xref: /openbmc/linux/drivers/edac/npcm_edac.c (revision df1c357f)
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.h>
7 #include <linux/platform_device.h>
8 #include <linux/regmap.h>
9 #include "edac_module.h"
10 
11 #define EDAC_MOD_NAME			"npcm-edac"
12 #define EDAC_MSG_SIZE			256
13 
14 /* chip serials */
15 #define NPCM7XX_CHIP			BIT(0)
16 #define NPCM8XX_CHIP			BIT(1)
17 
18 /* syndrome values */
19 #define UE_SYNDROME			0x03
20 
21 /* error injection */
22 #define ERROR_TYPE_CORRECTABLE		0
23 #define ERROR_TYPE_UNCORRECTABLE	1
24 #define ERROR_LOCATION_DATA		0
25 #define ERROR_LOCATION_CHECKCODE	1
26 #define ERROR_BIT_DATA_MAX		63
27 #define ERROR_BIT_CHECKCODE_MAX		7
28 
29 static char data_synd[] = {
30 	0xf4, 0xf1, 0xec, 0xea, 0xe9, 0xe6, 0xe5, 0xe3,
31 	0xdc, 0xda, 0xd9, 0xd6, 0xd5, 0xd3, 0xce, 0xcb,
32 	0xb5, 0xb0, 0xad, 0xab, 0xa8, 0xa7, 0xa4, 0xa2,
33 	0x9d, 0x9b, 0x98, 0x97, 0x94, 0x92, 0x8f, 0x8a,
34 	0x75, 0x70, 0x6d, 0x6b, 0x68, 0x67, 0x64, 0x62,
35 	0x5e, 0x5b, 0x58, 0x57, 0x54, 0x52, 0x4f, 0x4a,
36 	0x34, 0x31, 0x2c, 0x2a, 0x29, 0x26, 0x25, 0x23,
37 	0x1c, 0x1a, 0x19, 0x16, 0x15, 0x13, 0x0e, 0x0b
38 };
39 
40 static struct regmap *npcm_regmap;
41 
42 struct npcm_platform_data {
43 	/* chip serials */
44 	int chip;
45 
46 	/* memory controller registers */
47 	u32 ctl_ecc_en;
48 	u32 ctl_int_status;
49 	u32 ctl_int_ack;
50 	u32 ctl_int_mask_master;
51 	u32 ctl_int_mask_ecc;
52 	u32 ctl_ce_addr_l;
53 	u32 ctl_ce_addr_h;
54 	u32 ctl_ce_data_l;
55 	u32 ctl_ce_data_h;
56 	u32 ctl_ce_synd;
57 	u32 ctl_ue_addr_l;
58 	u32 ctl_ue_addr_h;
59 	u32 ctl_ue_data_l;
60 	u32 ctl_ue_data_h;
61 	u32 ctl_ue_synd;
62 	u32 ctl_source_id;
63 	u32 ctl_controller_busy;
64 	u32 ctl_xor_check_bits;
65 
66 	/* masks and shifts */
67 	u32 ecc_en_mask;
68 	u32 int_status_ce_mask;
69 	u32 int_status_ue_mask;
70 	u32 int_ack_ce_mask;
71 	u32 int_ack_ue_mask;
72 	u32 int_mask_master_non_ecc_mask;
73 	u32 int_mask_master_global_mask;
74 	u32 int_mask_ecc_non_event_mask;
75 	u32 ce_addr_h_mask;
76 	u32 ce_synd_mask;
77 	u32 ce_synd_shift;
78 	u32 ue_addr_h_mask;
79 	u32 ue_synd_mask;
80 	u32 ue_synd_shift;
81 	u32 source_id_ce_mask;
82 	u32 source_id_ce_shift;
83 	u32 source_id_ue_mask;
84 	u32 source_id_ue_shift;
85 	u32 controller_busy_mask;
86 	u32 xor_check_bits_mask;
87 	u32 xor_check_bits_shift;
88 	u32 writeback_en_mask;
89 	u32 fwc_mask;
90 };
91 
92 struct priv_data {
93 	void __iomem *reg;
94 	char message[EDAC_MSG_SIZE];
95 	const struct npcm_platform_data *pdata;
96 
97 	/* error injection */
98 	struct dentry *debugfs;
99 	u8 error_type;
100 	u8 location;
101 	u8 bit;
102 };
103 
104 static void handle_ce(struct mem_ctl_info *mci)
105 {
106 	struct priv_data *priv = mci->pvt_info;
107 	const struct npcm_platform_data *pdata;
108 	u32 val_h = 0, val_l, id, synd;
