blob: 5ff42d5b019d520100d45233999a009c5a9645fb [file] [log] [blame]
Loc Ho0d442932015-05-22 17:32:59 -06001/*
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"
Borislav Petkov09bd1b42015-09-22 13:13:46 +020032#include "edac_module.h"
Loc Ho0d442932015-05-22 17:32:59 -060033
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
58struct 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
71 struct mutex mc_lock;
72 int mc_active_mask;
73 int mc_registered_mask;
74};
75
76static void xgene_edac_pcp_rd(struct xgene_edac *edac, u32 reg, u32 *val)
77{
78 *val = readl(edac->pcp_csr + reg);
79}
80
81static void xgene_edac_pcp_clrbits(struct xgene_edac *edac, u32 reg,
82 u32 bits_mask)
83{
84 u32 val;
85
86 spin_lock(&edac->lock);
87 val = readl(edac->pcp_csr + reg);
88 val &= ~bits_mask;
89 writel(val, edac->pcp_csr + reg);
90 spin_unlock(&edac->lock);
91}
92
93static void xgene_edac_pcp_setbits(struct xgene_edac *edac, u32 reg,
94 u32 bits_mask)
95{
96 u32 val;
97
98 spin_lock(&edac->lock);
99 val = readl(edac->pcp_csr + reg);
100 val |= bits_mask;
101 writel(val, edac->pcp_csr + reg);
102 spin_unlock(&edac->lock);
103}
104
105/* Memory controller error CSR */
106#define MCU_MAX_RANK 8
107#define MCU_RANK_STRIDE 0x40
108
109#define MCUGECR 0x0110
110#define MCU_GECR_DEMANDUCINTREN_MASK BIT(0)
111#define MCU_GECR_BACKUCINTREN_MASK BIT(1)
112#define MCU_GECR_CINTREN_MASK BIT(2)
113#define MUC_GECR_MCUADDRERREN_MASK BIT(9)
114#define MCUGESR 0x0114
115#define MCU_GESR_ADDRNOMATCH_ERR_MASK BIT(7)
116#define MCU_GESR_ADDRMULTIMATCH_ERR_MASK BIT(6)
117#define MCU_GESR_PHYP_ERR_MASK BIT(3)
118#define MCUESRR0 0x0314
119#define MCU_ESRR_MULTUCERR_MASK BIT(3)
120#define MCU_ESRR_BACKUCERR_MASK BIT(2)
121#define MCU_ESRR_DEMANDUCERR_MASK BIT(1)
122#define MCU_ESRR_CERR_MASK BIT(0)
123#define MCUESRRA0 0x0318
124#define MCUEBLRR0 0x031c
125#define MCU_EBLRR_ERRBANK_RD(src) (((src) & 0x00000007) >> 0)
126#define MCUERCRR0 0x0320
127#define MCU_ERCRR_ERRROW_RD(src) (((src) & 0xFFFF0000) >> 16)
128#define MCU_ERCRR_ERRCOL_RD(src) ((src) & 0x00000FFF)
129#define MCUSBECNT0 0x0324
130#define MCU_SBECNT_COUNT(src) ((src) & 0xFFFF)
131
132#define CSW_CSWCR 0x0000
133#define CSW_CSWCR_DUALMCB_MASK BIT(0)
134
135#define MCBADDRMR 0x0000
136#define MCBADDRMR_MCU_INTLV_MODE_MASK BIT(3)
137#define MCBADDRMR_DUALMCU_MODE_MASK BIT(2)
138#define MCBADDRMR_MCB_INTLV_MODE_MASK BIT(1)
139#define MCBADDRMR_ADDRESS_MODE_MASK BIT(0)
140
141struct xgene_edac_mc_ctx {
142 struct list_head next;
143 char *name;
144 struct mem_ctl_info *mci;
145 struct xgene_edac *edac;
146 void __iomem *mcu_csr;
147 u32 mcu_id;
148};
149
150static ssize_t xgene_edac_mc_err_inject_write(struct file *file,
151 const char __user *data,
152 size_t count, loff_t *ppos)
153{
154 struct mem_ctl_info *mci = file->private_data;
155 struct xgene_edac_mc_ctx *ctx = mci->pvt_info;
156 int i;
157
158 for (i = 0; i < MCU_MAX_RANK; i++) {
159 writel(MCU_ESRR_MULTUCERR_MASK | MCU_ESRR_BACKUCERR_MASK |
160 MCU_ESRR_DEMANDUCERR_MASK | MCU_ESRR_CERR_MASK,
161 ctx->mcu_csr + MCUESRRA0 + i * MCU_RANK_STRIDE);
162 }
163 return count;
164}
165
166static const struct file_operations xgene_edac_mc_debug_inject_fops = {
167 .open = simple_open,
168 .write = xgene_edac_mc_err_inject_write,
169 .llseek = generic_file_llseek,
170};
171
172static void xgene_edac_mc_create_debugfs_node(struct mem_ctl_info *mci)
173{
174 if (!IS_ENABLED(CONFIG_EDAC_DEBUG))
175 return;
Borislav Petkov09bd1b42015-09-22 13:13:46 +0200176
Loc Ho0d442932015-05-22 17:32:59 -0600177 if (!mci->debugfs)
178 return;
Borislav Petkov09bd1b42015-09-22 13:13:46 +0200179
180 edac_debugfs_create_file("inject_ctrl", S_IWUSR, mci->debugfs, mci,
181 &xgene_edac_mc_debug_inject_fops);
Loc Ho0d442932015-05-22 17:32:59 -0600182}
183
184static void xgene_edac_mc_check(struct mem_ctl_info *mci)
185{
186 struct xgene_edac_mc_ctx *ctx = mci->pvt_info;
187 unsigned int pcp_hp_stat;
188 unsigned int pcp_lp_stat;
189 u32 reg;
190 u32 rank;
191 u32 bank;
192 u32 count;
193 u32 col_row;
194
195 xgene_edac_pcp_rd(ctx->edac, PCPHPERRINTSTS, &pcp_hp_stat);
196 xgene_edac_pcp_rd(ctx->edac, PCPLPERRINTSTS, &pcp_lp_stat);
197 if (!((MCU_UNCORR_ERR_MASK & pcp_hp_stat) ||
198 (MCU_CTL_ERR_MASK & pcp_hp_stat) ||
199 (MCU_CORR_ERR_MASK & pcp_lp_stat)))
200 return;
201
202 for (rank = 0; rank < MCU_MAX_RANK; rank++) {
203 reg = readl(ctx->mcu_csr + MCUESRR0 + rank * MCU_RANK_STRIDE);
