1 /*
2 * cpc925_edac.c, EDAC driver for IBM CPC925 Bridge and Memory Controller.
3 *
4 * Copyright (c) 2008 Wind River Systems, Inc.
5 *
6 * Authors: Cao Qingtao <qingtao.cao@windriver.com>
7 *
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 as
10 * published by the Free Software Foundation.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
15 * See the GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 */
21
22 #include <linux/module.h>
23 #include <linux/init.h>
24 #include <linux/io.h>
25 #include <linux/edac.h>
26 #include <linux/of.h>
27 #include <linux/platform_device.h>
28 #include <linux/gfp.h>
29
30 #include "edac_module.h"
31
32 #define CPC925_EDAC_REVISION " Ver: 1.0.0"
33 #define CPC925_EDAC_MOD_STR "cpc925_edac"
34
35 #define cpc925_printk(level, fmt, arg...) \
36 edac_printk(level, "CPC925", fmt, ##arg)
37
38 #define cpc925_mc_printk(mci, level, fmt, arg...) \
39 edac_mc_chipset_printk(mci, level, "CPC925", fmt, ##arg)
40
41 /*
42 * CPC925 registers are of 32 bits with bit0 defined at the
43 * most significant bit and bit31 at that of least significant.
44 */
45 #define CPC925_BITS_PER_REG 32
46 #define CPC925_BIT(nr) (1UL << (CPC925_BITS_PER_REG - 1 - nr))
47
48 /*
49 * EDAC device names for the error detections of
50 * CPU Interface and Hypertransport Link.
51 */
52 #define CPC925_CPU_ERR_DEV "cpu"
53 #define CPC925_HT_LINK_DEV "htlink"
54
55 /* Suppose DDR Refresh cycle is 15.6 microsecond */
56 #define CPC925_REF_FREQ 0xFA69
57 #define CPC925_SCRUB_BLOCK_SIZE 64 /* bytes */
58 #define CPC925_NR_CSROWS 8
59
60 /*
61 * All registers and bits definitions are taken from
62 * "CPC925 Bridge and Memory Controller User Manual, SA14-2761-02".
63 */
64
65 /*
66 * CPU and Memory Controller Registers
67 */
68 /************************************************************
69 * Processor Interface Exception Mask Register (APIMASK)
70 ************************************************************/
71 #define REG_APIMASK_OFFSET 0x30070
72 enum apimask_bits {
73 APIMASK_DART = CPC925_BIT(0), /* DART Exception */
74 APIMASK_ADI0 = CPC925_BIT(1), /* Handshake Error on PI0_ADI */
75 APIMASK_ADI1 = CPC925_BIT(2), /* Handshake Error on PI1_ADI */
76 APIMASK_STAT = CPC925_BIT(3), /* Status Exception */
77 APIMASK_DERR = CPC925_BIT(4), /* Data Error Exception */
78 APIMASK_ADRS0 = CPC925_BIT(5), /* Addressing Exception on PI0 */
79 APIMASK_ADRS1 = CPC925_BIT(6), /* Addressing Exception on PI1 */
80 /* BIT(7) Reserved */
81 APIMASK_ECC_UE_H = CPC925_BIT(8), /* UECC upper */
82 APIMASK_ECC_CE_H = CPC925_BIT(9), /* CECC upper */
83 APIMASK_ECC_UE_L = CPC925_BIT(10), /* UECC lower */
84 APIMASK_ECC_CE_L = CPC925_BIT(11), /* CECC lower */
85
86 CPU_MASK_ENABLE = (APIMASK_DART | APIMASK_ADI0 | APIMASK_ADI1 |
87 APIMASK_STAT | APIMASK_DERR | APIMASK_ADRS0 |
88 APIMASK_ADRS1),
89 ECC_MASK_ENABLE = (APIMASK_ECC_UE_H | APIMASK_ECC_CE_H |
90 APIMASK_ECC_UE_L | APIMASK_ECC_CE_L),
91 };
92 #define APIMASK_ADI(n) CPC925_BIT(((n)+1))
93
94 /************************************************************
95 * Processor Interface Exception Register (APIEXCP)
96 ************************************************************/
97 #define REG_APIEXCP_OFFSET 0x30060
98 enum apiexcp_bits {
99 APIEXCP_DART = CPC925_BIT(0), /* DART Exception */
100 APIEXCP_ADI0 = CPC925_BIT(1), /* Handshake Error on PI0_ADI */
101 APIEXCP_ADI1 = CPC925_BIT(2), /* Handshake Error on PI1_ADI */
102 APIEXCP_STAT = CPC925_BIT(3), /* Status Exception */
103 APIEXCP_DERR = CPC925_BIT(4), /* Data Error Exception */
104 APIEXCP_ADRS0 = CPC925_BIT(5), /* Addressing Exception on PI0 */
105 APIEXCP_ADRS1 = CPC925_BIT(6), /* Addressing Exception on PI1 */
106 /* BIT(7) Reserved */
107 APIEXCP_ECC_UE_H = CPC925_BIT(8), /* UECC upper */
108 APIEXCP_ECC_CE_H = CPC925_BIT(9), /* CECC upper */
109 APIEXCP_ECC_UE_L = CPC925_BIT(10), /* UECC lower */
110 APIEXCP_ECC_CE_L = CPC925_BIT(11), /* CECC lower */
111
112 CPU_EXCP_DETECTED = (APIEXCP_DART | APIEXCP_ADI0 | APIEXCP_ADI1 |
113 APIEXCP_STAT | APIEXCP_DERR | APIEXCP_ADRS0 |
114 APIEXCP_ADRS1),
115 UECC_EXCP_DETECTED = (APIEXCP_ECC_UE_H | APIEXCP_ECC_UE_L),
116 CECC_EXCP_DETECTED = (APIEXCP_ECC_CE_H | APIEXCP_ECC_CE_L),
117 ECC_EXCP_DETECTED = (UECC_EXCP_DETECTED | CECC_EXCP_DETECTED),
118 };
119
120 /************************************************************
121 * Memory Bus Configuration Register (MBCR)
122 ************************************************************/
123 #define REG_MBCR_OFFSET 0x2190
124 #define MBCR_64BITCFG_SHIFT 23
125 #define MBCR_64BITCFG_MASK (1UL << MBCR_64BITCFG_SHIFT)
126 #define MBCR_64BITBUS_SHIFT 22
