3C_ESR_DATATAG_MASK) {
1046 switch (L3C_ELR_ERRSYN(l3celr)) {
1047 case 0x13C:
1048 case 0x0B4:
1049 case 0x007:
1050 case 0x00D:
1051 case 0x00E:
1052 case 0x019:
1053 case 0x01A:
1054 case 0x01C:
1055 case 0x04E:
1056 case 0x041:
1057 return true;
1059 } else if (L3C_ELR_ERRWAY(l3celr) == 9)
1060 return true;
1062 return false;
1065 static void xgene_edac_l3_check(struct edac_device_ctl_info *edac_dev)
1067 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
1068 u32 l3cesr;
1069 u32 l3celr;
1070 u32 l3caelr;
1071 u32 l3cbelr;
1073 l3cesr = readl(ctx->dev_csr + L3C_ESR);
1074 if (!(l3cesr & (L3C_ESR_UCERR_MASK | L3C_ESR_CERR_MASK)))
1075 return;
1077 if (l3cesr & L3C_ESR_UCERR_MASK)
1078 dev_err(edac_dev->dev, "L3C uncorrectable error\n");
1079 if (l3cesr & L3C_ESR_CERR_MASK)
1080 dev_warn(edac_dev->dev, "L3C correctable error\n");
1082 l3celr = readl(ctx->dev_csr + L3C_ELR);
1083 l3caelr = readl(ctx->dev_csr + L3C_AELR);
1084 l3cbelr = readl(ctx->dev_csr + L3C_BELR);
1085 if (l3cesr & L3C_ESR_MULTIHIT_MASK)
1086 dev_err(edac_dev->dev, "L3C multiple hit error\n");
1087 if (l3cesr & L3C_ESR_UCEVICT_MASK)
1088 dev_err(edac_dev->dev,
1089 "L3C dropped eviction of line with error\n");
1090 if (l3cesr & L3C_ESR_MULTIUCERR_MASK)
1091 dev_err(edac_dev->dev, "L3C multiple uncorrectable error\n");
1092 if (l3cesr & L3C_ESR_DATATAG_MASK)
1093 dev_err(edac_dev->dev,
1094 "L3C data error syndrome 0x%X group 0x%X\n",
1095 L3C_ELR_ERRSYN(l3celr), L3C_ELR_ERRGRP(l3celr));
1096 else
1097 dev_err(edac_dev->dev,
1098 "L3C tag error syndrome 0x%X Way of Tag 0x%X Agent ID 0x%X Operation type 0x%X\n",
1099 L3C_ELR_ERRSYN(l3celr), L3C_ELR_ERRWAY(l3celr),
1100 L3C_ELR_AGENTID(l3celr), L3C_ELR_OPTYPE(l3celr));
1102 * NOTE: Address [41:38] in L3C_ELR_PADDRHIGH(l3celr).
1103 * Address [37:6] in l3caelr. Lower 6 bits are zero.
1105 dev_err(edac_dev->dev, "L3C error address 0x%08X.%08X bank %d\n",
1106 L3C_ELR_PADDRHIGH(l3celr) << 6 | (l3caelr >> 26),
1107 (l3caelr & 0x3FFFFFFF) << 6, L3C_BELR_BANK(l3cbelr));
1108 dev_err(edac_dev->dev,
1109 "L3C error status register value 0x%X\n", l3cesr);
1111 /* Clear L3C error interrupt */
1112 writel(0, ctx->dev_csr + L3C_ESR);
1114 if (ctx->version <= 1 &&
1115 xgene_edac_l3_promote_to_uc_err(l3cesr, l3celr)) {
1116 edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
1117 return;
1119 if (l3cesr & L3C_ESR_CERR_MASK)
1120 edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
1121 if (l3cesr & L3C_ESR_UCERR_MASK)
1122 edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
1125 static void xgene_edac_l3_hw_init(struct edac_device_ctl_info *edac_dev,
1126 bool enable)
1128 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
1129 u32 val;
1131 val = readl(ctx->dev_csr + L3C_ECR);
1132 val |= L3C_UCERREN | L3C_CERREN;
1133 /* On disable, we just disable interrupt but keep error enabled */
1134 if (edac_dev->op_state == OP_RUNNING_INTERRUPT) {
1135 if (enable)
1136 val |= L3C_ECR_UCINTREN | L3C_ECR_CINTREN;
1137 else
1138 val &= ~(L3C_ECR_UCINTREN | L3C_ECR_CINTREN);
1140 writel(val, ctx->dev_csr + L3C_ECR);
1142 if (edac_dev->op_state == OP_RUNNING_INTERRUPT) {
1143 /* Enable/disable L3 error top level interrupt */
1144 if (enable) {
1145 xgene_edac_pcp_clrbits(ctx->edac, PCPHPERRINTMSK,
1146 L3C_UNCORR_ERR_MASK);
1147 xgene_edac_pcp_clrbits(ctx->edac, PCPLPERRINTMSK,
1148 L3C_CORR_ERR_MASK);
1149 } else {
1150 xgene_edac_pcp_setbits(ctx->edac, PCPHPERRINTMSK,
1151 L3C_UNCORR_ERR_MASK);
1152 xgene_edac_pcp_setbits(ctx->edac, PCPLPERRINTMSK,
1153 L3C_CORR_ERR_MASK);
1158 static ssize_t xgene_edac_l3_inject_ctrl_write(struct file *file,
1159 const char __user *data,
1160 size_t count, loff_t *ppos)
1162 struct edac_device_ctl_info *edac_dev = file->private_data;
1163 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
1165 /* Generate all errors */
1166 writel(0xFFFFFFFF, ctx->dev_csr + L3C_ESR);
1167 return count;
1170 static const struct file_operations xgene_edac_l3_debug_inject_fops = {