204
205 /* Detect uncorrectable memory error */
206 if (reg & (MCU_ESRR_DEMANDUCERR_MASK |
207 MCU_ESRR_BACKUCERR_MASK)) {
208 /* Detected uncorrectable memory error */
209 edac_mc_chipset_printk(mci, KERN_ERR, "X-Gene",
210 "MCU uncorrectable error at rank %d\n", rank);
211
212 edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci,
213 1, 0, 0, 0, 0, 0, -1, mci->ctl_name, "");
214 }
215
216 /* Detect correctable memory error */
217 if (reg & MCU_ESRR_CERR_MASK) {
218 bank = readl(ctx->mcu_csr + MCUEBLRR0 +
219 rank * MCU_RANK_STRIDE);
220 col_row = readl(ctx->mcu_csr + MCUERCRR0 +
221 rank * MCU_RANK_STRIDE);
222 count = readl(ctx->mcu_csr + MCUSBECNT0 +
223 rank * MCU_RANK_STRIDE);
224 edac_mc_chipset_printk(mci, KERN_WARNING, "X-Gene",
225 "MCU correctable error at rank %d bank %d column %d row %d count %d\n",
226 rank, MCU_EBLRR_ERRBANK_RD(bank),
227 MCU_ERCRR_ERRCOL_RD(col_row),
228 MCU_ERCRR_ERRROW_RD(col_row),
229 MCU_SBECNT_COUNT(count));
230
231 edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci,
232 1, 0, 0, 0, 0, 0, -1, mci->ctl_name, "");
233 }
234
235 /* Clear all error registers */
236 writel(0x0, ctx->mcu_csr + MCUEBLRR0 + rank * MCU_RANK_STRIDE);
237 writel(0x0, ctx->mcu_csr + MCUERCRR0 + rank * MCU_RANK_STRIDE);
238 writel(0x0, ctx->mcu_csr + MCUSBECNT0 +
239 rank * MCU_RANK_STRIDE);
240 writel(reg, ctx->mcu_csr + MCUESRR0 + rank * MCU_RANK_STRIDE);
241 }
242
243 /* Detect memory controller error */
244 reg = readl(ctx->mcu_csr + MCUGESR);
245 if (reg) {
246 if (reg & MCU_GESR_ADDRNOMATCH_ERR_MASK)
247 edac_mc_chipset_printk(mci, KERN_WARNING, "X-Gene",
248 "MCU address miss-match error\n");
249 if (reg & MCU_GESR_ADDRMULTIMATCH_ERR_MASK)
250 edac_mc_chipset_printk(mci, KERN_WARNING, "X-Gene",
251 "MCU address multi-match error\n");
252
253 writel(reg, ctx->mcu_csr + MCUGESR);
254 }
255}
256
257static void xgene_edac_mc_irq_ctl(struct mem_ctl_info *mci, bool enable)
258{
259 struct xgene_edac_mc_ctx *ctx = mci->pvt_info;
260 unsigned int val;
261
262 if (edac_op_state != EDAC_OPSTATE_INT)
263 return;
264
265 mutex_lock(&ctx->edac->mc_lock);
266
267 /*
268 * As there is only single bit for enable error and interrupt mask,
269 * we must only enable top level interrupt after all MCUs are
270 * registered. Otherwise, if there is an error and the corresponding
271 * MCU has not registered, the interrupt will never get cleared. To
272 * determine all MCU have registered, we will keep track of active
273 * MCUs and registered MCUs.
274 */
275 if (enable) {
276 /* Set registered MCU bit */
277 ctx->edac->mc_registered_mask |= 1 << ctx->mcu_id;
278
279 /* Enable interrupt after all active MCU registered */
280 if (ctx->edac->mc_registered_mask ==
281 ctx->edac->mc_active_mask) {
282 /* Enable memory controller top level interrupt */
283 xgene_edac_pcp_clrbits(ctx->edac, PCPHPERRINTMSK,
284 MCU_UNCORR_ERR_MASK |
285 MCU_CTL_ERR_MASK);
286 xgene_edac_pcp_clrbits(ctx->edac, PCPLPERRINTMSK,
287 MCU_CORR_ERR_MASK);
288 }
289
290 /* Enable MCU interrupt and error reporting */
291 val = readl(ctx->mcu_csr + MCUGECR);
292 val |= MCU_GECR_DEMANDUCINTREN_MASK |
293 MCU_GECR_BACKUCINTREN_MASK |
294 MCU_GECR_CINTREN_MASK |
295 MUC_GECR_MCUADDRERREN_MASK;
296 writel(val, ctx->mcu_csr + MCUGECR);
297 } else {
298 /* Disable MCU interrupt */
299 val = readl(ctx->mcu_csr + MCUGECR);
300 val &= ~(MCU_GECR_DEMANDUCINTREN_MASK |
301 MCU_GECR_BACKUCINTREN_MASK |
302 MCU_GECR_CINTREN_MASK |
303 MUC_GECR_MCUADDRERREN_MASK);
304 writel(val, ctx->mcu_csr + MCUGECR);
305
306 /* Disable memory controller top level interrupt */
307 xgene_edac_pcp_setbits(ctx->edac, PCPHPERRINTMSK,
308 MCU_UNCORR_ERR_MASK | MCU_CTL_ERR_MASK);
309 xgene_edac_pcp_setbits(ctx->edac, PCPLPERRINTMSK,
310 MCU_CORR_ERR_MASK);
311
312 /* Clear registered MCU bit */
313 ctx->edac->mc_registered_mask &= ~(1 << ctx->mcu_id);
314 }
315
316 mutex_unlock(&ctx->edac->mc_lock);
317}
318
319static int xgene_edac_mc_is_active(struct xgene_edac_mc_ctx *ctx, int mc_idx)
320{
321 unsigned int reg;
322 u32 mcu_mask;
323
324 if (regmap_read(ctx->edac->csw_map, CSW_CSWCR, &reg))
325 return 0;
326
327 if (reg & CSW_CSWCR_DUALMCB_MASK) {
328 /*
329 * Dual MCB active - Determine if all 4 active or just MCU0
330 * and MCU2 active
331 */
332 if (regmap_read(ctx->edac->mcbb_map, MCBADDRMR, &reg))
333 return 0;
334 mcu_mask = (reg & MCBADDRMR_DUALMCU_MODE_MASK) ? 0xF : 0x5;