127 #define MBCR_64BITBUS_MASK (1UL << MBCR_64BITBUS_SHIFT)
128
129 /************************************************************
130 * Memory Bank Mode Register (MBMR)
131 ************************************************************/
132 #define REG_MBMR_OFFSET 0x21C0
133 #define MBMR_MODE_MAX_VALUE 0xF
134 #define MBMR_MODE_SHIFT 25
135 #define MBMR_MODE_MASK (MBMR_MODE_MAX_VALUE << MBMR_MODE_SHIFT)
136 #define MBMR_BBA_SHIFT 24
137 #define MBMR_BBA_MASK (1UL << MBMR_BBA_SHIFT)
138
139 /************************************************************
140 * Memory Bank Boundary Address Register (MBBAR)
141 ************************************************************/
142 #define REG_MBBAR_OFFSET 0x21D0
143 #define MBBAR_BBA_MAX_VALUE 0xFF
144 #define MBBAR_BBA_SHIFT 24
145 #define MBBAR_BBA_MASK (MBBAR_BBA_MAX_VALUE << MBBAR_BBA_SHIFT)
146
147 /************************************************************
148 * Memory Scrub Control Register (MSCR)
149 ************************************************************/
150 #define REG_MSCR_OFFSET 0x2400
151 #define MSCR_SCRUB_MOD_MASK 0xC0000000 /* scrub_mod - bit0:1*/
152 #define MSCR_BACKGR_SCRUB 0x40000000 /* 01 */
153 #define MSCR_SI_SHIFT 16 /* si - bit8:15*/
154 #define MSCR_SI_MAX_VALUE 0xFF
155 #define MSCR_SI_MASK (MSCR_SI_MAX_VALUE << MSCR_SI_SHIFT)
156
157 /************************************************************
158 * Memory Scrub Range Start Register (MSRSR)
159 ************************************************************/
160 #define REG_MSRSR_OFFSET 0x2410
161
162 /************************************************************
163 * Memory Scrub Range End Register (MSRER)
164 ************************************************************/
165 #define REG_MSRER_OFFSET 0x2420
166
167 /************************************************************
168 * Memory Scrub Pattern Register (MSPR)
169 ************************************************************/
170 #define REG_MSPR_OFFSET 0x2430
171
172 /************************************************************
173 * Memory Check Control Register (MCCR)
174 ************************************************************/
175 #define REG_MCCR_OFFSET 0x2440
176 enum mccr_bits {
177 MCCR_ECC_EN = CPC925_BIT(0), /* ECC high and low check */
178 };
179
180 /************************************************************
181 * Memory Check Range End Register (MCRER)
182 ************************************************************/
183 #define REG_MCRER_OFFSET 0x2450
184
185 /************************************************************
186 * Memory Error Address Register (MEAR)
187 ************************************************************/
188 #define REG_MEAR_OFFSET 0x2460
189 #define MEAR_BCNT_MAX_VALUE 0x3
190 #define MEAR_BCNT_SHIFT 30
191 #define MEAR_BCNT_MASK (MEAR_BCNT_MAX_VALUE << MEAR_BCNT_SHIFT)
192 #define MEAR_RANK_MAX_VALUE 0x7
193 #define MEAR_RANK_SHIFT 27
194 #define MEAR_RANK_MASK (MEAR_RANK_MAX_VALUE << MEAR_RANK_SHIFT)
195 #define MEAR_COL_MAX_VALUE 0x7FF
196 #define MEAR_COL_SHIFT 16
197 #define MEAR_COL_MASK (MEAR_COL_MAX_VALUE << MEAR_COL_SHIFT)
198 #define MEAR_BANK_MAX_VALUE 0x3
199 #define MEAR_BANK_SHIFT 14
200 #define MEAR_BANK_MASK (MEAR_BANK_MAX_VALUE << MEAR_BANK_SHIFT)
201 #define MEAR_ROW_MASK 0x00003FFF
202
203 /************************************************************
204 * Memory Error Syndrome Register (MESR)
205 ************************************************************/
206 #define REG_MESR_OFFSET 0x2470
207 #define MESR_ECC_SYN_H_MASK 0xFF00
208 #define MESR_ECC_SYN_L_MASK 0x00FF
209
210 /************************************************************
211 * Memory Mode Control Register (MMCR)
212 ************************************************************/
213 #define REG_MMCR_OFFSET 0x2500
214 enum mmcr_bits {
215 MMCR_REG_DIMM_MODE = CPC925_BIT(3),
216 };
217
218 /*
219 * HyperTransport Link Registers
220 */
221 /************************************************************
222 * Error Handling/Enumeration Scratch Pad Register (ERRCTRL)
223 ************************************************************/
224 #define REG_ERRCTRL_OFFSET 0x70140
225 enum errctrl_bits { /* nonfatal interrupts for */
226 ERRCTRL_SERR_NF = CPC925_BIT(0), /* system error */
227 ERRCTRL_CRC_NF = CPC925_BIT(1), /* CRC error */
228 ERRCTRL_RSP_NF = CPC925_BIT(2), /* Response error */
229 ERRCTRL_EOC_NF = CPC925_BIT(3), /* End-Of-Chain error */
230 ERRCTRL_OVF_NF = CPC925_BIT(4), /* Overflow error */
231 ERRCTRL_PROT_NF = CPC925_BIT(5), /* Protocol error */
232
233 ERRCTRL_RSP_ERR = CPC925_BIT(6), /* Response error received */
