xref: /openbmc/linux/drivers/edac/npcm_edac.c (revision d7955ce4)
1 // SPDX-License-Identifier: GPL-2.0-only
2 // Copyright (c) 2022 Nuvoton Technology Corporation
3 
4 #include <linux/debugfs.h>
5 #include <linux/iopoll.h>
6 #include <linux/of_device.h>
7 #include <linux/regmap.h>
8 #include "edac_module.h"
9 
10 #define EDAC_MOD_NAME			"npcm-edac"
11 #define EDAC_MSG_SIZE			256
12 
13 /* chip serials */
14 #define NPCM7XX_CHIP			BIT(0)
15 #define NPCM8XX_CHIP			BIT(1)
16 
17 /* syndrome values */
18 #define UE_SYNDROME			0x03
19 
20 /* error injection */
21 #define ERROR_TYPE_CORRECTABLE		0
22 #define ERROR_TYPE_UNCORRECTABLE	1
23 #define ERROR_LOCATION_DATA		0
24 #define ERROR_LOCATION_CHECKCODE	1
25 #define ERROR_BIT_DATA_MAX		63
26 #define ERROR_BIT_CHECKCODE_MAX		7
27 
28 static char data_synd[] = {
29 	0xf4, 0xf1, 0xec, 0xea, 0xe9, 0xe6, 0xe5, 0xe3,
30 	0xdc, 0xda, 0xd9, 0xd6, 0xd5, 0xd3, 0xce, 0xcb,
31 	0xb5, 0xb0, 0xad, 0xab, 0xa8, 0xa7, 0xa4, 0xa2,
32 	0x9d, 0x9b, 0x98, 0x97, 0x94, 0x92, 0x8f, 0x8a,
33 	0x75, 0x70, 0x6d, 0x6b, 0x68, 0x67, 0x64, 0x62,
34 	0x5e, 0x5b, 0x58, 0x57, 0x54, 0x52, 0x4f, 0x4a,
35 	0x34, 0x31, 0x2c, 0x2a, 0x29, 0x26, 0x25, 0x23,
36 	0x1c, 0x1a, 0x19, 0x16, 0x15, 0x13, 0x0e, 0x0b
37 };
38 
39 static struct regmap *npcm_regmap;
40 
41 struct npcm_platform_data {
42 	/* chip serials */
43 	int chip;
44 
45 	/* memory controller registers */
46 	u32 ctl_ecc_en;
47 	u32 ctl_int_status;
48 	u32 ctl_int_ack;
49 	u32 ctl_int_mask_master;
50 	u32 ctl_int_mask_ecc;
51 	u32 ctl_ce_addr_l;
52 	u32 ctl_ce_addr_h;
53 	u32 ctl_ce_data_l;
54 	u32 ctl_ce_data_h;
55 	u32 ctl_ce_synd;
56 	u32 ctl_ue_addr_l;
57 	u32 ctl_ue_addr_h;
58 	u32 ctl_ue_data_l;
59 	u32 ctl_ue_data_h;
60 	u32 ctl_ue_synd;
61 	u32 ctl_source_id;
62 	u32 ctl_controller_busy;
63 	u32 ctl_xor_check_bits;
64 
65 	/* masks and shifts */
66 	u32 ecc_en_mask;
67 	u32 int_status_ce_mask;
68 	u32 int_status_ue_mask;
69 	u32 int_ack_ce_mask;
70 	u32 int_ack_ue_mask;
71 	u32 int_mask_master_non_ecc_mask;
72 	u32 int_mask_master_global_mask;
73 	u32 int_mask_ecc_non_event_mask;
74 	u32 ce_addr_h_mask;
75 	u32 ce_synd_mask;
76 	u32 ce_synd_shift;
77 	u32 ue_addr_h_mask;
78 	u32 ue_synd_mask;
79 	u32 ue_synd_shift;
80 	u32 source_id_ce_mask;
81 	u32 source_id_ce_shift;
82 	u32 source_id_ue_mask;
83 	u32 source_id_ue_shift;
84 	u32 controller_busy_mask;
85 	u32 xor_check_bits_mask;
86 	u32 xor_check_bits_shift;
87 	u32 writeback_en_mask;
88 	u32 fwc_mask;
89 };
90 
91 struct priv_data {
92 	void __iomem *reg;
93 	char message[EDAC_MSG_SIZE];
94 	const struct npcm_platform_data *pdata;
95 
96 	/* error injection */
97 	struct dentry *debugfs;
98 	u8 error_type;
99 	u8 location;
100 	u8 bit;
101 };
102 
103 static void handle_ce(struct mem_ctl_info *mci)
104 {
105 	struct priv_data *priv = mci->pvt_info;
106 	const struct npcm_platform_data *pdata;