335 } else {
336 /*
337 * Single MCB active - Determine if MCU0/MCU1 or just MCU0
338 * active
339 */
340 if (regmap_read(ctx->edac->mcba_map, MCBADDRMR, &reg))
341 return 0;
342 mcu_mask = (reg & MCBADDRMR_DUALMCU_MODE_MASK) ? 0x3 : 0x1;
343 }
344
345 /* Save active MC mask if hasn't set already */
346 if (!ctx->edac->mc_active_mask)
347 ctx->edac->mc_active_mask = mcu_mask;
348
349 return (mcu_mask & (1 << mc_idx)) ? 1 : 0;
350}
351
352static int xgene_edac_mc_add(struct xgene_edac *edac, struct device_node *np)
353{
354 struct mem_ctl_info *mci;
355 struct edac_mc_layer layers[2];
356 struct xgene_edac_mc_ctx tmp_ctx;
357 struct xgene_edac_mc_ctx *ctx;
358 struct resource res;
359 int rc;
360
361 memset(&tmp_ctx, 0, sizeof(tmp_ctx));
362 tmp_ctx.edac = edac;
363
364 if (!devres_open_group(edac->dev, xgene_edac_mc_add, GFP_KERNEL))
365 return -ENOMEM;
366
367 rc = of_address_to_resource(np, 0, &res);
368 if (rc < 0) {
369 dev_err(edac->dev, "no MCU resource address\n");
370 goto err_group;
371 }
372 tmp_ctx.mcu_csr = devm_ioremap_resource(edac->dev, &res);
373 if (IS_ERR(tmp_ctx.mcu_csr)) {
374 dev_err(edac->dev, "unable to map MCU resource\n");
375 rc = PTR_ERR(tmp_ctx.mcu_csr);
376 goto err_group;
377 }
378
379 /* Ignore non-active MCU */
380 if (of_property_read_u32(np, "memory-controller", &tmp_ctx.mcu_id)) {
381 dev_err(edac->dev, "no memory-controller property\n");
382 rc = -ENODEV;
383 goto err_group;
384 }
385 if (!xgene_edac_mc_is_active(&tmp_ctx, tmp_ctx.mcu_id)) {
386 rc = -ENODEV;
387 goto err_group;
388 }
389
390 layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
391 layers[0].size = 4;
392 layers[0].is_virt_csrow = true;
393 layers[1].type = EDAC_MC_LAYER_CHANNEL;
394 layers[1].size = 2;
395 layers[1].is_virt_csrow = false;
396 mci = edac_mc_alloc(tmp_ctx.mcu_id, ARRAY_SIZE(layers), layers,
397 sizeof(*ctx));
398 if (!mci) {
399 rc = -ENOMEM;
400 goto err_group;
401 }
402
403 ctx = mci->pvt_info;
404 *ctx = tmp_ctx; /* Copy over resource value */
405 ctx->name = "xgene_edac_mc_err";
406 ctx->mci = mci;
407 mci->pdev = &mci->dev;
408 mci->ctl_name = ctx->name;
409 mci->dev_name = ctx->name;
410
411 mci->mtype_cap = MEM_FLAG_RDDR | MEM_FLAG_RDDR2 | MEM_FLAG_RDDR3 |
412 MEM_FLAG_DDR | MEM_FLAG_DDR2 | MEM_FLAG_DDR3;
413 mci->edac_ctl_cap = EDAC_FLAG_SECDED;
414 mci->edac_cap = EDAC_FLAG_SECDED;
415 mci->mod_name = EDAC_MOD_STR;
416 mci->mod_ver = "0.1";
417 mci->ctl_page_to_phys = NULL;
418 mci->scrub_cap = SCRUB_FLAG_HW_SRC;
419 mci->scrub_mode = SCRUB_HW_SRC;
420
421 if (edac_op_state == EDAC_OPSTATE_POLL)
422 mci->edac_check = xgene_edac_mc_check;
423
424 if (edac_mc_add_mc(mci)) {
425 dev_err(edac->dev, "edac_mc_add_mc failed\n");
426 rc = -EINVAL;
427 goto err_free;
428 }
429
430 xgene_edac_mc_create_debugfs_node(mci);
431
432 list_add(&ctx->next, &edac->mcus);
433
434 xgene_edac_mc_irq_ctl(mci, true);
435
436 devres_remove_group(edac->dev, xgene_edac_mc_add);
437
438 dev_info(edac->dev, "X-Gene EDAC MC registered\n");
439 return 0;
440
441err_free:
442 edac_mc_free(mci);
443err_group:
444 devres_release_group(edac->dev, xgene_edac_mc_add);
445 return rc;
446}
447
448static int xgene_edac_mc_remove(struct xgene_edac_mc_ctx *mcu)
449{
450 xgene_edac_mc_irq_ctl(mcu->mci, false);
451 edac_mc_del_mc(&mcu->mci->dev);
452 edac_mc_free(mcu->mci);
453 return 0;
454}
455
456/* CPU L1/L2 error CSR */
457#define MAX_CPU_PER_PMD 2
458#define CPU_CSR_STRIDE 0x00100000
459#define CPU_L2C_PAGE 0x000D0000
460#define CPU_MEMERR_L2C_PAGE 0x000E0000
461#define CPU_MEMERR_CPU_PAGE 0x000F0000
462
463#define MEMERR_CPU_ICFECR_PAGE_OFFSET 0x0000
464#define MEMERR_CPU_ICFESR_PAGE_OFFSET 0x0004
465#define MEMERR_CPU_ICFESR_ERRWAY_RD(src) (((src) & 0xFF000000) >> 24)
466#define MEMERR_CPU_ICFESR_ERRINDEX_RD(src) (((src) & 0x003F0000) >> 16)
467#define MEMERR_CPU_ICFESR_ERRINFO_RD(src) (((src) & 0x0000FF00) >> 8)
468#define MEMERR_CPU_ICFESR_ERRTYPE_RD(src) (((src) & 0x00000070) >> 4)
469#define MEMERR_CPU_ICFESR_MULTCERR_MASK BIT(2)
470#define MEMERR_CPU_ICFESR_CERR_MASK BIT(0)
471#define MEMERR_CPU_LSUESR_PAGE_OFFSET 0x000c
472#define MEMERR_CPU_LSUESR_ERRWAY_RD(src) (((src) & 0xFF000000) >> 24)
473#define MEMERR_CPU_LSUESR_ERRINDEX_RD(src) (((src) & 0x003F0000) >> 16)
474#define MEMERR_CPU_LSUESR_ERRINFO_RD(src) (((src) & 0x0000FF00) >> 8)
475#define MEMERR_CPU_LSUESR_ERRTYPE_RD(src) (((src) & 0x00000070) >> 4)