234 ERRCTRL_CHN_FAL = CPC925_BIT(7), /* Sync flooding detected */
235
236 HT_ERRCTRL_ENABLE = (ERRCTRL_SERR_NF | ERRCTRL_CRC_NF |
237 ERRCTRL_RSP_NF | ERRCTRL_EOC_NF |
238 ERRCTRL_OVF_NF | ERRCTRL_PROT_NF),
239 HT_ERRCTRL_DETECTED = (ERRCTRL_RSP_ERR | ERRCTRL_CHN_FAL),
240 };
241
242 /************************************************************
243 * Link Configuration and Link Control Register (LINKCTRL)
244 ************************************************************/
245 #define REG_LINKCTRL_OFFSET 0x70110
246 enum linkctrl_bits {
247 LINKCTRL_CRC_ERR = (CPC925_BIT(22) | CPC925_BIT(23)),
248 LINKCTRL_LINK_FAIL = CPC925_BIT(27),
249
250 HT_LINKCTRL_DETECTED = (LINKCTRL_CRC_ERR | LINKCTRL_LINK_FAIL),
251 };
252
253 /************************************************************
254 * Link FreqCap/Error/Freq/Revision ID Register (LINKERR)
255 ************************************************************/
256 #define REG_LINKERR_OFFSET 0x70120
257 enum linkerr_bits {
258 LINKERR_EOC_ERR = CPC925_BIT(17), /* End-Of-Chain error */
259 LINKERR_OVF_ERR = CPC925_BIT(18), /* Receive Buffer Overflow */
260 LINKERR_PROT_ERR = CPC925_BIT(19), /* Protocol error */
261
262 HT_LINKERR_DETECTED = (LINKERR_EOC_ERR | LINKERR_OVF_ERR |
263 LINKERR_PROT_ERR),
264 };
265
266 /************************************************************
267 * Bridge Control Register (BRGCTRL)
268 ************************************************************/
269 #define REG_BRGCTRL_OFFSET 0x70300
270 enum brgctrl_bits {
271 BRGCTRL_DETSERR = CPC925_BIT(0), /* SERR on Secondary Bus */
272 BRGCTRL_SECBUSRESET = CPC925_BIT(9), /* Secondary Bus Reset */
273 };
274
275 /* Private structure for edac memory controller */
276 struct cpc925_mc_pdata {
277 void __iomem *vbase;
278 unsigned long total_mem;
279 const char *name;
280 int edac_idx;
281 };
282
283 /* Private structure for common edac device */
284 struct cpc925_dev_info {
285 void __iomem *vbase;
286 struct platform_device *pdev;
287 char *ctl_name;
288 int edac_idx;
289 struct edac_device_ctl_info *edac_dev;
290 void (*init)(struct cpc925_dev_info *dev_info);
291 void (*exit)(struct cpc925_dev_info *dev_info);
292 void (*check)(struct edac_device_ctl_info *edac_dev);
293 };
294
295 /* Get total memory size from Open Firmware DTB */
get_total_mem(struct cpc925_mc_pdata * pdata)296 static void get_total_mem(struct cpc925_mc_pdata *pdata)
297 {
298 struct device_node *np = NULL;
299 const unsigned int *reg, *reg_end;
300 int len, sw, aw;
301 unsigned long start, size;
302
303 np = of_find_node_by_type(NULL, "memory");
304 if (!np)
305 return;
306
307 aw = of_n_addr_cells(np);
308 sw = of_n_size_cells(np);
309 reg = (const unsigned int *)of_get_property(np, "reg", &len);
310 reg_end = reg + len/4;
311
312 pdata->total_mem = 0;
313 do {
314 start = of_read_number(reg, aw);
315 reg += aw;
316 size = of_read_number(reg, sw);
317 reg += sw;
318 edac_dbg(1, "start 0x%lx, size 0x%lx\n", start, size);
319 pdata->total_mem += size;
320 } while (reg < reg_end);
321
322 of_node_put(np);
323 edac_dbg(0, "total_mem 0x%lx\n", pdata->total_mem);
324 }
325
cpc925_init_csrows(struct mem_ctl_info * mci)326 static void cpc925_init_csrows(struct mem_ctl_info *mci)
327 {
328 struct cpc925_mc_pdata *pdata = mci->pvt_info;
329 struct csrow_info *csrow;
330 struct dimm_info *dimm;
331 enum dev_type dtype;
332 int index, j;
333 u32 mbmr, mbbar, bba, grain;
334 unsigned long row_size, nr_pages, last_nr_pages = 0;
335
336 get_total_mem(pdata);
337
338 for (index = 0; index < mci->nr_csrows; index++) {
339 mbmr = __raw_readl(pdata->vbase + REG_MBMR_OFFSET +
340 0x20 * index);
341 mbbar = __raw_readl(pdata->vbase + REG_MBBAR_OFFSET +
342 0x20 + index);
343 bba = (((mbmr & MBMR_BBA_MASK) >> MBMR_BBA_SHIFT) << 8) |
344 ((mbbar & MBBAR_BBA_MASK) >> MBBAR_BBA_SHIFT);
345
346 if (bba == 0)
347 continue; /* not populated */
348
349 csrow = mci->csrows[index];
350
351 row_size = bba * (1UL << 28); /* 256M */
352 csrow->first_page = last_nr_pages;
353 nr_pages = row_size >> PAGE_SHIFT;
354 csrow->last_page = csrow->first_page + nr_pages - 1;
355 last_nr_pages = csrow->last_page + 1;
356
357 switch (csrow->nr_channels) {
358 case 1: /* Single channel */
359 grain = 32; /* four-beat burst of 32 bytes */
360 break;
361 case 2: /* Dual channel */
362 default:
363 grain = 64; /* four-beat burst of 64 bytes */
364 break;
365 }
366 switch ((mbmr & MBMR_MODE_MASK) >> MBMR_MODE_SHIFT) {
367 case 6: /* 0110, no way to differentiate X8 VS X16 */
368 case 5: /* 0101 */
369 case 8: /* 1000 */
370 dtype = DEV_X16;
371 break;
372 case 7: /* 0111 */
373 case 9: /* 1001 */