1171 .open = simple_open,
1172 .write = xgene_edac_l3_inject_ctrl_write,
1173 .llseek = generic_file_llseek
1176 static void
1177 xgene_edac_l3_create_debugfs_nodes(struct edac_device_ctl_info *edac_dev)
1179 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
1180 struct dentry *dbgfs_dir;
1181 char name[10];
1183 if (!IS_ENABLED(CONFIG_EDAC_DEBUG) || !ctx->edac->dfs)
1184 return;
1186 snprintf(name, sizeof(name), "l3c%d", ctx->edac_idx);
1187 dbgfs_dir = edac_debugfs_create_dir_at(name, ctx->edac->dfs);
1188 if (!dbgfs_dir)
1189 return;
1191 debugfs_create_file("l3_inject_ctrl", S_IWUSR, dbgfs_dir, edac_dev,
1192 &xgene_edac_l3_debug_inject_fops);
1195 static int xgene_edac_l3_add(struct xgene_edac *edac, struct device_node *np,
1196 int version)
1198 struct edac_device_ctl_info *edac_dev;
1199 struct xgene_edac_dev_ctx *ctx;
1200 struct resource res;
1201 void __iomem *dev_csr;
1202 int edac_idx;
1203 int rc = 0;
1205 if (!devres_open_group(edac->dev, xgene_edac_l3_add, GFP_KERNEL))
1206 return -ENOMEM;
1208 rc = of_address_to_resource(np, 0, &res);
1209 if (rc < 0) {
1210 dev_err(edac->dev, "no L3 resource address\n");
1211 goto err_release_group;
1213 dev_csr = devm_ioremap_resource(edac->dev, &res);
1214 if (IS_ERR(dev_csr)) {
1215 dev_err(edac->dev,
1216 "devm_ioremap_resource failed for L3 resource address\n");
1217 rc = PTR_ERR(dev_csr);
1218 goto err_release_group;
1221 edac_idx = edac_device_alloc_index();
1222 edac_dev = edac_device_alloc_ctl_info(sizeof(*ctx),
1223 "l3c", 1, "l3c", 1, 0, NULL, 0,
1224 edac_idx);
1225 if (!edac_dev) {
1226 rc = -ENOMEM;
1227 goto err_release_group;
1230 ctx = edac_dev->pvt_info;
1231 ctx->dev_csr = dev_csr;
1232 ctx->name = "xgene_l3_err";
1233 ctx->edac_idx = edac_idx;
1234 ctx->edac = edac;
1235 ctx->edac_dev = edac_dev;
1236 ctx->ddev = *edac->dev;
1237 ctx->version = version;
1238 edac_dev->dev = &ctx->ddev;
1239 edac_dev->ctl_name = ctx->name;
1240 edac_dev->dev_name = ctx->name;
1241 edac_dev->mod_name = EDAC_MOD_STR;
1243 if (edac_op_state == EDAC_OPSTATE_POLL)
1244 edac_dev->edac_check = xgene_edac_l3_check;
1246 xgene_edac_l3_create_debugfs_nodes(edac_dev);
1248 rc = edac_device_add_device(edac_dev);
1249 if (rc > 0) {
1250 dev_err(edac->dev, "failed edac_device_add_device()\n");
1251 rc = -ENOMEM;
1252 goto err_ctl_free;
1255 if (edac_op_state == EDAC_OPSTATE_INT)
1256 edac_dev->op_state = OP_RUNNING_INTERRUPT;
1258 list_add(&ctx->next, &edac->l3s);
1260 xgene_edac_l3_hw_init(edac_dev, 1);
1262 devres_remove_group(edac->dev, xgene_edac_l3_add);
1264 dev_info(edac->dev, "X-Gene EDAC L3 registered\n");
1265 return 0;
1267 err_ctl_free:
1268 edac_device_free_ctl_info(edac_dev);
1269 err_release_group:
1270 devres_release_group(edac->dev, xgene_edac_l3_add);
1271 return rc;
1274 static int xgene_edac_l3_remove(struct xgene_edac_dev_ctx *l3)
1276 struct edac_device_ctl_info *edac_dev = l3->edac_dev;
1278 xgene_edac_l3_hw_init(edac_dev, 0);
1279 edac_device_del_device(l3->edac->dev);
1280 edac_device_free_ctl_info(edac_dev);
1281 return 0;
1284 /* SoC error device */
1285 #define IOBAXIS0TRANSERRINTSTS 0x0000
1286 #define IOBAXIS0_M_ILLEGAL_ACCESS_MASK BIT(1)
1287 #define IOBAXIS0_ILLEGAL_ACCESS_MASK BIT(0)
1288 #define IOBAXIS0TRANSERRINTMSK 0x0004
1289 #define IOBAXIS0TRANSERRREQINFOL 0x0008
1290 #define IOBAXIS0TRANSERRREQINFOH 0x000c
1291 #define REQTYPE_RD(src) (((src) & BIT(0)))
1292 #define ERRADDRH_RD(src) (((src) & 0xffc00000) >> 22)
1293 #define IOBAXIS1TRANSERRINTSTS 0x0010
1294 #define IOBAXIS1TRANSERRINTMSK 0x0014
1295 #define IOBAXIS1TRANSERRREQINFOL 0x0018
1296 #define IOBAXIS1TRANSERRREQINFOH 0x001c
1297 #define IOBPATRANSERRINTSTS 0x0020
1298 #define IOBPA_M_REQIDRAM_CORRUPT_MASK BIT(7)
1299 #define IOBPA_REQIDRAM_CORRUPT_MASK BIT(6)
1300 #define IOBPA_M_TRANS_CORRUPT_MASK BIT(5)
1301 #define IOBPA_TRANS_CORRUPT_MASK BIT(4)
1302 #define IOBPA_M_WDATA_CORRUPT_MASK BIT(3)
1303 #define IOBPA_WDATA_CORRUPT_MASK BIT(2)
1304 #define IOBPA_M_RDATA_CORRUPT_MASK BIT(1)