476#define MEMERR_CPU_LSUESR_MULTCERR_MASK BIT(2)
477#define MEMERR_CPU_LSUESR_CERR_MASK BIT(0)
478#define MEMERR_CPU_LSUECR_PAGE_OFFSET 0x0008
479#define MEMERR_CPU_MMUECR_PAGE_OFFSET 0x0010
480#define MEMERR_CPU_MMUESR_PAGE_OFFSET 0x0014
481#define MEMERR_CPU_MMUESR_ERRWAY_RD(src) (((src) & 0xFF000000) >> 24)
482#define MEMERR_CPU_MMUESR_ERRINDEX_RD(src) (((src) & 0x007F0000) >> 16)
483#define MEMERR_CPU_MMUESR_ERRINFO_RD(src) (((src) & 0x0000FF00) >> 8)
484#define MEMERR_CPU_MMUESR_ERRREQSTR_LSU_MASK BIT(7)
485#define MEMERR_CPU_MMUESR_ERRTYPE_RD(src) (((src) & 0x00000070) >> 4)
486#define MEMERR_CPU_MMUESR_MULTCERR_MASK BIT(2)
487#define MEMERR_CPU_MMUESR_CERR_MASK BIT(0)
488#define MEMERR_CPU_ICFESRA_PAGE_OFFSET 0x0804
489#define MEMERR_CPU_LSUESRA_PAGE_OFFSET 0x080c
490#define MEMERR_CPU_MMUESRA_PAGE_OFFSET 0x0814
491
492#define MEMERR_L2C_L2ECR_PAGE_OFFSET 0x0000
493#define MEMERR_L2C_L2ESR_PAGE_OFFSET 0x0004
494#define MEMERR_L2C_L2ESR_ERRSYN_RD(src) (((src) & 0xFF000000) >> 24)
495#define MEMERR_L2C_L2ESR_ERRWAY_RD(src) (((src) & 0x00FC0000) >> 18)
496#define MEMERR_L2C_L2ESR_ERRCPU_RD(src) (((src) & 0x00020000) >> 17)
497#define MEMERR_L2C_L2ESR_ERRGROUP_RD(src) (((src) & 0x0000E000) >> 13)
498#define MEMERR_L2C_L2ESR_ERRACTION_RD(src) (((src) & 0x00001C00) >> 10)
499#define MEMERR_L2C_L2ESR_ERRTYPE_RD(src) (((src) & 0x00000300) >> 8)
500#define MEMERR_L2C_L2ESR_MULTUCERR_MASK BIT(3)
501#define MEMERR_L2C_L2ESR_MULTICERR_MASK BIT(2)
502#define MEMERR_L2C_L2ESR_UCERR_MASK BIT(1)
503#define MEMERR_L2C_L2ESR_ERR_MASK BIT(0)
504#define MEMERR_L2C_L2EALR_PAGE_OFFSET 0x0008
505#define CPUX_L2C_L2RTOCR_PAGE_OFFSET 0x0010
506#define MEMERR_L2C_L2EAHR_PAGE_OFFSET 0x000c
507#define CPUX_L2C_L2RTOSR_PAGE_OFFSET 0x0014
508#define MEMERR_L2C_L2RTOSR_MULTERR_MASK BIT(1)
509#define MEMERR_L2C_L2RTOSR_ERR_MASK BIT(0)
510#define CPUX_L2C_L2RTOALR_PAGE_OFFSET 0x0018
511#define CPUX_L2C_L2RTOAHR_PAGE_OFFSET 0x001c
512#define MEMERR_L2C_L2ESRA_PAGE_OFFSET 0x0804
513
514/*
515 * Processor Module Domain (PMD) context - Context for a pair of processsors.
516 * Each PMD consists of 2 CPUs and a shared L2 cache. Each CPU consists of
517 * its own L1 cache.
518 */
519struct xgene_edac_pmd_ctx {
520 struct list_head next;
521 struct device ddev;
522 char *name;
523 struct xgene_edac *edac;
524 struct edac_device_ctl_info *edac_dev;
525 void __iomem *pmd_csr;
526 u32 pmd;
Arnd Bergmann451bb7f2015-06-01 16:09:35 -0600527 int version;
Loc Ho0d442932015-05-22 17:32:59 -0600528};
529
530static void xgene_edac_pmd_l1_check(struct edac_device_ctl_info *edac_dev,
531 int cpu_idx)
532{
533 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
534 void __iomem *pg_f;
535 u32 val;
536
537 pg_f = ctx->pmd_csr + cpu_idx * CPU_CSR_STRIDE + CPU_MEMERR_CPU_PAGE;
538
539 val = readl(pg_f + MEMERR_CPU_ICFESR_PAGE_OFFSET);
540 if (val) {
541 dev_err(edac_dev->dev,
542 "CPU%d L1 memory error ICF 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X\n",
543 ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val,
544 MEMERR_CPU_ICFESR_ERRWAY_RD(val),
545 MEMERR_CPU_ICFESR_ERRINDEX_RD(val),
546 MEMERR_CPU_ICFESR_ERRINFO_RD(val));
547 if (val & MEMERR_CPU_ICFESR_CERR_MASK)
548 dev_err(edac_dev->dev,
549 "One or more correctable error\n");
550 if (val & MEMERR_CPU_ICFESR_MULTCERR_MASK)
551 dev_err(edac_dev->dev, "Multiple correctable error\n");
552 switch (MEMERR_CPU_ICFESR_ERRTYPE_RD(val)) {
553 case 1:
554 dev_err(edac_dev->dev, "L1 TLB multiple hit\n");
555 break;
556 case 2:
557 dev_err(edac_dev->dev, "Way select multiple hit\n");
558 break;
559 case 3:
560 dev_err(edac_dev->dev, "Physical tag parity error\n");
561 break;
562 case 4:
563 case 5:
564 dev_err(edac_dev->dev, "L1 data parity error\n");
565 break;
566 case 6:
567 dev_err(edac_dev->dev, "L1 pre-decode parity error\n");
568 break;
569 }
570
571 /* Clear any HW errors */
572 writel(val, pg_f + MEMERR_CPU_ICFESR_PAGE_OFFSET);
573
574 if (val & (MEMERR_CPU_ICFESR_CERR_MASK |
575 MEMERR_CPU_ICFESR_MULTCERR_MASK))
576 edac_device_handle_ce(edac_dev, 0, 0,
577 edac_dev->ctl_name);
578 }
579
580 val = readl(pg_f + MEMERR_CPU_LSUESR_PAGE_OFFSET);
581 if (val) {
582 dev_err(edac_dev->dev,
583 "CPU%d memory error LSU 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X\n",
584 ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val,
585 MEMERR_CPU_LSUESR_ERRWAY_RD(val),
586 MEMERR_CPU_LSUESR_ERRINDEX_RD(val),