374 dtype = DEV_X8;
375 break;
376 default:
377 dtype = DEV_UNKNOWN;
378 break;
379 }
380 for (j = 0; j < csrow->nr_channels; j++) {
381 dimm = csrow->channels[j]->dimm;
382 dimm->nr_pages = nr_pages / csrow->nr_channels;
383 dimm->mtype = MEM_RDDR;
384 dimm->edac_mode = EDAC_SECDED;
385 dimm->grain = grain;
386 dimm->dtype = dtype;
387 }
388 }
389 }
390
391 /* Enable memory controller ECC detection */
cpc925_mc_init(struct mem_ctl_info * mci)392 static void cpc925_mc_init(struct mem_ctl_info *mci)
393 {
394 struct cpc925_mc_pdata *pdata = mci->pvt_info;
395 u32 apimask;
396 u32 mccr;
397
398 /* Enable various ECC error exceptions */
399 apimask = __raw_readl(pdata->vbase + REG_APIMASK_OFFSET);
400 if ((apimask & ECC_MASK_ENABLE) == 0) {
401 apimask |= ECC_MASK_ENABLE;
402 __raw_writel(apimask, pdata->vbase + REG_APIMASK_OFFSET);
403 }
404
405 /* Enable ECC detection */
406 mccr = __raw_readl(pdata->vbase + REG_MCCR_OFFSET);
407 if ((mccr & MCCR_ECC_EN) == 0) {
408 mccr |= MCCR_ECC_EN;
409 __raw_writel(mccr, pdata->vbase + REG_MCCR_OFFSET);
410 }
411 }
412
413 /* Disable memory controller ECC detection */
cpc925_mc_exit(struct mem_ctl_info * mci)414 static void cpc925_mc_exit(struct mem_ctl_info *mci)
415 {
416 /*
417 * WARNING:
418 * We are supposed to clear the ECC error detection bits,
419 * and it will be no problem to do so. However, once they
420 * are cleared here if we want to re-install CPC925 EDAC
421 * module later, setting them up in cpc925_mc_init() will
422 * trigger machine check exception.
423 * Also, it's ok to leave ECC error detection bits enabled,
424 * since they are reset to 1 by default or by boot loader.
425 */
426
427 return;
428 }
429
430 /*
431 * Revert DDR column/row/bank addresses into page frame number and
432 * offset in page.
433 *
434 * Suppose memory mode is 0x0111(128-bit mode, identical DIMM pairs),
435 * physical address(PA) bits to column address(CA) bits mappings are:
436 * CA 0 1 2 3 4 5 6 7 8 9 10
437 * PA 59 58 57 56 55 54 53 52 51 50 49
438 *
439 * physical address(PA) bits to bank address(BA) bits mappings are:
440 * BA 0 1
441 * PA 43 44
442 *
443 * physical address(PA) bits to row address(RA) bits mappings are:
444 * RA 0 1 2 3 4 5 6 7 8 9 10 11 12
445 * PA 36 35 34 48 47 46 45 40 41 42 39 38 37
446 */
cpc925_mc_get_pfn(struct mem_ctl_info * mci,u32 mear,unsigned long * pfn,unsigned long * offset,int * csrow)447 static void cpc925_mc_get_pfn(struct mem_ctl_info *mci, u32 mear,
448 unsigned long *pfn, unsigned long *offset, int *csrow)
449 {
450 u32 bcnt, rank, col, bank, row;
451 u32 c;
452 unsigned long pa;
453 int i;
454
455 bcnt = (mear & MEAR_BCNT_MASK) >> MEAR_BCNT_SHIFT;
456 rank = (mear & MEAR_RANK_MASK) >> MEAR_RANK_SHIFT;
457 col = (mear & MEAR_COL_MASK) >> MEAR_COL_SHIFT;
458 bank = (mear & MEAR_BANK_MASK) >> MEAR_BANK_SHIFT;
459 row = mear & MEAR_ROW_MASK;
460
461 *csrow = rank;
462
463 #ifdef CONFIG_EDAC_DEBUG
464 if (mci->csrows[rank]->first_page == 0) {
465 cpc925_mc_printk(mci, KERN_ERR, "ECC occurs in a "
466 "non-populated csrow, broken hardware?\n");
467 return;
468 }
469 #endif
470
471 /* Revert csrow number */
472 pa = mci->csrows[rank]->first_page << PAGE_SHIFT;
473
474 /* Revert column address */
475 col += bcnt;
476 for (i = 0; i < 11; i++) {
477 c = col & 0x1;
478 col >>= 1;
479 pa |= c << (14 - i);
480 }
481
482 /* Revert bank address */
483 pa |= bank << 19;
484
485 /* Revert row address, in 4 steps */
486 for (i = 0; i < 3; i++) {
487 c = row & 0x1;
488 row >>= 1;
489 pa |= c << (26 - i);
490 }
491
492 for (i = 0; i < 3; i++) {
493 c = row & 0x1;
494 row >>= 1;
495 pa |= c << (21 + i);
496 }
497
498 for (i = 0; i < 4; i++) {
499 c = row & 0x1;
500 row >>= 1;
501 pa |= c << (18 - i);
502 }
503
504 for (i = 0; i < 3; i++) {
505 c = row & 0x1;
506 row >>= 1;
507 pa |= c << (29 - i);
508 }
509
510 *offset = pa & (PAGE_SIZE - 1);
511 *pfn = pa >> PAGE_SHIFT;
512
513 edac_dbg(0, "ECC physical address 0x%lx\n", pa);
514 }
515
cpc925_mc_find_channel(struct mem_ctl_info * mci,u16 syndrome)516 static int cpc925_mc_find_channel(struct mem_ctl_info *mci, u16 syndrome)
517 {
518 if ((syndrome & MESR_ECC_SYN_H_MASK) == 0)
519 return 0;
520
521 if ((syndrome & MESR_ECC_SYN_L_MASK) == 0)
522 return 1;
523
524 cpc925_mc_printk(mci, KERN_INFO, "Unexpected syndrome value: 0x%x\n",
525 syndrome);
526 return 1;
527 }
528
529 /* Check memory controller registers for ECC errors */
cpc925_mc_check(struct mem_ctl_info * mci)530 static void cpc925_mc_check(struct mem_ctl_info *mci)
531 {
532 struct cpc925_mc_pdata *pdata = mci->pvt_info;
533 u32 apiexcp;
534 u32 mear;
535 u32 mesr;
