Book a Demo!
CoCalc Logo Icon
StoreFeaturesDocsShareSupportNewsAboutPoliciesSign UpSign In
awilliam
GitHub Repository: awilliam/linux-vfio
Path: blob/master/drivers/edac/i5000_edac.c
15111 views
1
/*
2
* Intel 5000(P/V/X) class Memory Controllers kernel module
3
*
4
* This file may be distributed under the terms of the
5
* GNU General Public License.
6
*
7
* Written by Douglas Thompson Linux Networx (http://lnxi.com)
8
* [email protected]
9
*
10
* This module is based on the following document:
11
*
12
* Intel 5000X Chipset Memory Controller Hub (MCH) - Datasheet
13
* http://developer.intel.com/design/chipsets/datashts/313070.htm
14
*
15
*/
16
17
#include <linux/module.h>
18
#include <linux/init.h>
19
#include <linux/pci.h>
20
#include <linux/pci_ids.h>
21
#include <linux/slab.h>
22
#include <linux/edac.h>
23
#include <asm/mmzone.h>
24
25
#include "edac_core.h"
26
27
/*
28
* Alter this version for the I5000 module when modifications are made
29
*/
30
#define I5000_REVISION " Ver: 2.0.12"
31
#define EDAC_MOD_STR "i5000_edac"
32
33
#define i5000_printk(level, fmt, arg...) \
34
edac_printk(level, "i5000", fmt, ##arg)
35
36
#define i5000_mc_printk(mci, level, fmt, arg...) \
37
edac_mc_chipset_printk(mci, level, "i5000", fmt, ##arg)
38
39
#ifndef PCI_DEVICE_ID_INTEL_FBD_0
40
#define PCI_DEVICE_ID_INTEL_FBD_0 0x25F5
41
#endif
42
#ifndef PCI_DEVICE_ID_INTEL_FBD_1
43
#define PCI_DEVICE_ID_INTEL_FBD_1 0x25F6
44
#endif
45
46
/* Device 16,
47
* Function 0: System Address
48
* Function 1: Memory Branch Map, Control, Errors Register
49
* Function 2: FSB Error Registers
50
*
51
* All 3 functions of Device 16 (0,1,2) share the SAME DID
52
*/
53
#define PCI_DEVICE_ID_INTEL_I5000_DEV16 0x25F0
54
55
/* OFFSETS for Function 0 */
56
57
/* OFFSETS for Function 1 */
58
#define AMBASE 0x48
59
#define MAXCH 0x56
60
#define MAXDIMMPERCH 0x57
61
#define TOLM 0x6C
62
#define REDMEMB 0x7C
63
#define RED_ECC_LOCATOR(x) ((x) & 0x3FFFF)
64
#define REC_ECC_LOCATOR_EVEN(x) ((x) & 0x001FF)
65
#define REC_ECC_LOCATOR_ODD(x) ((x) & 0x3FE00)
66
#define MIR0 0x80
67
#define MIR1 0x84
68
#define MIR2 0x88
69
#define AMIR0 0x8C
70
#define AMIR1 0x90
71
#define AMIR2 0x94
72
73
#define FERR_FAT_FBD 0x98
74
#define NERR_FAT_FBD 0x9C
75
#define EXTRACT_FBDCHAN_INDX(x) (((x)>>28) & 0x3)
76
#define FERR_FAT_FBDCHAN 0x30000000
77
#define FERR_FAT_M3ERR 0x00000004
78
#define FERR_FAT_M2ERR 0x00000002
79
#define FERR_FAT_M1ERR 0x00000001