1305 #define IOBPA_RDATA_CORRUPT_MASK BIT(0)
1306 #define IOBBATRANSERRINTSTS 0x0030
1307 #define M_ILLEGAL_ACCESS_MASK BIT(15)
1308 #define ILLEGAL_ACCESS_MASK BIT(14)
1309 #define M_WIDRAM_CORRUPT_MASK BIT(13)
1310 #define WIDRAM_CORRUPT_MASK BIT(12)
1311 #define M_RIDRAM_CORRUPT_MASK BIT(11)
1312 #define RIDRAM_CORRUPT_MASK BIT(10)
1313 #define M_TRANS_CORRUPT_MASK BIT(9)
1314 #define TRANS_CORRUPT_MASK BIT(8)
1315 #define M_WDATA_CORRUPT_MASK BIT(7)
1316 #define WDATA_CORRUPT_MASK BIT(6)
1317 #define M_RBM_POISONED_REQ_MASK BIT(5)
1318 #define RBM_POISONED_REQ_MASK BIT(4)
1319 #define M_XGIC_POISONED_REQ_MASK BIT(3)
1320 #define XGIC_POISONED_REQ_MASK BIT(2)
1321 #define M_WRERR_RESP_MASK BIT(1)
1322 #define WRERR_RESP_MASK BIT(0)
1323 #define IOBBATRANSERRREQINFOL 0x0038
1324 #define IOBBATRANSERRREQINFOH 0x003c
1325 #define REQTYPE_F2_RD(src) ((src) & BIT(0))
1326 #define ERRADDRH_F2_RD(src) (((src) & 0xffc00000) >> 22)
1327 #define IOBBATRANSERRCSWREQID 0x0040
1328 #define XGICTRANSERRINTSTS 0x0050
1329 #define M_WR_ACCESS_ERR_MASK BIT(3)
1330 #define WR_ACCESS_ERR_MASK BIT(2)
1331 #define M_RD_ACCESS_ERR_MASK BIT(1)
1332 #define RD_ACCESS_ERR_MASK BIT(0)
1333 #define XGICTRANSERRINTMSK 0x0054
1334 #define XGICTRANSERRREQINFO 0x0058
1335 #define REQTYPE_MASK BIT(26)
1336 #define ERRADDR_RD(src) ((src) & 0x03ffffff)
1337 #define GLBL_ERR_STS 0x0800
1338 #define MDED_ERR_MASK BIT(3)
1339 #define DED_ERR_MASK BIT(2)
1340 #define MSEC_ERR_MASK BIT(1)
1341 #define SEC_ERR_MASK BIT(0)
1342 #define GLBL_SEC_ERRL 0x0810
1343 #define GLBL_SEC_ERRH 0x0818
1344 #define GLBL_MSEC_ERRL 0x0820
1345 #define GLBL_MSEC_ERRH 0x0828
1346 #define GLBL_DED_ERRL 0x0830
1347 #define GLBL_DED_ERRLMASK 0x0834
1348 #define GLBL_DED_ERRH 0x0838
1349 #define GLBL_DED_ERRHMASK 0x083c
1350 #define GLBL_MDED_ERRL 0x0840
1351 #define GLBL_MDED_ERRLMASK 0x0844
1352 #define GLBL_MDED_ERRH 0x0848
1353 #define GLBL_MDED_ERRHMASK 0x084c
1355 /* IO Bus Registers */
1356 #define RBCSR 0x0000
1357 #define STICKYERR_MASK BIT(0)
1358 #define RBEIR 0x0008
1359 #define AGENT_OFFLINE_ERR_MASK BIT(30)
1360 #define UNIMPL_RBPAGE_ERR_MASK BIT(29)
1361 #define WORD_ALIGNED_ERR_MASK BIT(28)
1362 #define PAGE_ACCESS_ERR_MASK BIT(27)
1363 #define WRITE_ACCESS_MASK BIT(26)
1364 #define RBERRADDR_RD(src) ((src) & 0x03FFFFFF)
1366 static const char * const soc_mem_err_v1[] = {
1367 "10GbE0",
1368 "10GbE1",
1369 "Security",
1370 "SATA45",
1371 "SATA23/ETH23",
1372 "SATA01/ETH01",
1373 "USB1",
1374 "USB0",
1375 "QML",
1376 "QM0",
1377 "QM1 (XGbE01)",
1378 "PCIE4",
1379 "PCIE3",
1380 "PCIE2",
1381 "PCIE1",
1382 "PCIE0",
1383 "CTX Manager",
1384 "OCM",
1385 "1GbE",
1386 "CLE",
1387 "AHBC",
1388 "PktDMA",
1389 "GFC",
1390 "MSLIM",
1391 "10GbE2",
1392 "10GbE3",
1393 "QM2 (XGbE23)",
1394 "IOB",
1395 "unknown",
1396 "unknown",
1397 "unknown",
1398 "unknown",
1401 static void xgene_edac_iob_gic_report(struct edac_device_ctl_info *edac_dev)
1403 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
1404 u32 err_addr_lo;
1405 u32 err_addr_hi;
1406 u32 reg;
1407 u32 info;
1409 /* GIC transaction error interrupt */
1410 reg = readl(ctx->dev_csr + XGICTRANSERRINTSTS);
1411 if (!reg)
1412 goto chk_iob_err;
1413 dev_err(edac_dev->dev, "XGIC transaction error\n");
1414 if (reg & RD_ACCESS_ERR_MASK)
1415 dev_err(edac_dev->dev, "XGIC read size error\n");
1416 if (reg & M_RD_ACCESS_ERR_MASK)
1417 dev_err(edac_dev->dev, "Multiple XGIC read size error\n");
1418 if (reg & WR_ACCESS_ERR_MASK)
1419 dev_err(edac_dev->dev, "XGIC write size error\n");
1420 if (reg & M_WR_ACCESS_ERR_MASK)
1421 dev_err(edac_dev->dev, "Multiple XGIC write size error\n");
1422 info = readl(ctx->dev_csr + XGICTRANSERRREQINFO);
1423 dev_err(edac_dev->dev, "XGIC %s access @ 0x%08X (0x%08X)\n",
1424 info & REQTYPE_MASK ? "read" : "write", ERRADDR_RD(info),
1425 info);
1426 writel(reg, ctx->dev_csr + XGICTRANSERRINTSTS);
1428 chk_iob_err:
1429 /* IOB memory error */