536 u16 syndrome;
537 unsigned long pfn = 0, offset = 0;
538 int csrow = 0, channel = 0;
539
540 /* APIEXCP is cleared when read */
541 apiexcp = __raw_readl(pdata->vbase + REG_APIEXCP_OFFSET);
542 if ((apiexcp & ECC_EXCP_DETECTED) == 0)
543 return;
544
545 mesr = __raw_readl(pdata->vbase + REG_MESR_OFFSET);
546 syndrome = mesr | (MESR_ECC_SYN_H_MASK | MESR_ECC_SYN_L_MASK);
547
548 mear = __raw_readl(pdata->vbase + REG_MEAR_OFFSET);
549
550 /* Revert column/row addresses into page frame number, etc */
551 cpc925_mc_get_pfn(mci, mear, &pfn, &offset, &csrow);
552
553 if (apiexcp & CECC_EXCP_DETECTED) {
554 cpc925_mc_printk(mci, KERN_INFO, "DRAM CECC Fault\n");
555 channel = cpc925_mc_find_channel(mci, syndrome);
556 edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
557 pfn, offset, syndrome,
558 csrow, channel, -1,
559 mci->ctl_name, "");
560 }
561
562 if (apiexcp & UECC_EXCP_DETECTED) {
563 cpc925_mc_printk(mci, KERN_INFO, "DRAM UECC Fault\n");
564 edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
565 pfn, offset, 0,
566 csrow, -1, -1,
567 mci->ctl_name, "");
568 }
569
570 cpc925_mc_printk(mci, KERN_INFO, "Dump registers:\n");
571 cpc925_mc_printk(mci, KERN_INFO, "APIMASK 0x%08x\n",
572 __raw_readl(pdata->vbase + REG_APIMASK_OFFSET));
573 cpc925_mc_printk(mci, KERN_INFO, "APIEXCP 0x%08x\n",
574 apiexcp);
575 cpc925_mc_printk(mci, KERN_INFO, "Mem Scrub Ctrl 0x%08x\n",
576 __raw_readl(pdata->vbase + REG_MSCR_OFFSET));
577 cpc925_mc_printk(mci, KERN_INFO, "Mem Scrub Rge Start 0x%08x\n",
578 __raw_readl(pdata->vbase + REG_MSRSR_OFFSET));
579 cpc925_mc_printk(mci, KERN_INFO, "Mem Scrub Rge End 0x%08x\n",
580 __raw_readl(pdata->vbase + REG_MSRER_OFFSET));
581 cpc925_mc_printk(mci, KERN_INFO, "Mem Scrub Pattern 0x%08x\n",
582 __raw_readl(pdata->vbase + REG_MSPR_OFFSET));
583 cpc925_mc_printk(mci, KERN_INFO, "Mem Chk Ctrl 0x%08x\n",
584 __raw_readl(pdata->vbase + REG_MCCR_OFFSET));
585 cpc925_mc_printk(mci, KERN_INFO, "Mem Chk Rge End 0x%08x\n",
586 __raw_readl(pdata->vbase + REG_MCRER_OFFSET));
587 cpc925_mc_printk(mci, KERN_INFO, "Mem Err Address 0x%08x\n",
588 mesr);
589 cpc925_mc_printk(mci, KERN_INFO, "Mem Err Syndrome 0x%08x\n",
590 syndrome);
591 }
592
593 /******************** CPU err device********************************/
cpc925_cpu_mask_disabled(void)594 static u32 cpc925_cpu_mask_disabled(void)
595 {
596 struct device_node *cpus;
597 struct device_node *cpunode = NULL;
598 static u32 mask = 0;
599
600 /* use cached value if available */
601 if (mask != 0)
602 return mask;
603
604 mask = APIMASK_ADI0 | APIMASK_ADI1;
605
606 cpus = of_find_node_by_path("/cpus");
607 if (cpus == NULL) {
608 cpc925_printk(KERN_DEBUG, "No /cpus node !\n");
609 return 0;
610 }
611
612 while ((cpunode = of_get_next_child(cpus, cpunode)) != NULL) {
613 const u32 *reg = of_get_property(cpunode, "reg", NULL);
614
615 if (strcmp(cpunode->type, "cpu")) {
616 cpc925_printk(KERN_ERR, "Not a cpu node in /cpus: %s\n", cpunode->name);
617 continue;
618 }
619
620 if (reg == NULL || *reg > 2) {
621 cpc925_printk(KERN_ERR, "Bad reg value at %pOF\n", cpunode);
622 continue;
623 }
624
625 mask &= ~APIMASK_ADI(*reg);
626 }
627
628 if (mask != (APIMASK_ADI0 | APIMASK_ADI1)) {
629 /* We assume that each CPU sits on it's own PI and that
630 * for present CPUs the reg property equals to the PI
631 * interface id */
632 cpc925_printk(KERN_WARNING,
633 "Assuming PI id is equal to CPU MPIC id!\n");
634 }
635
636 of_node_put(cpunode);
637 of_node_put(cpus);
638
639 return mask;
640 }
641
642 /* Enable CPU Errors detection */
cpc925_cpu_init(struct cpc925_dev_info * dev_info)643 static void cpc925_cpu_init(struct cpc925_dev_info *dev_info)
644 {
645 u32 apimask;
646 u32 cpumask;
647
648 apimask = __raw_readl(dev_info->vbase + REG_APIMASK_OFFSET);
649
650 cpumask = cpc925_cpu_mask_disabled();
651 if (apimask & cpumask) {
652 cpc925_printk(KERN_WARNING, "CPU(s) not present, "
653 "but enabled in APIMASK, disabling\n");
654 apimask &= ~cpumask;
655 }
656
657 if ((apimask & CPU_MASK_ENABLE) == 0)
658 apimask |= CPU_MASK_ENABLE;
659
660 __raw_writel(apimask, dev_info->vbase + REG_APIMASK_OFFSET);
661 }
662
663 /* Disable CPU Errors detection */
cpc925_cpu_exit(struct cpc925_dev_info * dev_info)664 static void cpc925_cpu_exit(struct cpc925_dev_info *dev_info)
665 {
666 /*
667 * WARNING:
668 * We are supposed to clear the CPU error detection bits,
669 * and it will be no problem to do so. However, once they
670 * are cleared here if we want to re-install CPC925 EDAC
671 * module later, setting them up in cpc925_cpu_init() will
672 * trigger machine check exception.