587 MEMERR_CPU_LSUESR_ERRINFO_RD(val));
588 if (val & MEMERR_CPU_LSUESR_CERR_MASK)
589 dev_err(edac_dev->dev,
590 "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;
614 }
615
616 /* Clear any HW errors */
617 writel(val, pg_f + MEMERR_CPU_LSUESR_PAGE_OFFSET);
618
619 if (val & (MEMERR_CPU_LSUESR_CERR_MASK |
620 MEMERR_CPU_LSUESR_MULTCERR_MASK))
621 edac_device_handle_ce(edac_dev, 0, 0,
622 edac_dev->ctl_name);
623 }
624
625 val = readl(pg_f + MEMERR_CPU_MMUESR_PAGE_OFFSET);
626 if (val) {
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" :
634 "ICF");
635 if (val & MEMERR_CPU_MMUESR_CERR_MASK)
636 dev_err(edac_dev->dev,
637 "One or more correctable error\n");
638 if (val & MEMERR_CPU_MMUESR_MULTCERR_MASK)
639 dev_err(edac_dev->dev, "Multiple correctable error\n");
640 switch (MEMERR_CPU_MMUESR_ERRTYPE_RD(val)) {
641 case 0:
642 dev_err(edac_dev->dev, "Stage 1 UTB hit error\n");
643 break;
644 case 1:
645 dev_err(edac_dev->dev, "Stage 1 UTB miss error\n");
646 break;
647 case 2:
648 dev_err(edac_dev->dev, "Stage 1 UTB allocate error\n");
649 break;
650 case 3:
651 dev_err(edac_dev->dev,
652 "TMO operation single bank error\n");
653 break;
654 case 4:
655 dev_err(edac_dev->dev, "Stage 2 UTB error\n");
656 break;
657 case 5:
658 dev_err(edac_dev->dev, "Stage 2 UTB miss error\n");
659 break;
660 case 6:
661 dev_err(edac_dev->dev, "Stage 2 UTB allocate error\n");
662 break;
663 case 7:
664 dev_err(edac_dev->dev,
665 "TMO operation multiple bank error\n");
666 break;
667 }
668
669 /* Clear any HW errors */
670 writel(val, pg_f + MEMERR_CPU_MMUESR_PAGE_OFFSET);
671
672 edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
673 }
674}
675
676static void xgene_edac_pmd_l2_check(struct edac_device_ctl_info *edac_dev)
677{
678 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
679 void __iomem *pg_d;
680 void __iomem *pg_e;
681 u32 val_hi;
682 u32 val_lo;
683 u32 val;
684
685 /* Check L2 */
686 pg_e = ctx->pmd_csr + CPU_MEMERR_L2C_PAGE;
687 val = readl(pg_e + MEMERR_L2C_L2ESR_PAGE_OFFSET);
688 if (val) {
689 val_lo = readl(pg_e + MEMERR_L2C_L2EALR_PAGE_OFFSET);
690 val_hi = readl(pg_e + MEMERR_L2C_L2EAHR_PAGE_OFFSET);
691 dev_err(edac_dev->dev,
692 "PMD%d memory error L2C L2ESR 0x%08X @ 0x%08X.%08X\n",
693 ctx->pmd, val, val_hi, val_lo);
694 dev_err(edac_dev->dev,
695 "ErrSyndrome 0x%02X ErrWay 0x%02X ErrCpu %d ErrGroup 0x%02X ErrAction 0x%02X\n",
696 MEMERR_L2C_L2ESR_ERRSYN_RD(val),
697 MEMERR_L2C_L2ESR_ERRWAY_RD(val),
698 MEMERR_L2C_L2ESR_ERRCPU_RD(val),
699 MEMERR_L2C_L2ESR_ERRGROUP_RD(val),
700 MEMERR_L2C_L2ESR_ERRACTION_RD(val));
701
702 if (val & MEMERR_L2C_L2ESR_ERR_MASK)
703 dev_err(edac_dev->dev,
704 "One or more correctable error\n");
705 if (val & MEMERR_L2C_L2ESR_MULTICERR_MASK)
706 dev_err(edac_dev->dev, "Multiple correctable error\n");
707 if (val & MEMERR_L2C_L2ESR_UCERR_MASK)
708 dev_err(edac_dev->dev,
709 "One or more uncorrectable error\n");
710 if (val & MEMERR_L2C_L2ESR_MULTUCERR_MASK)
711 dev_err(edac_dev->dev,
712 "Multiple uncorrectable error\n");
713
714 switch (MEMERR_L2C_L2ESR_ERRTYPE_RD(val)) {
715 case 0:
716 dev_err(edac_dev->dev, "Outbound SDB parity error\n");
717 break;
718 case 1:
719 dev_err(edac_dev->dev, "Inbound SDB parity error\n");
720 break;
721 case 2:
722 dev_err(edac_dev->dev, "Tag ECC error\n");
723 break;
724 case 3:
725 dev_err(edac_dev->dev, "Data ECC error\n");
726 break;
727 }
728
729 /* Clear any HW errors */
730 writel(val, pg_e + MEMERR_L2C_L2ESR_PAGE_OFFSET);
731
732 if (val & (MEMERR_L2C_L2ESR_ERR_MASK |
733 MEMERR_L2C_L2ESR_MULTICERR_MASK))
734 edac_device_handle_ce(edac_dev, 0, 0,
735 edac_dev->ctl_name);
736 if (val & (MEMERR_L2C_L2ESR_UCERR_MASK |
737 MEMERR_L2C_L2ESR_MULTUCERR_MASK))
738 edac_device_handle_ue(edac_dev, 0, 0,
739 edac_dev->ctl_name);
740 }
741
742 /* Check if any memory request timed out on L2 cache */
743 pg_d = ctx->pmd_csr + CPU_L2C_PAGE;
744 val = readl(pg_d + CPUX_L2C_L2RTOSR_PAGE_OFFSET);
745 if (val) {
746 val_lo = readl(pg_d + CPUX_L2C_L2RTOALR_PAGE_OFFSET);
747 val_hi = readl(pg_d + CPUX_L2C_L2RTOAHR_PAGE_OFFSET);
748 dev_err(edac_dev->dev,
749 "PMD%d L2C error L2C RTOSR 0x%08X @ 0x%08X.%08X\n",
750 ctx->pmd, val, val_hi, val_lo);
751 writel(val, pg_d + CPUX_L2C_L2RTOSR_PAGE_OFFSET);
752 }
753}
754
755static void xgene_edac_pmd_check(struct edac_device_ctl_info *edac_dev)
756{
757 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
758 unsigned int pcp_hp_stat;