1430 reg = readl(ctx->dev_csr + GLBL_ERR_STS);
1431 if (!reg)
1432 return;
1433 if (reg & SEC_ERR_MASK) {
1434 err_addr_lo = readl(ctx->dev_csr + GLBL_SEC_ERRL);
1435 err_addr_hi = readl(ctx->dev_csr + GLBL_SEC_ERRH);
1436 dev_err(edac_dev->dev,
1437 "IOB single-bit correctable memory at 0x%08X.%08X error\n",
1438 err_addr_lo, err_addr_hi);
1439 writel(err_addr_lo, ctx->dev_csr + GLBL_SEC_ERRL);
1440 writel(err_addr_hi, ctx->dev_csr + GLBL_SEC_ERRH);
1442 if (reg & MSEC_ERR_MASK) {
1443 err_addr_lo = readl(ctx->dev_csr + GLBL_MSEC_ERRL);
1444 err_addr_hi = readl(ctx->dev_csr + GLBL_MSEC_ERRH);
1445 dev_err(edac_dev->dev,
1446 "IOB multiple single-bit correctable memory at 0x%08X.%08X error\n",
1447 err_addr_lo, err_addr_hi);
1448 writel(err_addr_lo, ctx->dev_csr + GLBL_MSEC_ERRL);
1449 writel(err_addr_hi, ctx->dev_csr + GLBL_MSEC_ERRH);
1451 if (reg & (SEC_ERR_MASK | MSEC_ERR_MASK))
1452 edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
1454 if (reg & DED_ERR_MASK) {
1455 err_addr_lo = readl(ctx->dev_csr + GLBL_DED_ERRL);
1456 err_addr_hi = readl(ctx->dev_csr + GLBL_DED_ERRH);
1457 dev_err(edac_dev->dev,
1458 "IOB double-bit uncorrectable memory at 0x%08X.%08X error\n",
1459 err_addr_lo, err_addr_hi);
1460 writel(err_addr_lo, ctx->dev_csr + GLBL_DED_ERRL);
1461 writel(err_addr_hi, ctx->dev_csr + GLBL_DED_ERRH);
1463 if (reg & MDED_ERR_MASK) {
1464 err_addr_lo = readl(ctx->dev_csr + GLBL_MDED_ERRL);
1465 err_addr_hi = readl(ctx->dev_csr + GLBL_MDED_ERRH);
1466 dev_err(edac_dev->dev,
1467 "Multiple IOB double-bit uncorrectable memory at 0x%08X.%08X error\n",
1468 err_addr_lo, err_addr_hi);
1469 writel(err_addr_lo, ctx->dev_csr + GLBL_MDED_ERRL);
1470 writel(err_addr_hi, ctx->dev_csr + GLBL_MDED_ERRH);
1472 if (reg & (DED_ERR_MASK | MDED_ERR_MASK))
1473 edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
1476 static void xgene_edac_rb_report(struct edac_device_ctl_info *edac_dev)
1478 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
1479 u32 err_addr_lo;
1480 u32 err_addr_hi;
1481 u32 reg;
1483 /* If the register bus resource isn't available, just skip it */
1484 if (!ctx->edac->rb_map)
1485 goto rb_skip;
1488 * Check RB access errors
1489 * 1. Out of range
1490 * 2. Un-implemented page
1491 * 3. Un-aligned access
1492 * 4. Offline slave IP
1494 if (regmap_read(ctx->edac->rb_map, RBCSR, &reg))
1495 return;
1496 if (reg & STICKYERR_MASK) {
1497 bool write;
1498 u32 address;
1500 dev_err(edac_dev->dev, "IOB bus access error(s)\n");
1501 if (regmap_read(ctx->edac->rb_map, RBEIR, &reg))
1502 return;
1503 write = reg & WRITE_ACCESS_MASK ? 1 : 0;
1504 address = RBERRADDR_RD(reg);
1505 if (reg & AGENT_OFFLINE_ERR_MASK)
1506 dev_err(edac_dev->dev,
1507 "IOB bus %s access to offline agent error\n",
1508 write ? "write" : "read");
1509 if (reg & UNIMPL_RBPAGE_ERR_MASK)
1510 dev_err(edac_dev->dev,
1511 "IOB bus %s access to unimplemented page error\n",
1512 write ? "write" : "read");
1513 if (reg & WORD_ALIGNED_ERR_MASK)
1514 dev_err(edac_dev->dev,
1515 "IOB bus %s word aligned access error\n",
1516 write ? "write" : "read");
1517 if (reg & PAGE_ACCESS_ERR_MASK)
1518 dev_err(edac_dev->dev,
1519 "IOB bus %s to page out of range access error\n",
1520 write ? "write" : "read");
1521 if (regmap_write(ctx->edac->rb_map, RBEIR, 0))
1522 return;
1523 if (regmap_write(ctx->edac->rb_map, RBCSR, 0))
1524 return;
1526 rb_skip:
1528 /* IOB Bridge agent transaction error interrupt */
1529 reg = readl(ctx->dev_csr + IOBBATRANSERRINTSTS);
1530 if (!reg)
1531 return;
1533 dev_err(edac_dev->dev, "IOB bridge agent (BA) transaction error\n");
1534 if (reg & WRERR_RESP_MASK)
1535 dev_err(edac_dev->dev, "IOB BA write response error\n");
1536 if (reg & M_WRERR_RESP_MASK)
1537 dev_err(edac_dev->dev,
1538 "Multiple IOB BA write response error\n");
1539 if (reg & XGIC_POISONED_REQ_MASK)
1540 dev_err(edac_dev->dev, "IOB BA XGIC poisoned write error\n");
1541 if (reg & M_XGIC_POISONED_REQ_MASK)
1542 dev_err(edac_dev->dev,
1543 "Multiple IOB BA XGIC poisoned write error\n");