80
#define FERR_FAT_MASK (FERR_FAT_M1ERR | \
81
FERR_FAT_M2ERR | \
82
FERR_FAT_M3ERR)
83
84
#define FERR_NF_FBD 0xA0
85
86
/* Thermal and SPD or BFD errors */
87
#define FERR_NF_M28ERR 0x01000000
88
#define FERR_NF_M27ERR 0x00800000
89
#define FERR_NF_M26ERR 0x00400000
90
#define FERR_NF_M25ERR 0x00200000
91
#define FERR_NF_M24ERR 0x00100000
92
#define FERR_NF_M23ERR 0x00080000
93
#define FERR_NF_M22ERR 0x00040000
94
#define FERR_NF_M21ERR 0x00020000
95
96
/* Correctable errors */
97
#define FERR_NF_M20ERR 0x00010000
98
#define FERR_NF_M19ERR 0x00008000
99
#define FERR_NF_M18ERR 0x00004000
100
#define FERR_NF_M17ERR 0x00002000
101
102
/* Non-Retry or redundant Retry errors */
103
#define FERR_NF_M16ERR 0x00001000
104
#define FERR_NF_M15ERR 0x00000800
105
#define FERR_NF_M14ERR 0x00000400
106
#define FERR_NF_M13ERR 0x00000200
107
108
/* Uncorrectable errors */
109
#define FERR_NF_M12ERR 0x00000100
110
#define FERR_NF_M11ERR 0x00000080
111
#define FERR_NF_M10ERR 0x00000040
112
#define FERR_NF_M9ERR 0x00000020
113
#define FERR_NF_M8ERR 0x00000010
114
#define FERR_NF_M7ERR 0x00000008
115
#define FERR_NF_M6ERR 0x00000004
116
#define FERR_NF_M5ERR 0x00000002
117
#define FERR_NF_M4ERR 0x00000001
118
119
#define FERR_NF_UNCORRECTABLE (FERR_NF_M12ERR | \
120
FERR_NF_M11ERR | \
121
FERR_NF_M10ERR | \
122
FERR_NF_M9ERR | \
123
FERR_NF_M8ERR | \
124
FERR_NF_M7ERR | \
125
FERR_NF_M6ERR | \
126
FERR_NF_M5ERR | \
127
FERR_NF_M4ERR)
128
#define FERR_NF_CORRECTABLE (FERR_NF_M20ERR | \
129
FERR_NF_M19ERR | \
130
FERR_NF_M18ERR | \
131
FERR_NF_M17ERR)
132
#define FERR_NF_DIMM_SPARE (FERR_NF_M27ERR | \
133
FERR_NF_M28ERR)
134
#define FERR_NF_THERMAL (FERR_NF_M26ERR | \
135
FERR_NF_M25ERR | \
136
FERR_NF_M24ERR | \
137
FERR_NF_M23ERR)
138
#define FERR_NF_SPD_PROTOCOL (FERR_NF_M22ERR)
139
#define FERR_NF_NORTH_CRC (FERR_NF_M21ERR)
140
#define FERR_NF_NON_RETRY (FERR_NF_M13ERR | \
141
FERR_NF_M14ERR | \
142
FERR_NF_M15ERR)
143
144
#define NERR_NF_FBD 0xA4
145
#define FERR_NF_MASK (FERR_NF_UNCORRECTABLE | \
146
FERR_NF_CORRECTABLE | \
147
FERR_NF_DIMM_SPARE | \
148
FERR_NF_THERMAL | \
149
FERR_NF_SPD_PROTOCOL | \
150
FERR_NF_NORTH_CRC | \
151
FERR_NF_NON_RETRY)
152
153
#define EMASK_FBD 0xA8
154
#define EMASK_FBD_M28ERR 0x08000000
155
#define EMASK_FBD_M27ERR 0x04000000
156
#define EMASK_FBD_M26ERR 0x02000000
157
#define EMASK_FBD_M25ERR 0x01000000
158
#define EMASK_FBD_M24ERR 0x00800000