107 	u32 val_h = 0, val_l, id, synd;
108 	u64 addr = 0, data = 0;
109 
110 	pdata = priv->pdata;
111 	regmap_read(npcm_regmap, pdata->ctl_ce_addr_l, &val_l);
112 	if (pdata->chip == NPCM8XX_CHIP) {
113 		regmap_read(npcm_regmap, pdata->ctl_ce_addr_h, &val_h);
114 		val_h &= pdata->ce_addr_h_mask;
115 	}
116 	addr = ((addr | val_h) << 32) | val_l;
117 
118 	regmap_read(npcm_regmap, pdata->ctl_ce_data_l, &val_l);
119 	if (pdata->chip == NPCM8XX_CHIP)
120 		regmap_read(npcm_regmap, pdata->ctl_ce_data_h, &val_h);
121 	data = ((data | val_h) << 32) | val_l;
122 
123 	regmap_read(npcm_regmap, pdata->ctl_source_id, &id);
124 	id = (id & pdata->source_id_ce_mask) >> pdata->source_id_ce_shift;
125 
126 	regmap_read(npcm_regmap, pdata->ctl_ce_synd, &synd);
127 	synd = (synd & pdata->ce_synd_mask) >> pdata->ce_synd_shift;
128 
129 	snprintf(priv->message, EDAC_MSG_SIZE,
130 		 "addr = 0x%llx, data = 0x%llx, id = 0x%x", addr, data, id);
131 
132 	edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1, addr >> PAGE_SHIFT,
133 			     addr & ~PAGE_MASK, synd, 0, 0, -1, priv->message, "");
134 }
135 
136 static void handle_ue(struct mem_ctl_info *mci)
137 {
138 	struct priv_data *priv = mci->pvt_info;
139 	const struct npcm_platform_data *pdata;
140 	u32 val_h = 0, val_l, id, synd;
141 	u64 addr = 0, data = 0;
142 
143 	pdata = priv->pdata;
144 	regmap_read(npcm_regmap, pdata->ctl_ue_addr_l, &val_l);
145 	if (pdata->chip == NPCM8XX_CHIP) {
146 		regmap_read(npcm_regmap, pdata->ctl_ue_addr_h, &val_h);
147 		val_h &= pdata->ue_addr_h_mask;
148 	}
149 	addr = ((addr | val_h) << 32) | val_l;
150 
151 	regmap_read(npcm_regmap, pdata->ctl_ue_data_l, &val_l);
152 	if (pdata->chip == NPCM8XX_CHIP)
153 		regmap_read(npcm_regmap, pdata->ctl_ue_data_h, &val_h);
154 	data = ((data | val_h) << 32) | val_l;
155 
156 	regmap_read(npcm_regmap, pdata->ctl_source_id, &id);
157 	id = (id & pdata->source_id_ue_mask) >> pdata->source_id_ue_shift;
158 
159 	regmap_read(npcm_regmap, pdata->ctl_ue_synd, &synd);
160 	synd = (synd & pdata->ue_synd_mask) >> pdata->ue_synd_shift;
161 
162 	snprintf(priv->message, EDAC_MSG_SIZE,
163 		 "addr = 0x%llx, data = 0x%llx, id = 0x%x", addr, data, id);
164 
165 	edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, addr >> PAGE_SHIFT,
166 			     addr & ~PAGE_MASK, synd, 0, 0, -1, priv->message, "");
167 }
168 
169 static irqreturn_t edac_ecc_isr(int irq, void *dev_id)
170 {
171 	const struct npcm_platform_data *pdata;
172 	struct mem_ctl_info *mci = dev_id;
173 	u32 status;
174 
175 	pdata = ((struct priv_data *)mci->pvt_info)->pdata;
176 	regmap_read(npcm_regmap, pdata->ctl_int_status, &status);
177 	if (status & pdata->int_status_ce_mask) {
178 		handle_ce(mci);
179 
180 		/* acknowledge the CE interrupt */
181 		regmap_write(npcm_regmap, pdata->ctl_int_ack,