759 int i;
760
761 xgene_edac_pcp_rd(ctx->edac, PCPHPERRINTSTS, &pcp_hp_stat);
762 if (!((PMD0_MERR_MASK << ctx->pmd) & pcp_hp_stat))
763 return;
764
765 /* Check CPU L1 error */
766 for (i = 0; i < MAX_CPU_PER_PMD; i++)
767 xgene_edac_pmd_l1_check(edac_dev, i);
768
769 /* Check CPU L2 error */
770 xgene_edac_pmd_l2_check(edac_dev);
771}
772
773static void xgene_edac_pmd_cpu_hw_cfg(struct edac_device_ctl_info *edac_dev,
774 int cpu)
775{
776 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
777 void __iomem *pg_f = ctx->pmd_csr + cpu * CPU_CSR_STRIDE +
778 CPU_MEMERR_CPU_PAGE;
779
780 /*
781 * Enable CPU memory error:
782 * MEMERR_CPU_ICFESRA, MEMERR_CPU_LSUESRA, and MEMERR_CPU_MMUESRA
783 */
784 writel(0x00000301, pg_f + MEMERR_CPU_ICFECR_PAGE_OFFSET);
785 writel(0x00000301, pg_f + MEMERR_CPU_LSUECR_PAGE_OFFSET);
786 writel(0x00000101, pg_f + MEMERR_CPU_MMUECR_PAGE_OFFSET);
787}
788
Loc Ho0d442932015-05-22 17:32:59 -0600789static void xgene_edac_pmd_hw_cfg(struct edac_device_ctl_info *edac_dev)
790{
791 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
792 void __iomem *pg_d = ctx->pmd_csr + CPU_L2C_PAGE;
793 void __iomem *pg_e = ctx->pmd_csr + CPU_MEMERR_L2C_PAGE;
794
795 /* Enable PMD memory error - MEMERR_L2C_L2ECR and L2C_L2RTOCR */
796 writel(0x00000703, pg_e + MEMERR_L2C_L2ECR_PAGE_OFFSET);
797 /* Configure L2C HW request time out feature if supported */
Arnd Bergmann451bb7f2015-06-01 16:09:35 -0600798 if (ctx->version > 1)
Loc Ho0d442932015-05-22 17:32:59 -0600799 writel(0x00000119, pg_d + CPUX_L2C_L2RTOCR_PAGE_OFFSET);
800}
801
802static void xgene_edac_pmd_hw_ctl(struct edac_device_ctl_info *edac_dev,
803 bool enable)
804{
805 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
806 int i;
807
808 /* Enable PMD error interrupt */
809 if (edac_dev->op_state == OP_RUNNING_INTERRUPT) {
810 if (enable)
811 xgene_edac_pcp_clrbits(ctx->edac, PCPHPERRINTMSK,
812 PMD0_MERR_MASK << ctx->pmd);
813 else
814 xgene_edac_pcp_setbits(ctx->edac, PCPHPERRINTMSK,
815 PMD0_MERR_MASK << ctx->pmd);
816 }
817
818 if (enable) {
819 xgene_edac_pmd_hw_cfg(edac_dev);
820
821 /* Two CPUs per a PMD */
822 for (i = 0; i < MAX_CPU_PER_PMD; i++)
823 xgene_edac_pmd_cpu_hw_cfg(edac_dev, i);
824 }
825}
826
827static ssize_t xgene_edac_pmd_l1_inject_ctrl_write(struct file *file,
828 const char __user *data,
829 size_t count, loff_t *ppos)
830{
831 struct edac_device_ctl_info *edac_dev = file->private_data;
832 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
833 void __iomem *cpux_pg_f;
834 int i;
835
836 for (i = 0; i < MAX_CPU_PER_PMD; i++) {
837 cpux_pg_f = ctx->pmd_csr + i * CPU_CSR_STRIDE +
838 CPU_MEMERR_CPU_PAGE;
839
840 writel(MEMERR_CPU_ICFESR_MULTCERR_MASK |
841 MEMERR_CPU_ICFESR_CERR_MASK,
842 cpux_pg_f + MEMERR_CPU_ICFESRA_PAGE_OFFSET);
843 writel(MEMERR_CPU_LSUESR_MULTCERR_MASK |
844 MEMERR_CPU_LSUESR_CERR_MASK,
845 cpux_pg_f + MEMERR_CPU_LSUESRA_PAGE_OFFSET);
846 writel(MEMERR_CPU_MMUESR_MULTCERR_MASK |
847 MEMERR_CPU_MMUESR_CERR_MASK,
848 cpux_pg_f + MEMERR_CPU_MMUESRA_PAGE_OFFSET);
849 }
850 return count;
851}
852
853static ssize_t xgene_edac_pmd_l2_inject_ctrl_write(struct file *file,
854 const char __user *data,
855 size_t count, loff_t *ppos)
856{
857 struct edac_device_ctl_info *edac_dev = file->private_data;
858 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
859 void __iomem *pg_e = ctx->pmd_csr + CPU_MEMERR_L2C_PAGE;
860
861 writel(MEMERR_L2C_L2ESR_MULTUCERR_MASK |
862 MEMERR_L2C_L2ESR_MULTICERR_MASK |
863 MEMERR_L2C_L2ESR_UCERR_MASK |
864 MEMERR_L2C_L2ESR_ERR_MASK,
865 pg_e + MEMERR_L2C_L2ESRA_PAGE_OFFSET);
866 return count;
867}
868
869static const struct file_operations xgene_edac_pmd_debug_inject_fops[] = {
870 {
871 .open = simple_open,
872 .write = xgene_edac_pmd_l1_inject_ctrl_write,
873 .llseek = generic_file_llseek, },
874 {
875 .open = simple_open,
876 .write = xgene_edac_pmd_l2_inject_ctrl_write,
877 .llseek = generic_file_llseek, },
878 { }
879};
880
881static void xgene_edac_pmd_create_debugfs_nodes(
882 struct edac_device_ctl_info *edac_dev)
883{
884 struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
Borislav Petkov09bd1b42015-09-22 13:13:46 +0200885 struct dentry *dbgfs_dir;
Loc Ho0d442932015-05-22 17:32:59 -0600886 char name[30];
887
888 if (!IS_ENABLED(CONFIG_EDAC_DEBUG))
889 return;
890
891 /*
892 * Todo: Switch to common EDAC debug file system for edac device
893 * when available.