1544 if (reg & RBM_POISONED_REQ_MASK)
1545 dev_err(edac_dev->dev, "IOB BA RBM poisoned write error\n");
1546 if (reg & M_RBM_POISONED_REQ_MASK)
1547 dev_err(edac_dev->dev,
1548 "Multiple IOB BA RBM poisoned write error\n");
1549 if (reg & WDATA_CORRUPT_MASK)
1550 dev_err(edac_dev->dev, "IOB BA write error\n");
1551 if (reg & M_WDATA_CORRUPT_MASK)
1552 dev_err(edac_dev->dev, "Multiple IOB BA write error\n");
1553 if (reg & TRANS_CORRUPT_MASK)
1554 dev_err(edac_dev->dev, "IOB BA transaction error\n");
1555 if (reg & M_TRANS_CORRUPT_MASK)
1556 dev_err(edac_dev->dev, "Multiple IOB BA transaction error\n");
1557 if (reg & RIDRAM_CORRUPT_MASK)
1558 dev_err(edac_dev->dev,
1559 "IOB BA RDIDRAM read transaction ID error\n");
1560 if (reg & M_RIDRAM_CORRUPT_MASK)
1561 dev_err(edac_dev->dev,
1562 "Multiple IOB BA RDIDRAM read transaction ID error\n");
1563 if (reg & WIDRAM_CORRUPT_MASK)
1564 dev_err(edac_dev->dev,
1565 "IOB BA RDIDRAM write transaction ID error\n");
1566 if (reg & M_WIDRAM_CORRUPT_MASK)
1567 dev_err(edac_dev->dev,
1568 "Multiple IOB BA RDIDRAM write transaction ID error\n");
1569 if (reg & ILLEGAL_ACCESS_MASK)
1570 dev_err(edac_dev->dev,
1571 "IOB BA XGIC/RB illegal access error\n");
1572 if (reg & M_ILLEGAL_ACCESS_MASK)
1573 dev_err(edac_dev->dev,
1574 "Multiple IOB BA XGIC/RB illegal access error\n");
1576 err_addr_lo = readl(ctx->dev_csr + IOBBATRANSERRREQINFOL);
1577 err_addr_hi = readl(ctx->dev_csr + IOBBATRANSERRREQINFOH);
1578 dev_err(edac_dev->dev, "IOB BA %s access at 0x%02X.%08X (0x%08X)\n",
1579 REQTYPE_F2_RD(err_addr_hi) ? "read" : "write",
1580 ERRADDRH_F2_RD(err_addr_hi), err_addr_lo, err_addr_hi);
1581 if (reg & WRERR_RESP_MASK)
1582 dev_err(edac_dev->dev, "IOB BA requestor ID 0x%08X\n",
1583 readl(ctx->dev_csr + IOBBATRANSERRCSWREQID));
1584 writel(reg, ctx->dev_csr + IOBBATRANSERRINTSTS);
1587 static void xgene_edac_pa_report(struct edac_device_ctl_info *edac_dev)
1589 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
1590 u32 err_addr_lo;
1591 u32 err_addr_hi;
1592 u32 reg;
1594 /* IOB Processing agent transaction error interrupt */
1595 reg = readl(ctx->dev_csr + IOBPATRANSERRINTSTS);
1596 if (!reg)
1597 goto chk_iob_axi0;
1598 dev_err(edac_dev->dev, "IOB processing agent (PA) transaction error\n");
1599 if (reg & IOBPA_RDATA_CORRUPT_MASK)
1600 dev_err(edac_dev->dev, "IOB PA read data RAM error\n");
1601 if (reg & IOBPA_M_RDATA_CORRUPT_MASK)
1602 dev_err(edac_dev->dev,
1603 "Multiple IOB PA read data RAM error\n");
1604 if (reg & IOBPA_WDATA_CORRUPT_MASK)
1605 dev_err(edac_dev->dev, "IOB PA write data RAM error\n");
1606 if (reg & IOBPA_M_WDATA_CORRUPT_MASK)
1607 dev_err(edac_dev->dev,
1608 "Multiple IOB PA write data RAM error\n");
1609 if (reg & IOBPA_TRANS_CORRUPT_MASK)
1610 dev_err(edac_dev->dev, "IOB PA transaction error\n");
1611 if (reg & IOBPA_M_TRANS_CORRUPT_MASK)
1612 dev_err(edac_dev->dev, "Multiple IOB PA transaction error\n");
1613 if (reg & IOBPA_REQIDRAM_CORRUPT_MASK)
1614 dev_err(edac_dev->dev, "IOB PA transaction ID RAM error\n");
1615 if (reg & IOBPA_M_REQIDRAM_CORRUPT_MASK)
1616 dev_err(edac_dev->dev,
1617 "Multiple IOB PA transaction ID RAM error\n");
1618 writel(reg, ctx->dev_csr + IOBPATRANSERRINTSTS);
1620 chk_iob_axi0:
1621 /* IOB AXI0 Error */
1622 reg = readl(ctx->dev_csr + IOBAXIS0TRANSERRINTSTS);
1623 if (!reg)
1624 goto chk_iob_axi1;
1625 err_addr_lo = readl(ctx->dev_csr + IOBAXIS0TRANSERRREQINFOL);
1626 err_addr_hi = readl(ctx->dev_csr + IOBAXIS0TRANSERRREQINFOH);
1627 dev_err(edac_dev->dev,
1628 "%sAXI slave 0 illegal %s access @ 0x%02X.%08X (0x%08X)\n",
1629 reg & IOBAXIS0_M_ILLEGAL_ACCESS_MASK ? "Multiple " : "",
1630 REQTYPE_RD(err_addr_hi) ? "read" : "write",
1631 ERRADDRH_RD(err_addr_hi), err_addr_lo, err_addr_hi);
1632 writel(reg, ctx->dev_csr + IOBAXIS0TRANSERRINTSTS);
1634 chk_iob_axi1:
1635 /* IOB AXI1 Error */
1636 reg = readl(ctx->dev_csr + IOBAXIS1TRANSERRINTSTS);
1637 if (!reg)
1638 return;