673 * Also, it's ok to leave CPU error detection bits enabled,
674 * since they are reset to 1 by default.
675 */
676
677 return;
678 }
679
680 /* Check for CPU Errors */
cpc925_cpu_check(struct edac_device_ctl_info * edac_dev)681 static void cpc925_cpu_check(struct edac_device_ctl_info *edac_dev)
682 {
683 struct cpc925_dev_info *dev_info = edac_dev->pvt_info;
684 u32 apiexcp;
685 u32 apimask;
686
687 /* APIEXCP is cleared when read */
688 apiexcp = __raw_readl(dev_info->vbase + REG_APIEXCP_OFFSET);
689 if ((apiexcp & CPU_EXCP_DETECTED) == 0)
690 return;
691
692 if ((apiexcp & ~cpc925_cpu_mask_disabled()) == 0)
693 return;
694
695 apimask = __raw_readl(dev_info->vbase + REG_APIMASK_OFFSET);
696 cpc925_printk(KERN_INFO, "Processor Interface Fault\n"
697 "Processor Interface register dump:\n");
698 cpc925_printk(KERN_INFO, "APIMASK 0x%08x\n", apimask);
699 cpc925_printk(KERN_INFO, "APIEXCP 0x%08x\n", apiexcp);
700
701 edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
702 }
703
704 /******************** HT Link err device****************************/
705 /* Enable HyperTransport Link Error detection */
cpc925_htlink_init(struct cpc925_dev_info * dev_info)706 static void cpc925_htlink_init(struct cpc925_dev_info *dev_info)
707 {
708 u32 ht_errctrl;
709
710 ht_errctrl = __raw_readl(dev_info->vbase + REG_ERRCTRL_OFFSET);
711 if ((ht_errctrl & HT_ERRCTRL_ENABLE) == 0) {
712 ht_errctrl |= HT_ERRCTRL_ENABLE;
713 __raw_writel(ht_errctrl, dev_info->vbase + REG_ERRCTRL_OFFSET);
714 }
715 }
716
717 /* Disable HyperTransport Link Error detection */
cpc925_htlink_exit(struct cpc925_dev_info * dev_info)718 static void cpc925_htlink_exit(struct cpc925_dev_info *dev_info)
719 {
720 u32 ht_errctrl;
721
722 ht_errctrl = __raw_readl(dev_info->vbase + REG_ERRCTRL_OFFSET);
723 ht_errctrl &= ~HT_ERRCTRL_ENABLE;
724 __raw_writel(ht_errctrl, dev_info->vbase + REG_ERRCTRL_OFFSET);
725 }
726
727 /* Check for HyperTransport Link errors */
cpc925_htlink_check(struct edac_device_ctl_info * edac_dev)728 static void cpc925_htlink_check(struct edac_device_ctl_info *edac_dev)
729 {
730 struct cpc925_dev_info *dev_info = edac_dev->pvt_info;
731 u32 brgctrl = __raw_readl(dev_info->vbase + REG_BRGCTRL_OFFSET);
732 u32 linkctrl = __raw_readl(dev_info->vbase + REG_LINKCTRL_OFFSET);
733 u32 errctrl = __raw_readl(dev_info->vbase + REG_ERRCTRL_OFFSET);
734 u32 linkerr = __raw_readl(dev_info->vbase + REG_LINKERR_OFFSET);
735
736 if (!((brgctrl & BRGCTRL_DETSERR) ||
737 (linkctrl & HT_LINKCTRL_DETECTED) ||
738 (errctrl & HT_ERRCTRL_DETECTED) ||
739 (linkerr & HT_LINKERR_DETECTED)))
740 return;
741
742 cpc925_printk(KERN_INFO, "HT Link Fault\n"
743 "HT register dump:\n");
744 cpc925_printk(KERN_INFO, "Bridge Ctrl 0x%08x\n",
745 brgctrl);
746 cpc925_printk(KERN_INFO, "Link Config Ctrl 0x%08x\n",
747 linkctrl);
748 cpc925_printk(KERN_INFO, "Error Enum and Ctrl 0x%08x\n",
749 errctrl);
750 cpc925_printk(KERN_INFO, "Link Error 0x%08x\n",
751 linkerr);
752
753 /* Clear by write 1 */
754 if (brgctrl & BRGCTRL_DETSERR)
755 __raw_writel(BRGCTRL_DETSERR,
756 dev_info->vbase + REG_BRGCTRL_OFFSET);
757
758 if (linkctrl & HT_LINKCTRL_DETECTED)
759 __raw_writel(HT_LINKCTRL_DETECTED,
760 dev_info->vbase + REG_LINKCTRL_OFFSET);
761
762 /* Initiate Secondary Bus Reset to clear the chain failure */
763 if (errctrl & ERRCTRL_CHN_FAL)
764 __raw_writel(BRGCTRL_SECBUSRESET,
765 dev_info->vbase + REG_BRGCTRL_OFFSET);
766
767 if (errctrl & ERRCTRL_RSP_ERR)
768 __raw_writel(ERRCTRL_RSP_ERR,
769 dev_info->vbase + REG_ERRCTRL_OFFSET);
770
771 if (linkerr & HT_LINKERR_DETECTED)
772 __raw_writel(HT_LINKERR_DETECTED,
773 dev_info->vbase + REG_LINKERR_OFFSET);
774
775 edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
776 }
777
778 static struct cpc925_dev_info cpc925_devs[] = {
779 {
780 .ctl_name = CPC925_CPU_ERR_DEV,
781 .init = cpc925_cpu_init,
782 .exit = cpc925_cpu_exit,
783 .check = cpc925_cpu_check,
784 },
785 {
786 .ctl_name = CPC925_HT_LINK_DEV,
787 .init = cpc925_htlink_init,
788 .exit = cpc925_htlink_exit,
789 .check = cpc925_htlink_check,
790 },
791 { }
792 };
793
794 /*
795 * Add CPU Err detection and HyperTransport Link Err detection
796 * as common "edac_device", they have no corresponding device
797 * nodes in the Open Firmware DTB and we have to add platform
798 * devices for them. Also, they will share the MMIO with that
799 * of memory controller.