894 */
895 if (!ctx->edac->dfs) {
Borislav Petkov09bd1b42015-09-22 13:13:46 +0200896 ctx->edac->dfs = edac_debugfs_create_dir(edac_dev->dev->kobj.name);
Loc Ho0d442932015-05-22 17:32:59 -0600897 if (!ctx->edac->dfs)
898 return;
899 }
900 sprintf(name, "PMD%d", ctx->pmd);
Borislav Petkov09bd1b42015-09-22 13:13:46 +0200901 dbgfs_dir = edac_debugfs_create_dir_at(name, ctx->edac->dfs);
902 if (!dbgfs_dir)
Loc Ho0d442932015-05-22 17:32:59 -0600903 return;
904
Borislav Petkov09bd1b42015-09-22 13:13:46 +0200905 edac_debugfs_create_file("l1_inject_ctrl", S_IWUSR, dbgfs_dir, edac_dev,
906 &xgene_edac_pmd_debug_inject_fops[0]);
907 edac_debugfs_create_file("l2_inject_ctrl", S_IWUSR, dbgfs_dir, edac_dev,
908 &xgene_edac_pmd_debug_inject_fops[1]);
Loc Ho0d442932015-05-22 17:32:59 -0600909}
910
911static int xgene_edac_pmd_available(u32 efuse, int pmd)
912{
913 return (efuse & (1 << pmd)) ? 0 : 1;
914}
915
Arnd Bergmann451bb7f2015-06-01 16:09:35 -0600916static int xgene_edac_pmd_add(struct xgene_edac *edac, struct device_node *np,
917 int version)
Loc Ho0d442932015-05-22 17:32:59 -0600918{
919 struct edac_device_ctl_info *edac_dev;
920 struct xgene_edac_pmd_ctx *ctx;
921 struct resource res;
922 char edac_name[10];
923 u32 pmd;
924 int rc;
925 u32 val;
926
927 if (!devres_open_group(edac->dev, xgene_edac_pmd_add, GFP_KERNEL))
928 return -ENOMEM;
929
930 /* Determine if this PMD is disabled */
931 if (of_property_read_u32(np, "pmd-controller", &pmd)) {
932 dev_err(edac->dev, "no pmd-controller property\n");
933 rc = -ENODEV;
934 goto err_group;
935 }
936 rc = regmap_read(edac->efuse_map, 0, &val);
937 if (rc)
938 goto err_group;
939 if (!xgene_edac_pmd_available(val, pmd)) {
940 rc = -ENODEV;
941 goto err_group;
942 }
943
944 sprintf(edac_name, "l2c%d", pmd);
945 edac_dev = edac_device_alloc_ctl_info(sizeof(*ctx),
946 edac_name, 1, "l2c", 1, 2, NULL,
947 0, edac_device_alloc_index());
948 if (!edac_dev) {
949 rc = -ENOMEM;
950 goto err_group;
951 }
952
953 ctx = edac_dev->pvt_info;
954 ctx->name = "xgene_pmd_err";
955 ctx->pmd = pmd;
956 ctx->edac = edac;
957 ctx->edac_dev = edac_dev;
958 ctx->ddev = *edac->dev;
Arnd Bergmann451bb7f2015-06-01 16:09:35 -0600959 ctx->version = version;
Loc Ho0d442932015-05-22 17:32:59 -0600960 edac_dev->dev = &ctx->ddev;
961 edac_dev->ctl_name = ctx->name;
962 edac_dev->dev_name = ctx->name;
963 edac_dev->mod_name = EDAC_MOD_STR;
964
965 rc = of_address_to_resource(np, 0, &res);
966 if (rc < 0) {
967 dev_err(edac->dev, "no PMD resource address\n");
968 goto err_free;
969 }
970 ctx->pmd_csr = devm_ioremap_resource(edac->dev, &res);
971 if (IS_ERR(ctx->pmd_csr)) {
972 dev_err(edac->dev,
973 "devm_ioremap_resource failed for PMD resource address\n");
974 rc = PTR_ERR(ctx->pmd_csr);
975 goto err_free;
976 }
977
978 if (edac_op_state == EDAC_OPSTATE_POLL)
979 edac_dev->edac_check = xgene_edac_pmd_check;
980
981 xgene_edac_pmd_create_debugfs_nodes(edac_dev);
982
983 rc = edac_device_add_device(edac_dev);
984 if (rc > 0) {
985 dev_err(edac->dev, "edac_device_add_device failed\n");
986 rc = -ENOMEM;
987 goto err_free;
988 }
989
990 if (edac_op_state == EDAC_OPSTATE_INT)
991 edac_dev->op_state = OP_RUNNING_INTERRUPT;
992
993 list_add(&ctx->next, &edac->pmds);
994
995 xgene_edac_pmd_hw_ctl(edac_dev, 1);
996
997 devres_remove_group(edac->dev, xgene_edac_pmd_add);
998
999 dev_info(edac->dev, "X-Gene EDAC PMD%d registered\n", ctx->pmd);
1000 return 0;
1001
1002err_free:
1003 edac_device_free_ctl_info(edac_dev);
1004err_group:
1005 devres_release_group(edac->dev, xgene_edac_pmd_add);
1006 return rc;
1007}
1008
1009static int xgene_edac_pmd_remove(struct xgene_edac_pmd_ctx *pmd)
1010{
1011 struct edac_device_ctl_info *edac_dev = pmd->edac_dev;
1012
1013 xgene_edac_pmd_hw_ctl(edac_dev, 0);
1014 edac_device_del_device(edac_dev->dev);
1015 edac_device_free_ctl_info(edac_dev);
1016 return 0;
1017}
1018
1019static irqreturn_t xgene_edac_isr(int irq, void *dev_id)
1020{
1021 struct xgene_edac *ctx = dev_id;
1022 struct xgene_edac_pmd_ctx *pmd;
1023 unsigned int pcp_hp_stat;
1024 unsigned int pcp_lp_stat;
1025
1026 xgene_edac_pcp_rd(ctx, PCPHPERRINTSTS, &pcp_hp_stat);
1027 xgene_edac_pcp_rd(ctx, PCPLPERRINTSTS, &pcp_lp_stat);
1028 if ((MCU_UNCORR_ERR_MASK & pcp_hp_stat) ||
1029 (MCU_CTL_ERR_MASK & pcp_hp_stat) ||
1030 (MCU_CORR_ERR_MASK & pcp_lp_stat)) {
1031 struct xgene_edac_mc_ctx *mcu;
1032
1033 list_for_each_entry(mcu, &ctx->mcus, next) {
1034 xgene_edac_mc_check(mcu->mci);
1035 }
1036 }
1037
1038 list_for_each_entry(pmd, &ctx->pmds, next) {
1039 if ((PMD0_MERR_MASK << pmd->pmd) & pcp_hp_stat)
1040 xgene_edac_pmd_check(pmd->edac_dev);
1041 }
1042
1043 return IRQ_HANDLED;
1044}
1045
1046static int xgene_edac_probe(struct platform_device *pdev)
1047{