1639 err_addr_lo = readl(ctx->dev_csr + IOBAXIS1TRANSERRREQINFOL);
1640 err_addr_hi = readl(ctx->dev_csr + IOBAXIS1TRANSERRREQINFOH);
1641 dev_err(edac_dev->dev,
1642 "%sAXI slave 1 illegal %s access @ 0x%02X.%08X (0x%08X)\n",
1643 reg & IOBAXIS0_M_ILLEGAL_ACCESS_MASK ? "Multiple " : "",
1644 REQTYPE_RD(err_addr_hi) ? "read" : "write",
1645 ERRADDRH_RD(err_addr_hi), err_addr_lo, err_addr_hi);
1646 writel(reg, ctx->dev_csr + IOBAXIS1TRANSERRINTSTS);
1649 static void xgene_edac_soc_check(struct edac_device_ctl_info *edac_dev)
1651 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
1652 const char * const *soc_mem_err = NULL;
1653 u32 pcp_hp_stat;
1654 u32 pcp_lp_stat;
1655 u32 reg;
1656 int i;
1658 xgene_edac_pcp_rd(ctx->edac, PCPHPERRINTSTS, &pcp_hp_stat);
1659 xgene_edac_pcp_rd(ctx->edac, PCPLPERRINTSTS, &pcp_lp_stat);
1660 xgene_edac_pcp_rd(ctx->edac, MEMERRINTSTS, &reg);
1661 if (!((pcp_hp_stat & (IOB_PA_ERR_MASK | IOB_BA_ERR_MASK |
1662 IOB_XGIC_ERR_MASK | IOB_RB_ERR_MASK)) ||
1663 (pcp_lp_stat & CSW_SWITCH_TRACE_ERR_MASK) || reg))
1664 return;
1666 if (pcp_hp_stat & IOB_XGIC_ERR_MASK)
1667 xgene_edac_iob_gic_report(edac_dev);
1669 if (pcp_hp_stat & (IOB_RB_ERR_MASK | IOB_BA_ERR_MASK))
1670 xgene_edac_rb_report(edac_dev);
1672 if (pcp_hp_stat & IOB_PA_ERR_MASK)
1673 xgene_edac_pa_report(edac_dev);
1675 if (pcp_lp_stat & CSW_SWITCH_TRACE_ERR_MASK) {
1676 dev_info(edac_dev->dev,
1677 "CSW switch trace correctable memory parity error\n");
1678 edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
1681 if (!reg)
1682 return;
1683 if (ctx->version == 1)
1684 soc_mem_err = soc_mem_err_v1;
1685 if (!soc_mem_err) {
1686 dev_err(edac_dev->dev, "SoC memory parity error 0x%08X\n",
1687 reg);
1688 edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
1689 return;
1691 for (i = 0; i < 31; i++) {
1692 if (reg & (1 << i)) {
1693 dev_err(edac_dev->dev, "%s memory parity error\n",
1694 soc_mem_err[i]);
1695 edac_device_handle_ue(edac_dev, 0, 0,
1696 edac_dev->ctl_name);
1701 static void xgene_edac_soc_hw_init(struct edac_device_ctl_info *edac_dev,
1702 bool enable)
1704 struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
1706 /* Enable SoC IP error interrupt */
1707 if (edac_dev->op_state == OP_RUNNING_INTERRUPT) {
1708 if (enable) {
1709 xgene_edac_pcp_clrbits(ctx->edac, PCPHPERRINTMSK,
1710 IOB_PA_ERR_MASK |
1711 IOB_BA_ERR_MASK |
1712 IOB_XGIC_ERR_MASK |
1713 IOB_RB_ERR_MASK);
1714 xgene_edac_pcp_clrbits(ctx->edac, PCPLPERRINTMSK,
1715 CSW_SWITCH_TRACE_ERR_MASK);
1716 } else {
1717 xgene_edac_pcp_setbits(ctx->edac, PCPHPERRINTMSK,
1718 IOB_PA_ERR_MASK |
1719 IOB_BA_ERR_MASK |
1720 IOB_XGIC_ERR_MASK |
1721 IOB_RB_ERR_MASK);
1722 xgene_edac_pcp_setbits(ctx->edac, PCPLPERRINTMSK,
1723 CSW_SWITCH_TRACE_ERR_MASK);
1726 writel(enable ? 0x0 : 0xFFFFFFFF,
1727 ctx->dev_csr + IOBAXIS0TRANSERRINTMSK);
1728 writel(enable ? 0x0 : 0xFFFFFFFF,
1729 ctx->dev_csr + IOBAXIS1TRANSERRINTMSK);
1730 writel(enable ? 0x0 : 0xFFFFFFFF,
1731 ctx->dev_csr + XGICTRANSERRINTMSK);
1733 xgene_edac_pcp_setbits(ctx->edac, MEMERRINTMSK,
1734 enable ? 0x0 : 0xFFFFFFFF);
1738 static int xgene_edac_soc_add(struct xgene_edac *edac, struct device_node *np,
1739 int version)
1741 struct edac_device_ctl_info *edac_dev;
1742 struct xgene_edac_dev_ctx *ctx;
1743 void __iomem *dev_csr;
1744 struct resource res;
1745 int edac_idx;
1746 int rc;
1748 if (!devres_open_group(edac->dev, xgene_edac_soc_add, GFP_KERNEL))
1749 return -ENOMEM;
1751 rc = of_address_to_resource(np, 0, &res);
1752 if (rc < 0) {
1753 dev_err(edac->dev, "no SoC resource address\n");
1754 goto err_release_group;
1756 dev_csr = devm_ioremap_resource(edac->dev, &res);
1757 if (IS_ERR(dev_csr)) {
1758 dev_err(edac->dev,
1759 "devm_ioremap_resource failed for soc resource address\n");
1760 rc = PTR_ERR(dev_csr);
1761 goto err_release_group;
1764 edac_idx = edac_device_alloc_index();
1765 edac_dev = edac_device_alloc_ctl_info(sizeof(*ctx),
1766 "SOC", 1, "SOC", 1, 2, NULL, 0,
1767 edac_idx);
1768 if (!edac_dev) {