109 	u64 addr = 0, data = 0;
110 
111 	pdata = priv->pdata;
112 	regmap_read(npcm_regmap, pdata->ctl_ce_addr_l, &val_l);
113 	if (pdata->chip == NPCM8XX_CHIP) {
114 		regmap_read(npcm_regmap, pdata->ctl_ce_addr_h, &val_h);
115 		val_h &= pdata->ce_addr_h_mask;
116 	}
117 	addr = ((addr | val_h) << 32) | val_l;
118 
119 	regmap_read(npcm_regmap, pdata->ctl_ce_data_l, &val_l);
120 	if (pdata->chip == NPCM8XX_CHIP)
121 		regmap_read(npcm_regmap, pdata->ctl_ce_data_h, &val_h);
122 	data = ((data | val_h) << 32) | val_l;
123 
124 	regmap_read(npcm_regmap, pdata->ctl_source_id, &id);
125 	id = (id & pdata->source_id_ce_mask) >> pdata->source_id_ce_shift;
126 
127 	regmap_read(npcm_regmap, pdata->ctl_ce_synd, &synd);
128 	synd = (synd & pdata->ce_synd_mask) >> pdata->ce_synd_shift;
129 
130 	snprintf(priv->message, EDAC_MSG_SIZE,
131 		 "addr = 0x%llx, data = 0x%llx, id = 0x%x", addr, data, id);
132 
133 	edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1, addr >> PAGE_SHIFT,
134 			     addr & ~PAGE_MASK, synd, 0, 0, -1, priv->message, "");
135 }
136 
137 static void handle_ue(struct mem_ctl_info *mci)
138 {
139 	struct priv_data *priv = mci->pvt_info;
140 	const struct npcm_platform_data *pdata;
141 	u32 val_h = 0, val_l, id, synd;
142 	u64 addr = 0, data = 0;
143 
144 	pdata = priv->pdata;
145 	regmap_read(npcm_regmap, pdata->ctl_ue_addr_l, &val_l);
146 	if (pdata->chip == NPCM8XX_CHIP) {
147 		regmap_read(npcm_regmap, pdata->ctl_ue_addr_h, &val_h);
148 		val_h &= pdata->ue_addr_h_mask;
149 	}
150 	addr = ((addr | val_h) << 32) | val_l;
151 
152 	regmap_read(npcm_regmap, pdata->ctl_ue_data_l, &val_l);
153 	if (pdata->chip == NPCM8XX_CHIP)
154 		regmap_read(npcm_regmap, pdata->ctl_ue_data_h, &val_h);
155 	data = ((data | val_h) << 32) | val_l;
156 
157 	regmap_read(npcm_regmap, pdata->ctl_source_id, &id);
158 	id = (id & pdata->source_id_ue_mask) >> pdata->source_id_ue_shift;
159 
160 	regmap_read(npcm_regmap, pdata->ctl_ue_synd, &synd);
161 	synd = (synd & pdata->ue_synd_mask) >> pdata->ue_synd_shift;
162 
163 	snprintf(priv->message, EDAC_MSG_SIZE,
164 		 "addr = 0x%llx, data = 0x%llx, id = 0x%x", addr, data, id);
165 
166 	edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, addr >> PAGE_SHIFT,
167 			     addr & ~PAGE_MASK, synd, 0, 0, -1, priv->message, "");
168 }
169 
170 static irqreturn_t edac_ecc_isr(int irq, void *dev_id)
171 {
172 	const struct npcm_platform_data *pdata;
173 	struct mem_ctl_info *mci = dev_id;
174 	u32 status;
175 
176 	pdata = ((struct priv_data *)mci->pvt_info)->pdata;
177 	regmap_read(npcm_regmap, pdata->ctl_int_status, &status);
178 	if (status & pdata->int_status_ce_mask) {
179 		handle_ce(mci);
180 
181 		/* acknowledge the CE interrupt */
182 		regmap_write(npcm_regmap, pdata->ctl_int_ack,
183 			     pdata->int_ack_ce_mask);