800 */
cpc925_add_edac_devices(void __iomem * vbase)801 static void cpc925_add_edac_devices(void __iomem *vbase)
802 {
803 struct cpc925_dev_info *dev_info;
804
805 if (!vbase) {
806 cpc925_printk(KERN_ERR, "MMIO not established yet\n");
807 return;
808 }
809
810 for (dev_info = &cpc925_devs[0]; dev_info->init; dev_info++) {
811 dev_info->vbase = vbase;
812 dev_info->pdev = platform_device_register_simple(
813 dev_info->ctl_name, 0, NULL, 0);
814 if (IS_ERR(dev_info->pdev)) {
815 cpc925_printk(KERN_ERR,
816 "Can't register platform device for %s\n",
817 dev_info->ctl_name);
818 continue;
819 }
820
821 /*
822 * Don't have to allocate private structure but
823 * make use of cpc925_devs[] instead.
824 */
825 dev_info->edac_idx = edac_device_alloc_index();
826 dev_info->edac_dev =
827 edac_device_alloc_ctl_info(0, dev_info->ctl_name,
828 1, NULL, 0, 0, NULL, 0, dev_info->edac_idx);
829 if (!dev_info->edac_dev) {
830 cpc925_printk(KERN_ERR, "No memory for edac device\n");
831 goto err1;
832 }
833
834 dev_info->edac_dev->pvt_info = dev_info;
835 dev_info->edac_dev->dev = &dev_info->pdev->dev;
836 dev_info->edac_dev->ctl_name = dev_info->ctl_name;
837 dev_info->edac_dev->mod_name = CPC925_EDAC_MOD_STR;
838 dev_info->edac_dev->dev_name = dev_name(&dev_info->pdev->dev);
839
840 if (edac_op_state == EDAC_OPSTATE_POLL)
841 dev_info->edac_dev->edac_check = dev_info->check;
842
843 if (dev_info->init)
844 dev_info->init(dev_info);
845
846 if (edac_device_add_device(dev_info->edac_dev) > 0) {
847 cpc925_printk(KERN_ERR,
848 "Unable to add edac device for %s\n",
849 dev_info->ctl_name);
850 goto err2;
851 }
852
853 edac_dbg(0, "Successfully added edac device for %s\n",
854 dev_info->ctl_name);
855
856 continue;
857
858 err2:
859 if (dev_info->exit)
860 dev_info->exit(dev_info);
861 edac_device_free_ctl_info(dev_info->edac_dev);
862 err1:
863 platform_device_unregister(dev_info->pdev);
864 }
865 }
866
867 /*
868 * Delete the common "edac_device" for CPU Err Detection
869 * and HyperTransport Link Err Detection
870 */
cpc925_del_edac_devices(void)871 static void cpc925_del_edac_devices(void)
872 {
873 struct cpc925_dev_info *dev_info;
874
875 for (dev_info = &cpc925_devs[0]; dev_info->init; dev_info++) {
876 if (dev_info->edac_dev) {
877 edac_device_del_device(dev_info->edac_dev->dev);
878 edac_device_free_ctl_info(dev_info->edac_dev);
879 platform_device_unregister(dev_info->pdev);
880 }
881
882 if (dev_info->exit)
883 dev_info->exit(dev_info);
884
885 edac_dbg(0, "Successfully deleted edac device for %s\n",
886 dev_info->ctl_name);
887 }
888 }
889
890 /* Convert current back-ground scrub rate into byte/sec bandwidth */
cpc925_get_sdram_scrub_rate(struct mem_ctl_info * mci)891 static int cpc925_get_sdram_scrub_rate(struct mem_ctl_info *mci)
892 {
893 struct cpc925_mc_pdata *pdata = mci->pvt_info;
894 int bw;
895 u32 mscr;
896 u8 si;
897
898 mscr = __raw_readl(pdata->vbase + REG_MSCR_OFFSET);
899 si = (mscr & MSCR_SI_MASK) >> MSCR_SI_SHIFT;
900
901 edac_dbg(0, "Mem Scrub Ctrl Register 0x%x\n", mscr);
902
903 if (((mscr & MSCR_SCRUB_MOD_MASK) != MSCR_BACKGR_SCRUB) ||
904 (si == 0)) {
905 cpc925_mc_printk(mci, KERN_INFO, "Scrub mode not enabled\n");
906 bw = 0;
907 } else
908 bw = CPC925_SCRUB_BLOCK_SIZE * 0xFA67 / si;
909
910 return bw;
911 }
912
913 /* Return 0 for single channel; 1 for dual channel */
cpc925_mc_get_channels(void __iomem * vbase)914 static int cpc925_mc_get_channels(void __iomem *vbase)
915 {
916 int dual = 0;
917 u32 mbcr;
918
919 mbcr = __raw_readl(vbase + REG_MBCR_OFFSET);
920
921 /*
922 * Dual channel only when 128-bit wide physical bus
923 * and 128-bit configuration.
