/drivers/edac/cpc925_edac.c
C | 1019 lines | 678 code | 156 blank | 185 comment | 65 complexity | fc795fd3ce5d6729e98d402da033f4fe MD5 | raw file
Possible License(s): GPL-2.0, LGPL-2.0, AGPL-1.0
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_core.h"
31#include "edac_module.h"
32
33#define CPC925_EDAC_REVISION " Ver: 1.0.0"
34#define CPC925_EDAC_MOD_STR "cpc925_edac"
35
36#define cpc925_printk(level, fmt, arg...) \
37 edac_printk(level, "CPC925", fmt, ##arg)
38
39#define cpc925_mc_printk(mci, level, fmt, arg...) \
40 edac_mc_chipset_printk(mci, level, "CPC925", fmt, ##arg)
41
42/*
43 * CPC925 registers are of 32 bits with bit0 defined at the
44 * most significant bit and bit31 at that of least significant.
45 */
46#define CPC925_BITS_PER_REG 32
47#define CPC925_BIT(nr) (1UL << (CPC925_BITS_PER_REG - 1 - nr))
48
49/*
50 * EDAC device names for the error detections of
51 * CPU Interface and Hypertransport Link.
52 */
53#define CPC925_CPU_ERR_DEV "cpu"
54#define CPC925_HT_LINK_DEV "htlink"
55
56/* Suppose DDR Refresh cycle is 15.6 microsecond */
57#define CPC925_REF_FREQ 0xFA69
58#define CPC925_SCRUB_BLOCK_SIZE 64 /* bytes */
59#define CPC925_NR_CSROWS 8
60
61/*
62 * All registers and bits definitions are taken from
63 * "CPC925 Bridge and Memory Controller User Manual, SA14-2761-02".
64 */
65
66/*
67 * CPU and Memory Controller Registers
68 */
69/************************************************************
70 * Processor Interface Exception Mask Register (APIMASK)
71 ************************************************************/
72#define REG_APIMASK_OFFSET 0x30070
73enum apimask_bits {
74 APIMASK_DART = CPC925_BIT(0), /* DART Exception */
75 APIMASK_ADI0 = CPC925_BIT(1), /* Handshake Error on PI0_ADI */
76 APIMASK_ADI1 = CPC925_BIT(2), /* Handshake Error on PI1_ADI */
77 APIMASK_STAT = CPC925_BIT(3), /* Status Exception */
78 APIMASK_DERR = CPC925_BIT(4), /* Data Error Exception */
79 APIMASK_ADRS0 = CPC925_BIT(5), /* Addressing Exception on PI0 */
80 APIMASK_ADRS1 = CPC925_BIT(6), /* Addressing Exception on PI1 */
81 /* BIT(7) Reserved */
82 APIMASK_ECC_UE_H = CPC925_BIT(8), /* UECC upper */
83 APIMASK_ECC_CE_H = CPC925_BIT(9), /* CECC upper */
84 APIMASK_ECC_UE_L = CPC925_BIT(10), /* UECC lower */
85 APIMASK_ECC_CE_L = CPC925_BIT(11), /* CECC lower */
86
87 CPU_MASK_ENABLE = (APIMASK_DART | APIMASK_ADI0 | APIMASK_ADI1 |
88 APIMASK_STAT | APIMASK_DERR | APIMASK_ADRS0 |
89 APIMASK_ADRS1),
90 ECC_MASK_ENABLE = (APIMASK_ECC_UE_H | APIMASK_ECC_CE_H |
91 APIMASK_ECC_UE_L | APIMASK_ECC_CE_L),
92};
93
94/************************************************************
95 * Processor Interface Exception Register (APIEXCP)
96 ************************************************************/
97#define REG_APIEXCP_OFFSET 0x30060
98enum 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
176enum 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
214enum 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
225enum 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
246enum 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
257enum 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
270enum 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 */
276struct 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 */