184 		return IRQ_HANDLED;
185 	} else if (status & pdata->int_status_ue_mask) {
186 		handle_ue(mci);
187 
188 		/* acknowledge the UE interrupt */
189 		regmap_write(npcm_regmap, pdata->ctl_int_ack,
190 			     pdata->int_ack_ue_mask);
191 		return IRQ_HANDLED;
192 	}
193 
194 	WARN_ON_ONCE(1);
195 	return IRQ_NONE;
196 }
197 
198 static ssize_t force_ecc_error(struct file *file, const char __user *data,
199 			       size_t count, loff_t *ppos)
200 {
201 	struct device *dev = file->private_data;
202 	struct mem_ctl_info *mci = to_mci(dev);
203 	struct priv_data *priv = mci->pvt_info;
204 	const struct npcm_platform_data *pdata;
205 	u32 val, syndrome;
206 	int ret;
207 
208 	pdata = priv->pdata;
209 	edac_printk(KERN_INFO, EDAC_MOD_NAME,
210 		    "force an ECC error, type = %d, location = %d, bit = %d\n",
211 		    priv->error_type, priv->location, priv->bit);
212 
213 	/* ensure no pending writes */
214 	ret = regmap_read_poll_timeout(npcm_regmap, pdata->ctl_controller_busy,
215 				       val, !(val & pdata->controller_busy_mask),
216 				       1000, 10000);
217 	if (ret) {
218 		edac_printk(KERN_INFO, EDAC_MOD_NAME,
219 			    "wait pending writes timeout\n");
220 		return count;
221 	}
222 
223 	regmap_read(npcm_regmap, pdata->ctl_xor_check_bits, &val);
224 	val &= ~pdata->xor_check_bits_mask;
225 
226 	/* write syndrome to XOR_CHECK_BITS */
227 	if (priv->error_type == ERROR_TYPE_CORRECTABLE) {
228 		if (priv->location == ERROR_LOCATION_DATA &&
229 		    priv->bit > ERROR_BIT_DATA_MAX) {
230 			edac_printk(KERN_INFO, EDAC_MOD_NAME,
231 				    "data bit should not exceed %d (%d)\n",
232 				    ERROR_BIT_DATA_MAX, priv->bit);
233 			return count;
234 		}
235 
236 		if (priv->location == ERROR_LOCATION_CHECKCODE &&
237 		    priv->bit > ERROR_BIT_CHECKCODE_MAX) {
238 			edac_printk(KERN_INFO, EDAC_MOD_NAME,
239 				    "checkcode bit should not exceed %d (%d)\n",
240 				    ERROR_BIT_CHECKCODE_MAX, priv->bit);
241 			return count;
242 		}
243 
244 		syndrome = priv->location ? 1 << priv->bit
245 					  : data_synd[priv->bit];
246 
247 		regmap_write(npcm_regmap, pdata->ctl_xor_check_bits,
248 			     val | (syndrome << pdata->xor_check_bits_shift) |
249 			     pdata->writeback_en_mask);
250 	} else if (priv->error_type == ERROR_TYPE_UNCORRECTABLE) {
251 		regmap_write(npcm_regmap, pdata->ctl_xor_check_bits,
252 			     val | (UE_SYNDROME << pdata->xor_check_bits_shift));
253 	}
254 
255 	/* force write check */
256 	regmap_update_bits(npcm_regmap, pdata->ctl_xor_check_bits,
257 			   pdata->fwc_mask, pdata->fwc_mask);
258 
259 	return count;
260 }
261 
262 static const struct file_operations force_ecc_error_fops = {
263 	.open = simple_open,
264 	.write = force_ecc_error,
265 	.llseek = generic_file_llseek,
266 };
267 
268 /*
269  * Setup debugfs for error injection.