182 			     pdata->int_ack_ce_mask);
183 		return IRQ_HANDLED;
184 	} else if (status & pdata->int_status_ue_mask) {
185 		handle_ue(mci);
186 
187 		/* acknowledge the UE interrupt */
188 		regmap_write(npcm_regmap, pdata->ctl_int_ack,
189 			     pdata->int_ack_ue_mask);
190 		return IRQ_HANDLED;
191 	}
192 
193 	WARN_ON_ONCE(1);
194 	return IRQ_NONE;
195 }
196 
197 static ssize_t force_ecc_error(struct file *file, const char __user *data,
198 			       size_t count, loff_t *ppos)
199 {
200 	struct device *dev = file->private_data;
201 	struct mem_ctl_info *mci = to_mci(dev);
202 	struct priv_data *priv = mci->pvt_info;
203 	const struct npcm_platform_data *pdata;
204 	u32 val, syndrome;
205 	int ret;
206 
207 	pdata = priv->pdata;
208 	edac_printk(KERN_INFO, EDAC_MOD_NAME,
209 		    "force an ECC error, type = %d, location = %d, bit = %d\n",
210 		    priv->error_type, priv->location, priv->bit);
211 
212 	/* ensure no pending writes */
213 	ret = regmap_read_poll_timeout(npcm_regmap, pdata->ctl_controller_busy,
214 				       val, !(val & pdata->controller_busy_mask),
215 				       1000, 10000);
216 	if (ret) {
217 		edac_printk(KERN_INFO, EDAC_MOD_NAME,
218 			    "wait pending writes timeout\n");
219 		return count;
220 	}
221 
222 	regmap_read(npcm_regmap, pdata->ctl_xor_check_bits, &val);
223 	val &= ~pdata->xor_check_bits_mask;
224 
225 	/* write syndrome to XOR_CHECK_BITS */
226 	if (priv->error_type == ERROR_TYPE_CORRECTABLE) {
227 		if (priv->location == ERROR_LOCATION_DATA &&
228 		    priv->bit > ERROR_BIT_DATA_MAX) {
229 			edac_printk(KERN_INFO, EDAC_MOD_NAME,
230 				    "data bit should not exceed %d (%d)\n",
231 				    ERROR_BIT_DATA_MAX, priv->bit);
232 			return count;
233 		}
234 
235 		if (priv->location == ERROR_LOCATION_CHECKCODE &&
236 		    priv->bit > ERROR_BIT_CHECKCODE_MAX) {
237 			edac_printk(KERN_INFO, EDAC_MOD_NAME,
238 				    "checkcode bit should not exceed %d (%d)\n",
239 				    ERROR_BIT_CHECKCODE_MAX, priv->bit);
240 			return count;
241 		}
242 
243 		syndrome = priv->location ? 1 << priv->bit
244 					  : data_synd[priv->bit];
245 
246 		regmap_write(npcm_regmap, pdata->ctl_xor_check_bits,
247 			     val | (syndrome << pdata->xor_check_bits_shift) |
248 			     pdata->writeback_en_mask);
249 	} else if (priv->error_type == ERROR_TYPE_UNCORRECTABLE) {
250 		regmap_write(npcm_regmap, pdata->ctl_xor_check_bits,
251 			     val | (UE_SYNDROME << pdata->xor_check_bits_shift));
252 	}
253 
254 	/* force write check */
255 	regmap_update_bits(npcm_regmap, pdata->ctl_xor_check_bits,
256 			   pdata->fwc_mask, pdata->fwc_mask);
257 
258 	return count;
259 }
260 
261 static const struct file_operations force_ecc_error_fops = {
262 	.open = simple_open,
263 	.write = force_ecc_error,
264 	.llseek = generic_file_llseek,
265 };
266 
267 /*
268  * Setup debugfs for error injection.