284struct 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 */
296static 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 debugf1("%s: start 0x%lx, size 0x%lx\n", __func__,
319 start, size);
320 pdata->total_mem += size;
321 } while (reg < reg_end);
322
323 of_node_put(np);
324 debugf0("%s: total_mem 0x%lx\n", __func__, pdata->total_mem);
325}
326
327static void cpc925_init_csrows(struct mem_ctl_info *mci)
328{
329 struct cpc925_mc_pdata *pdata = mci->pvt_info;
330 struct csrow_info *csrow;
331 int index;
332 u32 mbmr, mbbar, bba;
333 unsigned long row_size, last_nr_pages = 0;
334
335 get_total_mem(pdata);
336
337 for (index = 0; index < mci->nr_csrows; index++) {
338 mbmr = __raw_readl(pdata->vbase + REG_MBMR_OFFSET +
339 0x20 * index);
340 mbbar = __raw_readl(pdata->vbase + REG_MBBAR_OFFSET +
341 0x20 + index);
342 bba = (((mbmr & MBMR_BBA_MASK) >> MBMR_BBA_SHIFT) << 8) |
343 ((mbbar & MBBAR_BBA_MASK) >> MBBAR_BBA_SHIFT);
344
345 if (bba == 0)
346 continue; /* not populated */
347
348 csrow = &mci->csrows[index];
349
350 row_size = bba * (1UL << 28); /* 256M */
351 csrow->first_page = last_nr_pages;
352 csrow->nr_pages = row_size >> PAGE_SHIFT;
353 csrow->last_page = csrow->first_page + csrow->nr_pages - 1;
354 last_nr_pages = csrow->last_page + 1;
355
356 csrow->mtype = MEM_RDDR;
357 csrow->edac_mode = EDAC_SECDED;
358
359 switch (csrow->nr_channels) {
360 case 1: /* Single channel */
361 csrow->grain = 32; /* four-beat burst of 32 bytes */
362 break;
363 case 2: /* Dual channel */
364 default:
365 csrow->grain = 64; /* four-beat burst of 64 bytes */
366 break;
367 }
368
369 switch ((mbmr & MBMR_MODE_MASK) >> MBMR_MODE_SHIFT) {
370 case 6: /* 0110, no way to differentiate X8 VS X16 */
371 case 5: /* 0101 */
372 case 8: /* 1000 */
373 csrow->dtype = DEV_X16;
374 break;
375 case 7: /* 0111 */
376 case 9: /* 1001 */
377 csrow->dtype = DEV_X8;
378 break;
379 default:
380 csrow->dtype = DEV_UNKNOWN;
381 break;
382 }
383 }
384}
385
386/* Enable memory controller ECC detection */
387static void cpc925_mc_init(struct mem_ctl_info *mci)
388{
389 struct cpc925_mc_pdata *pdata = mci->pvt_info;
390 u32 apimask;
391 u32 mccr;
392
393 /* Enable various ECC error exceptions */
394 apimask = __raw_readl(pdata->vbase + REG_APIMASK_OFFSET);
395 if ((apimask & ECC_MASK_ENABLE) == 0) {
396 apimask |= ECC_MASK_ENABLE;
397 __raw_writel(apimask, pdata->vbase + REG_APIMASK_OFFSET);
398 }
399
400 /* Enable ECC detection */
401 mccr = __raw_readl(pdata->vbase + REG_MCCR_OFFSET);
402 if ((mccr & MCCR_ECC_EN) == 0) {
403 mccr |= MCCR_ECC_EN;
404 __raw_writel(mccr, pdata->vbase + REG_MCCR_OFFSET);
405 }
406}
407
408/* Disable memory controller ECC detection */
409static void cpc925_mc_exit(struct mem_ctl_info *mci)
410{
411 /*
412 * WARNING:
413 * We are supposed to clear the ECC error detection bits,
414 * and it will be no problem to do so. However, once they
415 * are cleared here if we want to re-install CPC925 EDAC
416 * module later, setting them up in cpc925_mc_init() will
417 * trigger machine check exception.
418 * Also, it's ok to leave ECC error detection bits enabled,
419 * since they are reset to 1 by default or by boot loader.
420 */
421
422 return;
423}
424
425/*
426 * Revert DDR column/row/bank addresses into page frame number and
427 * offset in page.