924 */
925 if (((mbcr & MBCR_64BITCFG_MASK) == 0) &&
926 ((mbcr & MBCR_64BITBUS_MASK) == 0))
927 dual = 1;
928
929 edac_dbg(0, "%s channel\n", (dual > 0) ? "Dual" : "Single");
930
931 return dual;
932 }
933
cpc925_probe(struct platform_device * pdev)934 static int cpc925_probe(struct platform_device *pdev)
935 {
936 static int edac_mc_idx;
937 struct mem_ctl_info *mci;
938 struct edac_mc_layer layers[2];
939 void __iomem *vbase;
940 struct cpc925_mc_pdata *pdata;
941 struct resource *r;
942 int res = 0, nr_channels;
943
944 edac_dbg(0, "%s platform device found!\n", pdev->name);
945
946 if (!devres_open_group(&pdev->dev, cpc925_probe, GFP_KERNEL)) {
947 res = -ENOMEM;
948 goto out;
949 }
950
951 r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
952 if (!r) {
953 cpc925_printk(KERN_ERR, "Unable to get resource\n");
954 res = -ENOENT;
955 goto err1;
956 }
957
958 if (!devm_request_mem_region(&pdev->dev,
959 r->start,
960 resource_size(r),
961 pdev->name)) {
962 cpc925_printk(KERN_ERR, "Unable to request mem region\n");
963 res = -EBUSY;
964 goto err1;
965 }
966
967 vbase = devm_ioremap(&pdev->dev, r->start, resource_size(r));
968 if (!vbase) {
969 cpc925_printk(KERN_ERR, "Unable to ioremap device\n");
970 res = -ENOMEM;
971 goto err2;
972 }
973
974 nr_channels = cpc925_mc_get_channels(vbase) + 1;
975
976 layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
977 layers[0].size = CPC925_NR_CSROWS;
978 layers[0].is_virt_csrow = true;
979 layers[1].type = EDAC_MC_LAYER_CHANNEL;
980 layers[1].size = nr_channels;
981 layers[1].is_virt_csrow = false;
982 mci = edac_mc_alloc(edac_mc_idx, ARRAY_SIZE(layers), layers,
983 sizeof(struct cpc925_mc_pdata));
984 if (!mci) {
985 cpc925_printk(KERN_ERR, "No memory for mem_ctl_info\n");
986 res = -ENOMEM;
987 goto err2;
988 }
989
990 pdata = mci->pvt_info;
991 pdata->vbase = vbase;
992 pdata->edac_idx = edac_mc_idx++;
993 pdata->name = pdev->name;
994
995 mci->pdev = &pdev->dev;
996 platform_set_drvdata(pdev, mci);
997 mci->dev_name = dev_name(&pdev->dev);
998 mci->mtype_cap = MEM_FLAG_RDDR | MEM_FLAG_DDR;
999 mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED;
1000 mci->edac_cap = EDAC_FLAG_SECDED;
1001 mci->mod_name = CPC925_EDAC_MOD_STR;
1002 mci->ctl_name = pdev->name;
1003
1004 if (edac_op_state == EDAC_OPSTATE_POLL)
1005 mci->edac_check = cpc925_mc_check;
1006
1007 mci->ctl_page_to_phys = NULL;
1008 mci->scrub_mode = SCRUB_SW_SRC;
1009 mci->set_sdram_scrub_rate = NULL;
1010 mci->get_sdram_scrub_rate = cpc925_get_sdram_scrub_rate;
1011
1012 cpc925_init_csrows(mci);
1013
1014 /* Setup memory controller registers */
1015 cpc925_mc_init(mci);
1016
1017 if (edac_mc_add_mc(mci) > 0) {
1018 cpc925_mc_printk(mci, KERN_ERR, "Failed edac_mc_add_mc()\n");
1019 goto err3;
1020 }
1021
1022 cpc925_add_edac_devices(vbase);
1023
1024 /* get this far and it's successful */
1025 edac_dbg(0, "success\n");
1026
1027 res = 0;
1028 goto out;
1029
1030 err3:
1031 cpc925_mc_exit(mci);
1032 edac_mc_free(mci);
1033 err2:
1034 devm_release_mem_region(&pdev->dev, r->start, resource_size(r));
1035 err1:
1036 devres_release_group(&pdev->dev, cpc925_probe);
1037 out:
1038 return res;
1039 }
1040
cpc925_remove(struct platform_device * pdev)1041 static int cpc925_remove(struct platform_device *pdev)
1042 {
1043 struct mem_ctl_info *mci = platform_get_drvdata(pdev);
1044
1045 /*
1046 * Delete common edac devices before edac mc, because
1047 * the former share the MMIO of the latter.
1048 */
1049 cpc925_del_edac_devices();
1050 cpc925_mc_exit(mci);
1051
1052 edac_mc_del_mc(&pdev->dev);
1053 edac_mc_free(mci);
1054
1055 return 0;
1056 }
1057
1058 static struct platform_driver cpc925_edac_driver = {
1059 .probe = cpc925_probe,
1060 .remove = cpc925_remove,
1061 .driver = {
1062 .name = "cpc925_edac",
1063 }
1064 };
1065
cpc925_edac_init(void)1066 static int __init cpc925_edac_init(void)
1067 {
1068 int ret = 0;
1069
1070 printk(KERN_INFO "IBM CPC925 EDAC driver " CPC925_EDAC_REVISION "\n");
1071 printk(KERN_INFO "\t(c) 2008 Wind River Systems, Inc\n");
1072
1073 /* Only support POLL mode so far */
1074 edac_op_state = EDAC_OPSTATE_POLL;
1075
1076 ret = platform_driver_register(&cpc925_edac_driver);
1077 if (ret) {
1078 printk(KERN_WARNING "Failed to register %s\n",
1079 CPC925_EDAC_MOD_STR);
1080 }
1081
1082 return ret;
1083 }
1084
cpc925_edac_exit(void)1085 static void __exit cpc925_edac_exit(void)
1086 {
1087 platform_driver_unregister(&cpc925_edac_driver);
1088 }
1089
1090 module_init(cpc925_edac_init);
1091 module_exit(cpc925_edac_exit);
1092
1093 MODULE_LICENSE("GPL");
1094 MODULE_AUTHOR("Cao Qingtao <qingtao.cao@windriver.com>");
1095 MODULE_DESCRIPTION("IBM CPC925 Bridge and MC EDAC kernel module");
1096