270  *
271  * Nodes:
272  *   error_type		- 0: CE, 1: UE
273  *   location		- 0: data, 1: checkcode
274  *   bit		- 0 ~ 63 for data and 0 ~ 7 for checkcode
275  *   force_ecc_error	- trigger
276  *
277  * Examples:
278  *   1. Inject a correctable error (CE) at checkcode bit 7.
279  *      ~# echo 0 > /sys/kernel/debug/edac/npcm-edac/error_type
280  *      ~# echo 1 > /sys/kernel/debug/edac/npcm-edac/location
281  *      ~# echo 7 > /sys/kernel/debug/edac/npcm-edac/bit
282  *      ~# echo 1 > /sys/kernel/debug/edac/npcm-edac/force_ecc_error
283  *
284  *   2. Inject an uncorrectable error (UE).
285  *      ~# echo 1 > /sys/kernel/debug/edac/npcm-edac/error_type
286  *      ~# echo 1 > /sys/kernel/debug/edac/npcm-edac/force_ecc_error
287  */
288 static void setup_debugfs(struct mem_ctl_info *mci)
289 {
290 	struct priv_data *priv = mci->pvt_info;
291 
292 	priv->debugfs = edac_debugfs_create_dir(mci->mod_name);
293 	if (!priv->debugfs)
294 		return;
295 
296 	edac_debugfs_create_x8("error_type", 0644, priv->debugfs, &priv->error_type);
297 	edac_debugfs_create_x8("location", 0644, priv->debugfs, &priv->location);
298 	edac_debugfs_create_x8("bit", 0644, priv->debugfs, &priv->bit);
299 	edac_debugfs_create_file("force_ecc_error", 0200, priv->debugfs,
300 				 &mci->dev, &force_ecc_error_fops);
301 }
302 
303 static int setup_irq(struct mem_ctl_info *mci, struct platform_device *pdev)
304 {
305 	const struct npcm_platform_data *pdata;
306 	int ret, irq;
307 
308 	pdata = ((struct priv_data *)mci->pvt_info)->pdata;
309 	irq = platform_get_irq(pdev, 0);
310 	if (irq < 0) {
311 		edac_printk(KERN_ERR, EDAC_MOD_NAME, "IRQ not defined in DTS\n");
312 		return irq;
313 	}
314 
315 	ret = devm_request_irq(&pdev->dev, irq, edac_ecc_isr, 0,
316 			       dev_name(&pdev->dev), mci);
317 	if (ret < 0) {
318 		edac_printk(KERN_ERR, EDAC_MOD_NAME, "failed to request IRQ\n");
319 		return ret;
320 	}
321 
322 	/* enable the functional group of ECC and mask the others */
323 	regmap_write(npcm_regmap, pdata->ctl_int_mask_master,
324 		     pdata->int_mask_master_non_ecc_mask);
325 
326 	if (pdata->chip == NPCM8XX_CHIP)
327 		regmap_write(npcm_regmap, pdata->ctl_int_mask_ecc,
328 			     pdata->int_mask_ecc_non_event_mask);
329 
330 	return 0;
331 }
332 
333 static const struct regmap_config npcm_regmap_cfg = {
334 	.reg_bits	= 32,
335 	.reg_stride	= 4,
336 	.val_bits	= 32,
337 };
338 
339 static int edac_probe(struct platform_device *pdev)
340 {
341 	const struct npcm_platform_data *pdata;
342 	struct device *dev = &pdev->dev;
343 	struct edac_mc_layer layers[1];
344 	struct mem_ctl_info *mci;
345 	struct priv_data *priv;
346 	void __iomem *reg;
347 	u32 val;
348 	int rc;
349 
350 	reg = devm_platform_ioremap_resource(pdev, 0);
351 	if (IS_ERR(reg))
352 		return PTR_ERR(reg);
353 
354 	npcm_regmap = devm_regmap_init_mmio(dev, reg, &npcm_regmap_cfg);