428 *
429 * Suppose memory mode is 0x0111(128-bit mode, identical DIMM pairs),
430 * physical address(PA) bits to column address(CA) bits mappings are:
431 * CA 0 1 2 3 4 5 6 7 8 9 10
432 * PA 59 58 57 56 55 54 53 52 51 50 49
433 *
434 * physical address(PA) bits to bank address(BA) bits mappings are:
435 * BA 0 1
436 * PA 43 44
437 *
438 * physical address(PA) bits to row address(RA) bits mappings are:
439 * RA 0 1 2 3 4 5 6 7 8 9 10 11 12
440 * PA 36 35 34 48 47 46 45 40 41 42 39 38 37
441 */
442static void cpc925_mc_get_pfn(struct mem_ctl_info *mci, u32 mear,
443 unsigned long *pfn, unsigned long *offset, int *csrow)
444{
445 u32 bcnt, rank, col, bank, row;
446 u32 c;
447 unsigned long pa;
448 int i;
449
450 bcnt = (mear & MEAR_BCNT_MASK) >> MEAR_BCNT_SHIFT;
451 rank = (mear & MEAR_RANK_MASK) >> MEAR_RANK_SHIFT;
452 col = (mear & MEAR_COL_MASK) >> MEAR_COL_SHIFT;
453 bank = (mear & MEAR_BANK_MASK) >> MEAR_BANK_SHIFT;
454 row = mear & MEAR_ROW_MASK;
455
456 *csrow = rank;
457
458#ifdef CONFIG_EDAC_DEBUG
459 if (mci->csrows[rank].first_page == 0) {
460 cpc925_mc_printk(mci, KERN_ERR, "ECC occurs in a "
461 "non-populated csrow, broken hardware?\n");
462 return;
463 }
464#endif
465
466 /* Revert csrow number */
467 pa = mci->csrows[rank].first_page << PAGE_SHIFT;
468
469 /* Revert column address */
470 col += bcnt;
471 for (i = 0; i < 11; i++) {
472 c = col & 0x1;
473 col >>= 1;
474 pa |= c << (14 - i);
475 }
476
477 /* Revert bank address */
478 pa |= bank << 19;
479
480 /* Revert row address, in 4 steps */
481 for (i = 0; i < 3; i++) {
482 c = row & 0x1;
483 row >>= 1;
484 pa |= c << (26 - i);
485 }
486
487 for (i = 0; i < 3; i++) {
488 c = row & 0x1;
489 row >>= 1;
490 pa |= c << (21 + i);
491 }
492
493 for (i = 0; i < 4; i++) {
494 c = row & 0x1;
495 row >>= 1;
496 pa |= c << (18 - i);
497 }
498
499 for (i = 0; i < 3; i++) {
500 c = row & 0x1;
501 row >>= 1;
502 pa |= c << (29 - i);
503 }
504
505 *offset = pa & (PAGE_SIZE - 1);
506 *pfn = pa >> PAGE_SHIFT;
507
508 debugf0("%s: ECC physical address 0x%lx\n", __func__, pa);
509}
510
511static int cpc925_mc_find_channel(struct mem_ctl_info *mci, u16 syndrome)
512{
513 if ((syndrome & MESR_ECC_SYN_H_MASK) == 0)
514 return 0;
515
516 if ((syndrome & MESR_ECC_SYN_L_MASK) == 0)
517 return 1;
518
519 cpc925_mc_printk(mci, KERN_INFO, "Unexpected syndrome value: 0x%x\n",
520 syndrome);
521 return 1;
522}
523
524/* Check memory controller registers for ECC errors */
525static void cpc925_mc_check(struct mem_ctl_info *mci)
526{
527 struct cpc925_mc_pdata *pdata = mci->pvt_info;
528 u32 apiexcp;
529 u32 mear;
530 u32 mesr;
531 u16 syndrome;
532 unsigned long pfn = 0, offset = 0;
533 int csrow = 0, channel = 0;
534
535 /* APIEXCP is cleared when read */