1769 rc = -ENOMEM;
1770 goto err_release_group;
1773 ctx = edac_dev->pvt_info;
1774 ctx->dev_csr = dev_csr;
1775 ctx->name = "xgene_soc_err";
1776 ctx->edac_idx = edac_idx;
1777 ctx->edac = edac;
1778 ctx->edac_dev = edac_dev;
1779 ctx->ddev = *edac->dev;
1780 ctx->version = version;
1781 edac_dev->dev = &ctx->ddev;
1782 edac_dev->ctl_name = ctx->name;
1783 edac_dev->dev_name = ctx->name;
1784 edac_dev->mod_name = EDAC_MOD_STR;
1786 if (edac_op_state == EDAC_OPSTATE_POLL)
1787 edac_dev->edac_check = xgene_edac_soc_check;
1789 rc = edac_device_add_device(edac_dev);
1790 if (rc > 0) {
1791 dev_err(edac->dev, "failed edac_device_add_device()\n");
1792 rc = -ENOMEM;
1793 goto err_ctl_free;
1796 if (edac_op_state == EDAC_OPSTATE_INT)
1797 edac_dev->op_state = OP_RUNNING_INTERRUPT;
1799 list_add(&ctx->next, &edac->socs);
1801 xgene_edac_soc_hw_init(edac_dev, 1);
1803 devres_remove_group(edac->dev, xgene_edac_soc_add);
1805 dev_info(edac->dev, "X-Gene EDAC SoC registered\n");
1807 return 0;
1809 err_ctl_free:
1810 edac_device_free_ctl_info(edac_dev);
1811 err_release_group:
1812 devres_release_group(edac->dev, xgene_edac_soc_add);
1813 return rc;
1816 static int xgene_edac_soc_remove(struct xgene_edac_dev_ctx *soc)
1818 struct edac_device_ctl_info *edac_dev = soc->edac_dev;
1820 xgene_edac_soc_hw_init(edac_dev, 0);
1821 edac_device_del_device(soc->edac->dev);
1822 edac_device_free_ctl_info(edac_dev);
1823 return 0;
1826 static irqreturn_t xgene_edac_isr(int irq, void *dev_id)
1828 struct xgene_edac *ctx = dev_id;
1829 struct xgene_edac_pmd_ctx *pmd;
1830 struct xgene_edac_dev_ctx *node;
1831 unsigned int pcp_hp_stat;
1832 unsigned int pcp_lp_stat;
1834 xgene_edac_pcp_rd(ctx, PCPHPERRINTSTS, &pcp_hp_stat);
1835 xgene_edac_pcp_rd(ctx, PCPLPERRINTSTS, &pcp_lp_stat);
1836 if ((MCU_UNCORR_ERR_MASK & pcp_hp_stat) ||
1837 (MCU_CTL_ERR_MASK & pcp_hp_stat) ||
1838 (MCU_CORR_ERR_MASK & pcp_lp_stat)) {
1839 struct xgene_edac_mc_ctx *mcu;
1841 list_for_each_entry(mcu, &ctx->mcus, next)
1842 xgene_edac_mc_check(mcu->mci);
1845 list_for_each_entry(pmd, &ctx->pmds, next) {
1846 if ((PMD0_MERR_MASK << pmd->pmd) & pcp_hp_stat)
1847 xgene_edac_pmd_check(pmd->edac_dev);
1850 list_for_each_entry(node, &ctx->l3s, next)
1851 xgene_edac_l3_check(node->edac_dev);
1853 list_for_each_entry(node, &ctx->socs, next)
1854 xgene_edac_soc_check(node->edac_dev);
1856 return IRQ_HANDLED;
1859 static int xgene_edac_probe(struct platform_device *pdev)
1861 struct xgene_edac *edac;
1862 struct device_node *child;
1863 struct resource *res;
1864 int rc;
1866 edac = devm_kzalloc(&pdev->dev, sizeof(*edac), GFP_KERNEL);
1867 if (!edac)
1868 return -ENOMEM;
1870 edac->dev = &pdev->dev;
1871 platform_set_drvdata(pdev, edac);
1872 INIT_LIST_HEAD(&edac->mcus);
1873 INIT_LIST_HEAD(&edac->pmds);
1874 INIT_LIST_HEAD(&edac->l3s);
1875 INIT_LIST_HEAD(&edac->socs);
1876 spin_lock_init(&edac->lock);
1877 mutex_init(&edac->mc_lock);
1879 edac->csw_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
1880 "regmap-csw");
1881 if (IS_ERR(edac->csw_map)) {
1882 dev_err(edac->dev, "unable to get syscon regmap csw\n");
1883 rc = PTR_ERR(edac->csw_map);
1884 goto out_err;
1887 edac->mcba_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
1888 "regmap-mcba");
1889 if (IS_ERR(edac->mcba_map)) {
1890 dev_err(edac->dev, "unable to get syscon regmap mcba\n");
1891 rc = PTR_ERR(edac->mcba_map);
1892 goto out_err;
1895 edac->mcbb_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
1896 "regmap-mcbb");
1897 if (IS_ERR(edac->mcbb_map)) {
1898 dev_err(edac->dev, "unable to get syscon regmap mcbb\n");
1899 rc = PTR_ERR(edac->mcbb_map);
1900 goto out_err;
1902 edac->efuse_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
1903 "regmap-efuse");
1904 if (IS_ERR(edac->efuse_map)) {
1905 dev_err(edac->dev, "unable to get syscon regmap efuse\n");
1906 rc = PTR_ERR(edac->efuse_map);
1907 goto out_err;
1911 * NOTE: The register bus resource is optional for compatibility
1912 * reason.