269  *
270  * Nodes:
271  *   error_type		- 0: CE, 1: UE
272  *   location		- 0: data, 1: checkcode
273  *   bit		- 0 ~ 63 for data and 0 ~ 7 for checkcode
274  *   force_ecc_error	- trigger
275  *
276  * Examples:
277  *   1. Inject a correctable error (CE) at checkcode bit 7.
278  *      ~# echo 0 > /sys/kernel/debug/edac/npcm-edac/error_type
279  *      ~# echo 1 > /sys/kernel/debug/edac/npcm-edac/location
280  *      ~# echo 7 > /sys/kernel/debug/edac/npcm-edac/bit
281  *      ~# echo 1 > /sys/kernel/debug/edac/npcm-edac/force_ecc_error
282  *
283  *   2. Inject an uncorrectable error (UE).
284  *      ~# echo 1 > /sys/kernel/debug/edac/npcm-edac/error_type
285  *      ~# echo 1 > /sys/kernel/debug/edac/npcm-edac/force_ecc_error
286  */
287 static void setup_debugfs(struct mem_ctl_info *mci)
288 {
289 	struct priv_data *priv = mci->pvt_info;
290 
291 	priv->debugfs = edac_debugfs_create_dir(mci->mod_name);
292 	if (!priv->debugfs)
293 		return;
294 
295 	edac_debugfs_create_x8("error_type", 0644, priv->debugfs, &priv->error_type);
296 	edac_debugfs_create_x8("location", 0644, priv->debugfs, &priv->location);
297 	edac_debugfs_create_x8("bit", 0644, priv->debugfs, &priv->bit);
298 	edac_debugfs_create_file("force_ecc_error", 0200, priv->debugfs,
299 				 &mci->dev, &force_ecc_error_fops);
300 }
301 
302 static int setup_irq(struct mem_ctl_info *mci, struct platform_device *pdev)
303 {
304 	const struct npcm_platform_data *pdata;
305 	int ret, irq;
306 
307 	pdata = ((struct priv_data *)mci->pvt_info)->pdata;
308 	irq = platform_get_irq(pdev, 0);
309 	if (irq < 0) {
310 		edac_printk(KERN_ERR, EDAC_MOD_NAME, "IRQ not defined in DTS\n");
311 		return irq;
312 	}
313 
314 	ret = devm_request_irq(&pdev->dev, irq, edac_ecc_isr, 0,
315 			       dev_name(&pdev->dev), mci);
316 	if (ret < 0) {
317 		edac_printk(KERN_ERR, EDAC_MOD_NAME, "failed to request IRQ\n");
318 		return ret;
319 	}
320 
321 	/* enable the functional group of ECC and mask the others */
322 	regmap_write(npcm_regmap, pdata->ctl_int_mask_master,
323 		     pdata->int_mask_master_non_ecc_mask);
324 
325 	if (pdata->chip == NPCM8XX_CHIP)
326 		regmap_write(npcm_regmap, pdata->ctl_int_mask_ecc,
327 			     pdata->int_mask_ecc_non_event_mask);
328 
329 	return 0;
330 }
331 
332 static const struct regmap_config npcm_regmap_cfg = {
333 	.reg_bits	= 32,
334 	.reg_stride	= 4,
335 	.val_bits	= 32,
336 };
337 
338 static int edac_probe(struct platform_device *pdev)
339 {
340 	const struct npcm_platform_data *pdata;
341 	struct device *dev = &pdev->dev;
342 	struct edac_mc_layer layers[1];
343 	struct mem_ctl_info *mci;
344 	struct priv_data *priv;
345 	void __iomem *reg;
346 	u32 val;
347 	int rc;
348 
349 	reg = devm_platform_ioremap_resource(pdev, 0);
350 	if (IS_ERR(reg))
351 		return PTR_ERR(reg);
352 
353 	npcm_regmap = devm_regmap_init_mmio(dev, reg, &npcm_regmap_cfg);