159
#define EMASK_FBD_M23ERR 0x00400000
160
#define EMASK_FBD_M22ERR 0x00200000
161
#define EMASK_FBD_M21ERR 0x00100000
162
#define EMASK_FBD_M20ERR 0x00080000
163
#define EMASK_FBD_M19ERR 0x00040000
164
#define EMASK_FBD_M18ERR 0x00020000
165
#define EMASK_FBD_M17ERR 0x00010000
166
167
#define EMASK_FBD_M15ERR 0x00004000
168
#define EMASK_FBD_M14ERR 0x00002000
169
#define EMASK_FBD_M13ERR 0x00001000
170
#define EMASK_FBD_M12ERR 0x00000800
171
#define EMASK_FBD_M11ERR 0x00000400
172
#define EMASK_FBD_M10ERR 0x00000200
173
#define EMASK_FBD_M9ERR 0x00000100
174
#define EMASK_FBD_M8ERR 0x00000080
175
#define EMASK_FBD_M7ERR 0x00000040
176
#define EMASK_FBD_M6ERR 0x00000020
177
#define EMASK_FBD_M5ERR 0x00000010
178
#define EMASK_FBD_M4ERR 0x00000008
179
#define EMASK_FBD_M3ERR 0x00000004
180
#define EMASK_FBD_M2ERR 0x00000002
181
#define EMASK_FBD_M1ERR 0x00000001
182
183
#define ENABLE_EMASK_FBD_FATAL_ERRORS (EMASK_FBD_M1ERR | \
184
EMASK_FBD_M2ERR | \
185
EMASK_FBD_M3ERR)
186
187
#define ENABLE_EMASK_FBD_UNCORRECTABLE (EMASK_FBD_M4ERR | \
188
EMASK_FBD_M5ERR | \
189
EMASK_FBD_M6ERR | \
190
EMASK_FBD_M7ERR | \
191
EMASK_FBD_M8ERR | \
192
EMASK_FBD_M9ERR | \
193
EMASK_FBD_M10ERR | \
194
EMASK_FBD_M11ERR | \
195
EMASK_FBD_M12ERR)
196
#define ENABLE_EMASK_FBD_CORRECTABLE (EMASK_FBD_M17ERR | \
197
EMASK_FBD_M18ERR | \
198
EMASK_FBD_M19ERR | \
199
EMASK_FBD_M20ERR)
200
#define ENABLE_EMASK_FBD_DIMM_SPARE (EMASK_FBD_M27ERR | \
201
EMASK_FBD_M28ERR)
202
#define ENABLE_EMASK_FBD_THERMALS (EMASK_FBD_M26ERR | \
203
EMASK_FBD_M25ERR | \
204
EMASK_FBD_M24ERR | \
205
EMASK_FBD_M23ERR)
206
#define ENABLE_EMASK_FBD_SPD_PROTOCOL (EMASK_FBD_M22ERR)
207
#define ENABLE_EMASK_FBD_NORTH_CRC (EMASK_FBD_M21ERR)
208
#define ENABLE_EMASK_FBD_NON_RETRY (EMASK_FBD_M15ERR | \
209
EMASK_FBD_M14ERR | \
210
EMASK_FBD_M13ERR)
211
212
#define ENABLE_EMASK_ALL (ENABLE_EMASK_FBD_NON_RETRY | \
213
ENABLE_EMASK_FBD_NORTH_CRC | \
214
ENABLE_EMASK_FBD_SPD_PROTOCOL | \
215
ENABLE_EMASK_FBD_THERMALS | \
216
ENABLE_EMASK_FBD_DIMM_SPARE | \
217
ENABLE_EMASK_FBD_FATAL_ERRORS | \
218
ENABLE_EMASK_FBD_CORRECTABLE | \
219
ENABLE_EMASK_FBD_UNCORRECTABLE)
220
221
#define ERR0_FBD 0xAC
222
#define ERR1_FBD 0xB0
223
#define ERR2_FBD 0xB4
224
#define MCERR_FBD 0xB8
225
#define NRECMEMA 0xBE
226
#define NREC_BANK(x) (((x)>>12) & 0x7)
227
#define NREC_RDWR(x) (((x)>>11) & 1)
228
#define NREC_RANK(x) (((x)>>8) & 0x7)
229