536 apiexcp = __raw_readl(pdata->vbase + REG_APIEXCP_OFFSET);
537 if ((apiexcp & ECC_EXCP_DETECTED) == 0)
538 return;
539
540 mesr = __raw_readl(pdata->vbase + REG_MESR_OFFSET);
541 syndrome = mesr | (MESR_ECC_SYN_H_MASK | MESR_ECC_SYN_L_MASK);
542
543 mear = __raw_readl(pdata->vbase + REG_MEAR_OFFSET);
544
545 /* Revert column/row addresses into page frame number, etc */
546 cpc925_mc_get_pfn(mci, mear, &pfn, &offset, &csrow);
547
548 if (apiexcp & CECC_EXCP_DETECTED) {
549 cpc925_mc_printk(mci, KERN_INFO, "DRAM CECC Fault\n");
550 channel = cpc925_mc_find_channel(mci, syndrome);
551 edac_mc_handle_ce(mci, pfn, offset, syndrome,
552 csrow, channel, mci->ctl_name);
553 }
554
555 if (apiexcp & UECC_EXCP_DETECTED) {
556 cpc925_mc_printk(mci, KERN_INFO, "DRAM UECC Fault\n");
557 edac_mc_handle_ue(mci, pfn, offset, csrow, mci->ctl_name);
558 }
559
560 cpc925_mc_printk(mci, KERN_INFO, "Dump registers:\n");
561 cpc925_mc_printk(mci, KERN_INFO, "APIMASK 0x%08x\n",
562 __raw_readl(pdata->vbase + REG_APIMASK_OFFSET));
563 cpc925_mc_printk(mci, KERN_INFO, "APIEXCP 0x%08x\n",
564 apiexcp);
565 cpc925_mc_printk(mci, KERN_INFO, "Mem Scrub Ctrl 0x%08x\n",
566 __raw_readl(pdata->vbase + REG_MSCR_OFFSET));
567 cpc925_mc_printk(mci, KERN_INFO, "Mem Scrub Rge Start 0x%08x\n",
568 __raw_readl(pdata->vbase + REG_MSRSR_OFFSET));
569 cpc925_mc_printk(mci, KERN_INFO, "Mem Scrub Rge End 0x%08x\n",
570 __raw_readl(pdata->vbase + REG_MSRER_OFFSET));
571 cpc925_mc_printk(mci, KERN_INFO, "Mem Scrub Pattern 0x%08x\n",
572 __raw_readl(pdata->vbase + REG_MSPR_OFFSET));
573 cpc925_mc_printk(mci, KERN_INFO, "Mem Chk Ctrl 0x%08x\n",
574 __raw_readl(pdata->vbase + REG_MCCR_OFFSET));
575 cpc925_mc_printk(mci, KERN_INFO, "Mem Chk Rge End 0x%08x\n",
576 __raw_readl(pdata->vbase + REG_MCRER_OFFSET));
577 cpc925_mc_printk(mci, KERN_INFO, "Mem Err Address 0x%08x\n",
578 mesr);
579 cpc925_mc_printk(mci, KERN_INFO, "Mem Err Syndrome 0x%08x\n",
580 syndrome);
581}
582
583/******************** CPU err device********************************/
584/* Enable CPU Errors detection */
585static void cpc925_cpu_init(struct cpc925_dev_info *dev_info)
586{
587 u32 apimask;
588
589 apimask = __raw_readl(dev_info->vbase + REG_APIMASK_OFFSET);
590 if ((apimask & CPU_MASK_ENABLE) == 0) {
591 apimask |= CPU_MASK_ENABLE;
592 __raw_writel(apimask, dev_info->vbase + REG_APIMASK_OFFSET);
593 }
594}
595
596/* Disable CPU Errors detection */
597static void cpc925_cpu_exit(struct cpc925_dev_info *dev_info)
598{
599 /*
600 * WARNING:
601 * We are supposed to clear the CPU error detection bits,
602 * and it will be no problem to do so. However, once they
603 * are cleared here if we want to re-install CPC925 EDAC
604 * module later, setting them up in cpc925_cpu_init() will
605 * trigger machine check exception.