354 	if (IS_ERR(npcm_regmap))
355 		return PTR_ERR(npcm_regmap);
356 
357 	pdata = of_device_get_match_data(dev);
358 	if (!pdata)
359 		return -EINVAL;
360 
361 	/* bail out if ECC is not enabled */
362 	regmap_read(npcm_regmap, pdata->ctl_ecc_en, &val);
363 	if (!(val & pdata->ecc_en_mask)) {
364 		edac_printk(KERN_ERR, EDAC_MOD_NAME, "ECC is not enabled\n");
365 		return -EPERM;
366 	}
367 
368 	edac_op_state = EDAC_OPSTATE_INT;
369 
370 	layers[0].type = EDAC_MC_LAYER_ALL_MEM;
371 	layers[0].size = 1;
372 
373 	mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers,
374 			    sizeof(struct priv_data));
375 	if (!mci)
376 		return -ENOMEM;
377 
378 	mci->pdev = &pdev->dev;
379 	priv = mci->pvt_info;
380 	priv->reg = reg;
381 	priv->pdata = pdata;
382 	platform_set_drvdata(pdev, mci);
383 
384 	mci->mtype_cap = MEM_FLAG_DDR4;
385 	mci->edac_ctl_cap = EDAC_FLAG_SECDED;
386 	mci->scrub_cap = SCRUB_FLAG_HW_SRC;
387 	mci->scrub_mode = SCRUB_HW_SRC;
388 	mci->edac_cap = EDAC_FLAG_SECDED;
389 	mci->ctl_name = "npcm_ddr_controller";
390 	mci->dev_name = dev_name(&pdev->dev);
391 	mci->mod_name = EDAC_MOD_NAME;
392 	mci->ctl_page_to_phys = NULL;
393 
394 	rc = setup_irq(mci, pdev);
395 	if (rc)
396 		goto free_edac_mc;
397 
398 	rc = edac_mc_add_mc(mci);
399 	if (rc)
400 		goto free_edac_mc;
401 
402 	if (IS_ENABLED(CONFIG_EDAC_DEBUG) && pdata->chip == NPCM8XX_CHIP)
403 		setup_debugfs(mci);
404 
405 	return rc;
406 
407 free_edac_mc:
408 	edac_mc_free(mci);
409 	return rc;
410 }
411 
412 static int edac_remove(struct platform_device *pdev)
413 {
414 	struct mem_ctl_info *mci = platform_get_drvdata(pdev);
415 	struct priv_data *priv = mci->pvt_info;
416 	const struct npcm_platform_data *pdata;
417 
418 	pdata = priv->pdata;
419 	if (IS_ENABLED(CONFIG_EDAC_DEBUG) && pdata->chip == NPCM8XX_CHIP)
420 		edac_debugfs_remove_recursive(priv->debugfs);
421 
422 	edac_mc_del_mc(&pdev->dev);
423 	edac_mc_free(mci);
424 
425 	regmap_write(npcm_regmap, pdata->ctl_int_mask_master,
426 		     pdata->int_mask_master_global_mask);
427 	regmap_update_bits(npcm_regmap, pdata->ctl_ecc_en, pdata->ecc_en_mask, 0);
428 
429 	return 0;
430 }
431 
432 static const struct npcm_platform_data npcm750_edac = {
433 	.chip				= NPCM7XX_CHIP,
434 
435 	/* memory controller registers */
436 	.ctl_ecc_en			= 0x174,
437 	.ctl_int_status			= 0x1d0,
438 	.ctl_int_ack			= 0x1d4,
439 	.ctl_int_mask_master		= 0x1d8,
440 	.ctl_ce_addr_l			= 0x188,
441 	.ctl_ce_data_l			= 0x190,
442 	.ctl_ce_synd			= 0x18c,
443 	.ctl_ue_addr_l			= 0x17c,
444 	.ctl_ue_data_l			= 0x184,
445 	.ctl_ue_synd			= 0x180,
446 	.ctl_source_id			= 0x194,
447 
448 	/* masks and shifts */
449 	.ecc_en_mask			= BIT(24),
450 	.int_status_ce_mask		= GENMASK(4, 3),
451 	.int_status_ue_mask		= GENMASK(6, 5),
452 	.int_ack_ce_mask		= GENMASK(4, 3),
453 	.int_ack_ue_mask		= GENMASK(6, 5),