1048 struct xgene_edac *edac;
1049 struct device_node *child;
1050 struct resource *res;
1051 int rc;
1052
1053 edac = devm_kzalloc(&pdev->dev, sizeof(*edac), GFP_KERNEL);
1054 if (!edac)
1055 return -ENOMEM;
1056
1057 edac->dev = &pdev->dev;
1058 platform_set_drvdata(pdev, edac);
1059 INIT_LIST_HEAD(&edac->mcus);
1060 INIT_LIST_HEAD(&edac->pmds);
1061 spin_lock_init(&edac->lock);
1062 mutex_init(&edac->mc_lock);
1063
1064 edac->csw_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
1065 "regmap-csw");
1066 if (IS_ERR(edac->csw_map)) {
1067 dev_err(edac->dev, "unable to get syscon regmap csw\n");
1068 rc = PTR_ERR(edac->csw_map);
1069 goto out_err;
1070 }
1071
1072 edac->mcba_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
1073 "regmap-mcba");
1074 if (IS_ERR(edac->mcba_map)) {
1075 dev_err(edac->dev, "unable to get syscon regmap mcba\n");
1076 rc = PTR_ERR(edac->mcba_map);
1077 goto out_err;
1078 }
1079
1080 edac->mcbb_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
1081 "regmap-mcbb");
1082 if (IS_ERR(edac->mcbb_map)) {
1083 dev_err(edac->dev, "unable to get syscon regmap mcbb\n");
1084 rc = PTR_ERR(edac->mcbb_map);
1085 goto out_err;
1086 }
1087 edac->efuse_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
1088 "regmap-efuse");
1089 if (IS_ERR(edac->efuse_map)) {
1090 dev_err(edac->dev, "unable to get syscon regmap efuse\n");
1091 rc = PTR_ERR(edac->efuse_map);
1092 goto out_err;
1093 }
1094
1095 res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
1096 edac->pcp_csr = devm_ioremap_resource(&pdev->dev, res);
1097 if (IS_ERR(edac->pcp_csr)) {
1098 dev_err(&pdev->dev, "no PCP resource address\n");
1099 rc = PTR_ERR(edac->pcp_csr);
1100 goto out_err;
1101 }
1102
1103 if (edac_op_state == EDAC_OPSTATE_INT) {
1104 int irq;
1105 int i;
1106
1107 for (i = 0; i < 3; i++) {
1108 irq = platform_get_irq(pdev, i);
1109 if (irq < 0) {
1110 dev_err(&pdev->dev, "No IRQ resource\n");
1111 rc = -EINVAL;
1112 goto out_err;
1113 }
1114 rc = devm_request_irq(&pdev->dev, irq,
1115 xgene_edac_isr, IRQF_SHARED,
1116 dev_name(&pdev->dev), edac);
1117 if (rc) {
1118 dev_err(&pdev->dev,
1119 "Could not request IRQ %d\n", irq);
1120 goto out_err;
1121 }
1122 }
1123 }
1124
1125 for_each_child_of_node(pdev->dev.of_node, child) {
1126 if (!of_device_is_available(child))
1127 continue;
1128 if (of_device_is_compatible(child, "apm,xgene-edac-mc"))
1129 xgene_edac_mc_add(edac, child);
1130 if (of_device_is_compatible(child, "apm,xgene-edac-pmd"))
Arnd Bergmann451bb7f2015-06-01 16:09:35 -06001131 xgene_edac_pmd_add(edac, child, 1);
1132 if (of_device_is_compatible(child, "apm,xgene-edac-pmd-v2"))
1133 xgene_edac_pmd_add(edac, child, 2);
Loc Ho0d442932015-05-22 17:32:59 -06001134 }
1135
1136 return 0;
1137
1138out_err:
1139 return rc;
1140}
1141
1142static int xgene_edac_remove(struct platform_device *pdev)
1143{
1144 struct xgene_edac *edac = dev_get_drvdata(&pdev->dev);
1145 struct xgene_edac_mc_ctx *mcu;
1146 struct xgene_edac_mc_ctx *temp_mcu;
1147 struct xgene_edac_pmd_ctx *pmd;
1148 struct xgene_edac_pmd_ctx *temp_pmd;
1149
1150 list_for_each_entry_safe(mcu, temp_mcu, &edac->mcus, next) {
1151 xgene_edac_mc_remove(mcu);
1152 }
1153
1154 list_for_each_entry_safe(pmd, temp_pmd, &edac->pmds, next) {
1155 xgene_edac_pmd_remove(pmd);
1156 }
1157 return 0;
1158}
1159
1160static const struct of_device_id xgene_edac_of_match[] = {
1161 { .compatible = "apm,xgene-edac" },
1162 {},
1163};
1164MODULE_DEVICE_TABLE(of, xgene_edac_of_match);
1165
1166static struct platform_driver xgene_edac_driver = {
1167 .probe = xgene_edac_probe,
1168 .remove = xgene_edac_remove,
1169 .driver = {
1170 .name = "xgene-edac",
Loc Ho0d442932015-05-22 17:32:59 -06001171 .of_match_table = xgene_edac_of_match,
1172 },
1173};
1174
1175static int __init xgene_edac_init(void)
1176{
1177 int rc;
1178
1179 /* Make sure error reporting method is sane */
1180 switch (edac_op_state) {
1181 case EDAC_OPSTATE_POLL:
1182 case EDAC_OPSTATE_INT:
1183 break;
1184 default:
1185 edac_op_state = EDAC_OPSTATE_INT;
1186 break;
1187 }
1188
1189 rc = platform_driver_register(&xgene_edac_driver);
1190 if (rc) {
1191 edac_printk(KERN_ERR, EDAC_MOD_STR,
1192 "EDAC fails to register\n");
1193 goto reg_failed;
1194 }
1195
1196 return 0;
1197
1198reg_failed:
1199 return rc;
1200}
1201module_init(xgene_edac_init);
1202
1203static void __exit xgene_edac_exit(void)
1204{
1205 platform_driver_unregister(&xgene_edac_driver);
1206}
1207module_exit(xgene_edac_exit);
1208
1209MODULE_LICENSE("GPL");
1210MODULE_AUTHOR("Feng Kan <fkan@apm.com>");
1211MODULE_DESCRIPTION("APM X-Gene EDAC driver");
1212module_param(edac_op_state, int, 0444);
1213MODULE_PARM_DESC(edac_op_state,
1214 "EDAC error reporting state: 0=Poll, 2=Interrupt");