606 * Also, it's ok to leave CPU error detection bits enabled,
607 * since they are reset to 1 by default.
608 */
609
610 return;
611}
612
613/* Check for CPU Errors */
614static void cpc925_cpu_check(struct edac_device_ctl_info *edac_dev)
615{
616 struct cpc925_dev_info *dev_info = edac_dev->pvt_info;
617 u32 apiexcp;
618 u32 apimask;
619
620 /* APIEXCP is cleared when read */
621 apiexcp = __raw_readl(dev_info->vbase + REG_APIEXCP_OFFSET);
622 if ((apiexcp & CPU_EXCP_DETECTED) == 0)
623 return;
624
625 apimask = __raw_readl(dev_info->vbase + REG_APIMASK_OFFSET);
626 cpc925_printk(KERN_INFO, "Processor Interface Fault\n"
627 "Processor Interface register dump:\n");
628 cpc925_printk(KERN_INFO, "APIMASK 0x%08x\n", apimask);
629 cpc925_printk(KERN_INFO, "APIEXCP 0x%08x\n", apiexcp);
630
631 edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
632}
633
634/******************** HT Link err device****************************/
635/* Enable HyperTransport Link Error detection */
636static void cpc925_htlink_init(struct cpc925_dev_info *dev_info)
637{
638 u32 ht_errctrl;
639
640 ht_errctrl = __raw_readl(dev_info->vbase + REG_ERRCTRL_OFFSET);
641 if ((ht_errctrl & HT_ERRCTRL_ENABLE) == 0) {
642 ht_errctrl |= HT_ERRCTRL_ENABLE;
643 __raw_writel(ht_errctrl, dev_info->vbase + REG_ERRCTRL_OFFSET);
644 }
645}
646
647/* Disable HyperTransport Link Error detection */
648static void cpc925_htlink_exit(struct cpc925_dev_info *dev_info)
649{
650 u32 ht_errctrl;
651
652 ht_errctrl = __raw_readl(dev_info->vbase + REG_ERRCTRL_OFFSET);
653 ht_errctrl &= ~HT_ERRCTRL_ENABLE;
654 __raw_writel(ht_errctrl, dev_info->vbase + REG_ERRCTRL_OFFSET);
655}
656
657/* Check for HyperTransport Link errors */
658static void cpc925_htlink_check(struct edac_device_ctl_info *edac_dev)
659{
660 struct cpc925_dev_info *dev_info = edac_dev->pvt_info;
661 u32 brgctrl = __raw_readl(dev_info->vbase + REG_BRGCTRL_OFFSET);
662 u32 linkctrl = __raw_readl(dev_info->vbase + REG_LINKCTRL_OFFSET);
663 u32 errctrl = __raw_readl(dev_info->vbase + REG_ERRCTRL_OFFSET);
664 u32 linkerr = __raw_readl(dev_info->vbase + REG_LINKERR_OFFSET);
665
666 if (!((brgctrl & BRGCTRL_DETSERR) ||
667 (linkctrl & HT_LINKCTRL_DETECTED) ||
668 (errctrl & HT_ERRCTRL_DETECTED) ||
669 (linkerr & HT_LINKERR_DETECTED)))
670 return;
671
672 cpc925_printk(KERN_INFO, "HT Link Fault\n"
673 "HT register dump:\n");
674 cpc925_printk(KERN_INFO, "Bridge Ctrl 0x%08x\n",
675 brgctrl);
676 cpc925_printk(KERN_INFO, "Link Config Ctrl 0x%08x\n",
677 linkctrl);
678 cpc925_printk(KERN_INFO, "Error Enum and Ctrl 0x%08x\n",
679 errctrl);
680 cpc925_printk(KERN_INFO, "Link Error 0x%08x\n",
681 linkerr);
682
683 /* Clear by write 1 */
684 if (brgctrl & BRGCTRL_DETSERR)
685 __raw_writel(BRGCTRL_DETSERR,
686 dev_info->vbase + REG_BRGCTRL_OFFSET);
687
688 if (linkctrl & HT_LINKCTRL_DETECTED)
689 __raw_writel(HT_LINKCTRL_DETECTED,
690 dev_info->vbase + REG_LINKCTRL_OFFSET);
691
692 /* Initiate Secondary Bus Reset to clear the chain failure */