454 	.int_mask_master_non_ecc_mask	= GENMASK(30, 7) | GENMASK(2, 0),
455 	.int_mask_master_global_mask	= BIT(31),
456 	.ce_synd_mask			= GENMASK(6, 0),
457 	.ce_synd_shift			= 0,
458 	.ue_synd_mask			= GENMASK(6, 0),
459 	.ue_synd_shift			= 0,
460 	.source_id_ce_mask		= GENMASK(29, 16),
461 	.source_id_ce_shift		= 16,
462 	.source_id_ue_mask		= GENMASK(13, 0),
463 	.source_id_ue_shift		= 0,
464 };
465 
466 static const struct npcm_platform_data npcm845_edac = {
467 	.chip =				NPCM8XX_CHIP,
468 
469 	/* memory controller registers */
470 	.ctl_ecc_en			= 0x16c,
471 	.ctl_int_status			= 0x228,
472 	.ctl_int_ack			= 0x244,
473 	.ctl_int_mask_master		= 0x220,
474 	.ctl_int_mask_ecc		= 0x260,
475 	.ctl_ce_addr_l			= 0x18c,
476 	.ctl_ce_addr_h			= 0x190,
477 	.ctl_ce_data_l			= 0x194,
478 	.ctl_ce_data_h			= 0x198,
479 	.ctl_ce_synd			= 0x190,
480 	.ctl_ue_addr_l			= 0x17c,
481 	.ctl_ue_addr_h			= 0x180,
482 	.ctl_ue_data_l			= 0x184,
483 	.ctl_ue_data_h			= 0x188,
484 	.ctl_ue_synd			= 0x180,
485 	.ctl_source_id			= 0x19c,
486 	.ctl_controller_busy		= 0x20c,
487 	.ctl_xor_check_bits		= 0x174,
488 
489 	/* masks and shifts */
490 	.ecc_en_mask			= GENMASK(17, 16),
491 	.int_status_ce_mask		= GENMASK(1, 0),
492 	.int_status_ue_mask		= GENMASK(3, 2),
493 	.int_ack_ce_mask		= GENMASK(1, 0),
494 	.int_ack_ue_mask		= GENMASK(3, 2),
495 	.int_mask_master_non_ecc_mask	= GENMASK(30, 3) | GENMASK(1, 0),
496 	.int_mask_master_global_mask	= BIT(31),
497 	.int_mask_ecc_non_event_mask	= GENMASK(8, 4),
498 	.ce_addr_h_mask			= GENMASK(1, 0),
499 	.ce_synd_mask			= GENMASK(15, 8),
500 	.ce_synd_shift			= 8,
501 	.ue_addr_h_mask			= GENMASK(1, 0),
502 	.ue_synd_mask			= GENMASK(15, 8),
503 	.ue_synd_shift			= 8,
504 	.source_id_ce_mask		= GENMASK(29, 16),
505 	.source_id_ce_shift		= 16,
506 	.source_id_ue_mask		= GENMASK(13, 0),
507 	.source_id_ue_shift		= 0,
508 	.controller_busy_mask		= BIT(0),
509 	.xor_check_bits_mask		= GENMASK(23, 16),
510 	.xor_check_bits_shift		= 16,
511 	.writeback_en_mask		= BIT(24),
512 	.fwc_mask			= BIT(8),
513 };
514 
515 static const struct of_device_id npcm_edac_of_match[] = {
516 	{
517 		.compatible = "nuvoton,npcm750-memory-controller",
518 		.data = &npcm750_edac
519 	},
520 	{
521 		.compatible = "nuvoton,npcm845-memory-controller",
522 		.data = &npcm845_edac
523 	},
524 	{},
525 };
526 
527 MODULE_DEVICE_TABLE(of, npcm_edac_of_match);
528 
529 static struct platform_driver npcm_edac_driver = {
530 	.driver = {
531 		.name = "npcm-edac",
532 		.of_match_table = npcm_edac_of_match,
533 	},
534 	.probe = edac_probe,
535 	.remove = edac_remove,
536 };
537 
538 module_platform_driver(npcm_edac_driver);
539 
540 MODULE_AUTHOR("Medad CChien <medadyoung@gmail.com>");
541 MODULE_AUTHOR("Marvin Lin <kflin@nuvoton.com>");
542 MODULE_DESCRIPTION("Nuvoton NPCM EDAC Driver");
543 MODULE_LICENSE("GPL");
544