355 	if (IS_ERR(npcm_regmap))
356 		return PTR_ERR(npcm_regmap);
357 
358 	pdata = of_device_get_match_data(dev);
359 	if (!pdata)
360 		return -EINVAL;
361 
362 	/* bail out if ECC is not enabled */
363 	regmap_read(npcm_regmap, pdata->ctl_ecc_en, &val);
364 	if (!(val & pdata->ecc_en_mask)) {
365 		edac_printk(KERN_ERR, EDAC_MOD_NAME, "ECC is not enabled\n");
366 		return -EPERM;
367 	}
368 
369 	edac_op_state = EDAC_OPSTATE_INT;
370 
371 	layers[0].type = EDAC_MC_LAYER_ALL_MEM;
372 	layers[0].size = 1;
373 
374 	mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers,
375 			    sizeof(struct priv_data));
376 	if (!mci)
377 		return -ENOMEM;
378 
379 	mci->pdev = &pdev->dev;
380 	priv = mci->pvt_info;
381 	priv->reg = reg;
382 	priv->pdata = pdata;
383 	platform_set_drvdata(pdev, mci);
384 
385 	mci->mtype_cap = MEM_FLAG_DDR4;
386 	mci->edac_ctl_cap = EDAC_FLAG_SECDED;
387 	mci->scrub_cap = SCRUB_FLAG_HW_SRC;
388 	mci->scrub_mode = SCRUB_HW_SRC;
389 	mci->edac_cap = EDAC_FLAG_SECDED;
390 	mci->ctl_name = "npcm_ddr_controller";
391 	mci->dev_name = dev_name(&pdev->dev);
392 	mci->mod_name = EDAC_MOD_NAME;
393 	mci->ctl_page_to_phys = NULL;
394 
395 	rc = setup_irq(mci, pdev);
396 	if (rc)
397 		goto free_edac_mc;
398 
399 	rc = edac_mc_add_mc(mci);
400 	if (rc)
401 		goto free_edac_mc;
402 
403 	if (IS_ENABLED(CONFIG_EDAC_DEBUG) && pdata->chip == NPCM8XX_CHIP)
404 		setup_debugfs(mci);
405 
406 	return rc;
407 
408 free_edac_mc:
409 	edac_mc_free(mci);
410 	return rc;
411 }
412 
413 static int edac_remove(struct platform_device *pdev)
414 {
415 	struct mem_ctl_info *mci = platform_get_drvdata(pdev);
416 	struct priv_data *priv = mci->pvt_info;
417 	const struct npcm_platform_data *pdata;
418 
419 	pdata = priv->pdata;
420 	if (IS_ENABLED(CONFIG_EDAC_DEBUG) && pdata->chip == NPCM8XX_CHIP)
421 		edac_debugfs_remove_recursive(priv->debugfs);
422 
423 	edac_mc_del_mc(&pdev->dev);
424 	edac_mc_free(mci);
425 
426 	regmap_write(npcm_regmap, pdata->ctl_int_mask_master,
427 		     pdata->int_mask_master_global_mask);
428 	regmap_update_bits(npcm_regmap, pdata->ctl_ecc_en, pdata->ecc_en_mask, 0);
429 
430 	return 0;
431 }
432 
433 static const struct npcm_platform_data npcm750_edac = {
434 	.chip				= NPCM7XX_CHIP,
435 
436 	/* memory controller registers */
437 	.ctl_ecc_en			= 0x174,
438 	.ctl_int_status			= 0x1d0,
439 	.ctl_int_ack			= 0x1d4,
440 	.ctl_int_mask_master		= 0x1d8,
441 	.ctl_ce_addr_l			= 0x188,
442 	.ctl_ce_data_l			= 0x190,
443 	.ctl_ce_synd			= 0x18c,
444 	.ctl_ue_addr_l			= 0x17c,
445 	.ctl_ue_data_l			= 0x184,
446 	.ctl_ue_synd			= 0x180,
447 	.ctl_source_id			= 0x194,
448 
449 	/* masks and shifts */
450 	.ecc_en_mask			= BIT(24),
451 	.int_status_ce_mask		= GENMASK(4, 3),
452 	.int_status_ue_mask		= GENMASK(6, 5),
453 	.int_ack_ce_mask		= GENMASK(4, 3),
454 	.int_ack_ue_mask		= GENMASK(6, 5),