693 if (errctrl & ERRCTRL_CHN_FAL)
694 __raw_writel(BRGCTRL_SECBUSRESET,
695 dev_info->vbase + REG_BRGCTRL_OFFSET);
696
697 if (errctrl & ERRCTRL_RSP_ERR)
698 __raw_writel(ERRCTRL_RSP_ERR,
699 dev_info->vbase + REG_ERRCTRL_OFFSET);
700
701 if (linkerr & HT_LINKERR_DETECTED)
702 __raw_writel(HT_LINKERR_DETECTED,
703 dev_info->vbase + REG_LINKERR_OFFSET);
704
705 edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
706}
707
708static struct cpc925_dev_info cpc925_devs[] = {
709 {
710 .ctl_name = CPC925_CPU_ERR_DEV,
711 .init = cpc925_cpu_init,
712 .exit = cpc925_cpu_exit,
713 .check = cpc925_cpu_check,
714 },
715 {
716 .ctl_name = CPC925_HT_LINK_DEV,
717 .init = cpc925_htlink_init,
718 .exit = cpc925_htlink_exit,
719 .check = cpc925_htlink_check,
720 },
721 {0}, /* Terminated by NULL */
722};
723
724/*
725 * Add CPU Err detection and HyperTransport Link Err detection
726 * as common "edac_device", they have no corresponding device
727 * nodes in the Open Firmware DTB and we have to add platform
728 * devices for them. Also, they will share the MMIO with that
729 * of memory controller.
730 */
731static void cpc925_add_edac_devices(void __iomem *vbase)
732{
733 struct cpc925_dev_info *dev_info;
734
735 if (!vbase) {
736 cpc925_printk(KERN_ERR, "MMIO not established yet\n");
737 return;
738 }
739
740 for (dev_info = &cpc925_devs[0]; dev_info->init; dev_info++) {
741 dev_info->vbase = vbase;
742 dev_info->pdev = platform_device_register_simple(
743 dev_info->ctl_name, 0, NULL, 0);
744 if (IS_ERR(dev_info->pdev)) {
745 cpc925_printk(KERN_ERR,
746 "Can't register platform device for %s\n",
747 dev_info->ctl_name);
748 continue;
749 }
750
751 /*
752 * Don't have to allocate private structure but
753 * make use of cpc925_devs[] instead.
754 */
755 dev_info->edac_idx = edac_device_alloc_index();
756 dev_info->edac_dev =
757 edac_device_alloc_ctl_info(0, dev_info->ctl_name,
758 1, NULL, 0, 0, NULL, 0, dev_info->edac_idx);
759 if (!dev_info->edac_dev) {
760 cpc925_printk(KERN_ERR, "No memory for edac device\n");
761 goto err1;
762 }
763
764 dev_info->edac_dev->pvt_info = dev_info;
765 dev_info->edac_dev->dev = &dev_info->pdev->dev;
766 dev_info->edac_dev->ctl_name = dev_info->ctl_name;
767 dev_info->edac_dev->mod_name = CPC925_EDAC_MOD_STR;
768 dev_info->edac_dev->dev_name = dev_name(&dev_info->pdev->dev);
769
770 if (edac_op_state == EDAC_OPSTATE_POLL)
771 dev_info->edac_dev->edac_check = dev_info->check;
772
773 if (dev_info->init)
774 dev_info->init(dev_info);
775
776 if (edac_device_add_device(dev_info->edac_dev) > 0) {
777 cpc925_printk(KERN_ERR,
778 "Unable to add edac device for %s\n",
779 dev_info->ctl_name);
780 goto err2;
781 }
782
783 debugf0("%s: Successfully added edac device for %s\n",
784 __func__, dev_info->ctl_name);
785
786 continue;
787
788err2:
789 if (dev_info->exit)
790 dev_info->exit(dev_info);
791 edac_device_free_ctl_info(dev_info->edac_dev);
792err1:
793 platform_device_unregister(dev_info->pdev);
794 }
795}
796
797/*
798 * Delete the common "edac_device" for CPU Err Detection
799 * and HyperTransport Link Err Detection
800 */
801static void cpc925_del_edac_devices(void)
802{
803 struct cpc925_dev_info *dev_info;
804
805 for (dev_info = &cpc925_devs[0]; dev_info->init; dev_info++) {
806 if (dev_info->edac_dev) {
807 edac_device_del_device(dev_info->edac_dev->dev);
808 edac_device_free_ctl_info(dev_info->edac_dev);
809 platform_device_unregister(dev_info->pdev);
810 }
811
812 if (dev_info->exit)
813 dev_info->exit(dev_info);
814
815 debugf0("%s: Successfully deleted edac device for %s\n",
816 __func__, dev_info->ctl_name);
817 }
818}
819
820/* Convert current back-ground scrub rate into byte/sec bandwidth */
821static int cpc925_get_sdram_scrub_rate(struct mem_ctl_info *mci)
822{
823 struct cpc925_mc_pdata *pdata = mci->pvt_info;
824 int bw;
825 u32 mscr;
826 u8 si;
827
828 mscr = __raw_readl(pdata->vbase + REG_MSCR_OFFSET);
829 si = (mscr & MSCR_SI_MASK) >> MSCR_SI_SHIFT;
830
831 debugf0("%s, Mem Scrub Ctrl Register 0x%x\n", __func__, mscr);
832
833 if (((mscr & MSCR_SCRUB_MOD_MASK) != MSCR_BACKGR_SCRUB) ||
834 (si == 0)) {
835 cpc925_mc_printk(mci, KERN_INFO, "Scrub mode not enabled\n");
836 bw = 0;
837 } else
838 bw = CPC925_SCRUB_BLOCK_SIZE * 0xFA67 / si;
839
840 return bw;
841}
842
843/* Return 0 for single channel; 1 for dual channel */
844static int cpc925_mc_get_channels(void __iomem *vbase)
845{
846 int dual = 0;
847 u32 mbcr;
848
849 mbcr = __raw_readl(vbase + REG_MBCR_OFFSET);
850
851 /*
852 * Dual channel only when 128-bit wide physical bus
853 * and 128-bit configuration.
854 */
855 if (((mbcr & MBCR_64BITCFG_MASK) == 0) &&
856 ((mbcr & MBCR_64BITBUS_MASK) == 0))
857 dual = 1;
858
859 debugf0("%s: %s channel\n", __func__,
860 (dual > 0) ? "Dual" : "Single");
861
862 return dual;
863}
864
865static int __devinit cpc925_probe(struct platform_device *pdev)
866{
867 static int edac_mc_idx;
868 struct mem_ctl_info *mci;
869 void __iomem *vbase;
870 struct cpc925_mc_pdata *pdata;
871 struct resource *r;
872 int res = 0, nr_channels;
873
874 debugf0("%s: %s platform device found!\n", __func__, pdev->name);
875
876 if (!devres_open_group(&pdev->dev, cpc925_probe, GFP_KERNEL)) {
877 res = -ENOMEM;
878 goto out;
879 }
880
881 r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
882 if (!r) {
883 cpc925_printk(KERN_ERR, "Unable to get resource\n");
884 res = -ENOENT;
885 goto err1;
886 }
887
888 if (!devm_request_mem_region(&pdev->dev,
889 r->start,
890 resource_size(r),
891 pdev->name)) {
892 cpc925_printk(KERN_ERR, "Unable to request mem region\n");
893 res = -EBUSY;
894 goto err1;
895 }
896
897 vbase = devm_ioremap(&pdev->dev, r->start, resource_size(r));
898 if (!vbase) {
899 cpc925_printk(KERN_ERR, "Unable to ioremap device\n");
900 res = -ENOMEM;
901 goto err2;
902 }
903
904 nr_channels = cpc925_mc_get_channels(vbase);
905 mci = edac_mc_alloc(sizeof(struct cpc925_mc_pdata),