1914 edac->rb_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
1915 "regmap-rb");
1916 if (IS_ERR(edac->rb_map)) {
1917 dev_warn(edac->dev, "missing syscon regmap rb\n");
1918 edac->rb_map = NULL;
1921 res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
1922 edac->pcp_csr = devm_ioremap_resource(&pdev->dev, res);
1923 if (IS_ERR(edac->pcp_csr)) {
1924 dev_err(&pdev->dev, "no PCP resource address\n");
1925 rc = PTR_ERR(edac->pcp_csr);
1926 goto out_err;
1929 if (edac_op_state == EDAC_OPSTATE_INT) {
1930 int irq;
1931 int i;
1933 for (i = 0; i < 3; i++) {
1934 irq = platform_get_irq(pdev, i);
1935 if (irq < 0) {
1936 dev_err(&pdev->dev, "No IRQ resource\n");
1937 rc = -EINVAL;
1938 goto out_err;
1940 rc = devm_request_irq(&pdev->dev, irq,
1941 xgene_edac_isr, IRQF_SHARED,
1942 dev_name(&pdev->dev), edac);
1943 if (rc) {
1944 dev_err(&pdev->dev,
1945 "Could not request IRQ %d\n", irq);
1946 goto out_err;
1951 edac->dfs = edac_debugfs_create_dir(pdev->dev.kobj.name);
1953 for_each_child_of_node(pdev->dev.of_node, child) {
1954 if (!of_device_is_available(child))
1955 continue;
1956 if (of_device_is_compatible(child, "apm,xgene-edac-mc"))
1957 xgene_edac_mc_add(edac, child);
1958 if (of_device_is_compatible(child, "apm,xgene-edac-pmd"))
1959 xgene_edac_pmd_add(edac, child, 1);
1960 if (of_device_is_compatible(child, "apm,xgene-edac-pmd-v2"))
1961 xgene_edac_pmd_add(edac, child, 2);
1962 if (of_device_is_compatible(child, "apm,xgene-edac-l3"))
1963 xgene_edac_l3_add(edac, child, 1);
1964 if (of_device_is_compatible(child, "apm,xgene-edac-l3-v2"))
1965 xgene_edac_l3_add(edac, child, 2);
1966 if (of_device_is_compatible(child, "apm,xgene-edac-soc"))
1967 xgene_edac_soc_add(edac, child, 0);
1968 if (of_device_is_compatible(child, "apm,xgene-edac-soc-v1"))
1969 xgene_edac_soc_add(edac, child, 1);
1972 return 0;
1974 out_err:
1975 return rc;
1978 static int xgene_edac_remove(struct platform_device *pdev)
1980 struct xgene_edac *edac = dev_get_drvdata(&pdev->dev);
1981 struct xgene_edac_mc_ctx *mcu;
1982 struct xgene_edac_mc_ctx *temp_mcu;
1983 struct xgene_edac_pmd_ctx *pmd;
1984 struct xgene_edac_pmd_ctx *temp_pmd;
1985 struct xgene_edac_dev_ctx *node;
1986 struct xgene_edac_dev_ctx *temp_node;
1988 list_for_each_entry_safe(mcu, temp_mcu, &edac->mcus, next)
1989 xgene_edac_mc_remove(mcu);
1991 list_for_each_entry_safe(pmd, temp_pmd, &edac->pmds, next)
1992 xgene_edac_pmd_remove(pmd);
1994 list_for_each_entry_safe(node, temp_node, &edac->l3s, next)
1995 xgene_edac_l3_remove(node);
1997 list_for_each_entry_safe(node, temp_node, &edac->socs, next)
1998 xgene_edac_soc_remove(node);
2000 return 0;
2003 static const struct of_device_id xgene_edac_of_match[] = {
2004 { .compatible = "apm,xgene-edac" },
2007 MODULE_DEVICE_TABLE(of, xgene_edac_of_match);
2009 static struct platform_driver xgene_edac_driver = {
2010 .probe = xgene_edac_probe,
2011 .remove = xgene_edac_remove,
2012 .driver = {
2013 .name = "xgene-edac",
2014 .of_match_table = xgene_edac_of_match,
2018 static int __init xgene_edac_init(void)
2020 int rc;
2022 /* Make sure error reporting method is sane */
2023 switch (edac_op_state) {
2024 case EDAC_OPSTATE_POLL:
2025 case EDAC_OPSTATE_INT:
2026 break;
2027 default:
2028 edac_op_state = EDAC_OPSTATE_INT;
2029 break;
2032 rc = platform_driver_register(&xgene_edac_driver);
2033 if (rc) {
2034 edac_printk(KERN_ERR, EDAC_MOD_STR,
2035 "EDAC fails to register\n");
2036 goto reg_failed;
2039 return 0;
2041 reg_failed:
2042 return rc;
2044 module_init(xgene_edac_init);
2046 static void __exit xgene_edac_exit(void)
2048 platform_driver_unregister(&xgene_edac_driver);
2050 module_exit(xgene_edac_exit);
2052 MODULE_LICENSE("GPL");
2053 MODULE_AUTHOR("Feng Kan <fkan@apm.com>");
2054 MODULE_DESCRIPTION("APM X-Gene EDAC driver");
2055 module_param(edac_op_state, int, 0444);
2056 MODULE_PARM_DESC(edac_op_state,
2057 "EDAC error reporting state: 0=Poll, 2=Interrupt");