455 	.int_mask_master_non_ecc_mask	= GENMASK(30, 7) | GENMASK(2, 0),
456 	.int_mask_master_global_mask	= BIT(31),
457 	.ce_synd_mask			= GENMASK(6, 0),
458 	.ce_synd_shift			= 0,
459 	.ue_synd_mask			= GENMASK(6, 0),
460 	.ue_synd_shift			= 0,
461 	.source_id_ce_mask		= GENMASK(29, 16),
462 	.source_id_ce_shift		= 16,
463 	.source_id_ue_mask		= GENMASK(13, 0),
464 	.source_id_ue_shift		= 0,
465 };
466 
467 static const struct npcm_platform_data npcm845_edac = {
468 	.chip =				NPCM8XX_CHIP,
469 
470 	/* memory controller registers */
471 	.ctl_ecc_en			= 0x16c,
472 	.ctl_int_status			= 0x228,
473 	.ctl_int_ack			= 0x244,
474 	.ctl_int_mask_master		= 0x220,
475 	.ctl_int_mask_ecc		= 0x260,
476 	.ctl_ce_addr_l			= 0x18c,
477 	.ctl_ce_addr_h			= 0x190,
478 	.ctl_ce_data_l			= 0x194,
479 	.ctl_ce_data_h			= 0x198,
480 	.ctl_ce_synd			= 0x190,
481 	.ctl_ue_addr_l			= 0x17c,
482 	.ctl_ue_addr_h			= 0x180,
483 	.ctl_ue_data_l			= 0x184,
484 	.ctl_ue_data_h			= 0x188,
485 	.ctl_ue_synd			= 0x180,
486 	.ctl_source_id			= 0x19c,
487 	.ctl_controller_busy		= 0x20c,
488 	.ctl_xor_check_bits		= 0x174,
489 
490 	/* masks and shifts */
491 	.ecc_en_mask			= GENMASK(17, 16),
492 	.int_status_ce_mask		= GENMASK(1, 0),
493 	.int_status_ue_mask		= GENMASK(3, 2),
494 	.int_ack_ce_mask		= GENMASK(1, 0),
495 	.int_ack_ue_mask		= GENMASK(3, 2),
496 	.int_mask_master_non_ecc_mask	= GENMASK(30, 3) | GENMASK(1, 0),
497 	.int_mask_master_global_mask	= BIT(31),
498 	.int_mask_ecc_non_event_mask	= GENMASK(8, 4),
499 	.ce_addr_h_mask			= GENMASK(1, 0),
500 	.ce_synd_mask			= GENMASK(15, 8),
501 	.ce_synd_shift			= 8,
502 	.ue_addr_h_mask			= GENMASK(1, 0),
503 	.ue_synd_mask			= GENMASK(15, 8),
504 	.ue_synd_shift			= 8,
505 	.source_id_ce_mask		= GENMASK(29, 16),
506 	.source_id_ce_shift		= 16,
507 	.source_id_ue_mask		= GENMASK(13, 0),
508 	.source_id_ue_shift		= 0,
509 	.controller_busy_mask		= BIT(0),
510 	.xor_check_bits_mask		= GENMASK(23, 16),
511 	.xor_check_bits_shift		= 16,
512 	.writeback_en_mask		= BIT(24),
513 	.fwc_mask			= BIT(8),
514 };
515 
516 static const struct of_device_id npcm_edac_of_match[] = {
517 	{
518 		.compatible = "nuvoton,npcm750-memory-controller",
519 		.data = &npcm750_edac
520 	},
521 	{
522 		.compatible = "nuvoton,npcm845-memory-controller",
523 		.data = &npcm845_edac
524 	},
525 	{},
526 };
527 
528 MODULE_DEVICE_TABLE(of, npcm_edac_of_match);
529 
530 static struct platform_driver npcm_edac_driver = {
531 	.driver = {
532 		.name = "npcm-edac",
533 		.of_match_table = npcm_edac_of_match,
534 	},
535 	.probe = edac_probe,
536 	.remove = edac_remove,
537 };
538 
539 module_platform_driver(npcm_edac_driver);
540 
541 MODULE_AUTHOR("Medad CChien <medadyoung@gmail.com>");
542 MODULE_AUTHOR("Marvin Lin <kflin@nuvoton.com>");
543 MODULE_DESCRIPTION("Nuvoton NPCM EDAC Driver");
544 MODULE_LICENSE("GPL");
545