906 CPC925_NR_CSROWS, nr_channels + 1, edac_mc_idx);
907 if (!mci) {
908 cpc925_printk(KERN_ERR, "No memory for mem_ctl_info\n");
909 res = -ENOMEM;
910 goto err2;
911 }
912
913 pdata = mci->pvt_info;
914 pdata->vbase = vbase;
915 pdata->edac_idx = edac_mc_idx++;
916 pdata->name = pdev->name;
917
918 mci->dev = &pdev->dev;
919 platform_set_drvdata(pdev, mci);
920 mci->dev_name = dev_name(&pdev->dev);
921 mci->mtype_cap = MEM_FLAG_RDDR | MEM_FLAG_DDR;
922 mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED;
923 mci->edac_cap = EDAC_FLAG_SECDED;
924 mci->mod_name = CPC925_EDAC_MOD_STR;
925 mci->mod_ver = CPC925_EDAC_REVISION;
926 mci->ctl_name = pdev->name;
927
928 if (edac_op_state == EDAC_OPSTATE_POLL)
929 mci->edac_check = cpc925_mc_check;
930
931 mci->ctl_page_to_phys = NULL;
932 mci->scrub_mode = SCRUB_SW_SRC;
933 mci->set_sdram_scrub_rate = NULL;
934 mci->get_sdram_scrub_rate = cpc925_get_sdram_scrub_rate;
935
936 cpc925_init_csrows(mci);
937
938 /* Setup memory controller registers */
939 cpc925_mc_init(mci);
940
941 if (edac_mc_add_mc(mci) > 0) {
942 cpc925_mc_printk(mci, KERN_ERR, "Failed edac_mc_add_mc()\n");
943 goto err3;
944 }
945
946 cpc925_add_edac_devices(vbase);
947
948 /* get this far and it's successful */
949 debugf0("%s: success\n", __func__);
950
951 res = 0;
952 goto out;
953
954err3:
955 cpc925_mc_exit(mci);
956 edac_mc_free(mci);
957err2:
958 devm_release_mem_region(&pdev->dev, r->start, resource_size(r));
959err1:
960 devres_release_group(&pdev->dev, cpc925_probe);
961out:
962 return res;
963}
964
965static int cpc925_remove(struct platform_device *pdev)
966{
967 struct mem_ctl_info *mci = platform_get_drvdata(pdev);
968
969 /*
970 * Delete common edac devices before edac mc, because
971 * the former share the MMIO of the latter.
972 */
973 cpc925_del_edac_devices();
974 cpc925_mc_exit(mci);
975
976 edac_mc_del_mc(&pdev->dev);
977 edac_mc_free(mci);
978
979 return 0;
980}
981
982static struct platform_driver cpc925_edac_driver = {
983 .probe = cpc925_probe,
984 .remove = cpc925_remove,
985 .driver = {
986 .name = "cpc925_edac",
987 }
988};
989
990static int __init cpc925_edac_init(void)
991{
992 int ret = 0;
993
994 printk(KERN_INFO "IBM CPC925 EDAC driver " CPC925_EDAC_REVISION "\n");
995 printk(KERN_INFO "\t(c) 2008 Wind River Systems, Inc\n");
996
997 /* Only support POLL mode so far */
998 edac_op_state = EDAC_OPSTATE_POLL;
999
1000 ret = platform_driver_register(&cpc925_edac_driver);
1001 if (ret) {
1002 printk(KERN_WARNING "Failed to register %s\n",
1003 CPC925_EDAC_MOD_STR);
1004 }
1005
1006 return ret;
1007}
1008
1009static void __exit cpc925_edac_exit(void)
1010{
1011 platform_driver_unregister(&cpc925_edac_driver);
1012}
1013
1014module_init(cpc925_edac_init);
1015module_exit(cpc925_edac_exit);
1016
1017MODULE_LICENSE("GPL");
1018MODULE_AUTHOR("Cao Qingtao <qingtao.cao@windriver.com>");
1019MODULE_DESCRIPTION("IBM CPC925 Bridge and MC EDAC kernel module");