#define NRECMEMB 0xC0
230
#define NREC_CAS(x) (((x)>>16) & 0xFFFFFF)
231
#define NREC_RAS(x) ((x) & 0x7FFF)
232
#define NRECFGLOG 0xC4
233
#define NREEECFBDA 0xC8
234
#define NREEECFBDB 0xCC
235
#define NREEECFBDC 0xD0
236
#define NREEECFBDD 0xD4
237
#define NREEECFBDE 0xD8
238
#define REDMEMA 0xDC
239
#define RECMEMA 0xE2
240
#define REC_BANK(x) (((x)>>12) & 0x7)
241
#define REC_RDWR(x) (((x)>>11) & 1)
242
#define REC_RANK(x) (((x)>>8) & 0x7)
243
#define RECMEMB 0xE4
244
#define REC_CAS(x) (((x)>>16) & 0xFFFFFF)
245
#define REC_RAS(x) ((x) & 0x7FFF)
246
#define RECFGLOG 0xE8
247
#define RECFBDA 0xEC
248
#define RECFBDB 0xF0
249
#define RECFBDC 0xF4
250
#define RECFBDD 0xF8
251
#define RECFBDE 0xFC
252
253
/* OFFSETS for Function 2 */
254
255
/*
256
* Device 21,
257
* Function 0: Memory Map Branch 0
258
*
259
* Device 22,
260
* Function 0: Memory Map Branch 1
261
*/
262
#define PCI_DEVICE_ID_I5000_BRANCH_0 0x25F5
263
#define PCI_DEVICE_ID_I5000_BRANCH_1 0x25F6
264
265
#define AMB_PRESENT_0 0x64
266
#define AMB_PRESENT_1 0x66
267
#define MTR0 0x80
268
#define MTR1 0x84
269
#define MTR2 0x88
270
#define MTR3 0x8C
271
272
#define NUM_MTRS 4
273
#define CHANNELS_PER_BRANCH (2)
274
275
/* Defines to extract the vaious fields from the
276
* MTRx - Memory Technology Registers
277
*/
278
#define MTR_DIMMS_PRESENT(mtr) ((mtr) & (0x1 << 8))
279
#define MTR_DRAM_WIDTH(mtr) ((((mtr) >> 6) & 0x1) ? 8 : 4)
280
#define MTR_DRAM_BANKS(mtr) ((((mtr) >> 5) & 0x1) ? 8 : 4)
281
#define MTR_DRAM_BANKS_ADDR_BITS(mtr) ((MTR_DRAM_BANKS(mtr) == 8) ? 3 : 2)
282
#define MTR_DIMM_RANK(mtr) (((mtr) >> 4) & 0x1)
283
#define MTR_DIMM_RANK_ADDR_BITS(mtr) (MTR_DIMM_RANK(mtr) ? 2 : 1)
284
#define MTR_DIMM_ROWS(mtr) (((mtr) >> 2) & 0x3)
285
#define MTR_DIMM_ROWS_ADDR_BITS(mtr) (MTR_DIMM_ROWS(mtr) + 13)
286
#define MTR_DIMM_COLS(mtr) ((mtr) & 0x3)
287
#define MTR_DIMM_COLS_ADDR_BITS(mtr) (MTR_DIMM_COLS(mtr) + 10)
288
289
#ifdef CONFIG_EDAC_DEBUG
290
static char *numrow_toString[] = {
291
"8,192 - 13 rows",
292
"16,384 - 14 rows",
293
"32,768 - 15 rows",
294
"reserved"
295
};
296
297
static char *numcol_toString[] = {
298
"1,024 - 10 columns",
299
"2,048 - 11 columns",
300
"4,096 - 12 columns",
301
"reserved"
302
};
303
#endif
304
305
/* enables the report of miscellaneous messages as CE errors - default off */
306
static int misc_messages;
307
308
/* Enumeration of supported devices */
309
enum i5000_chips {
310
I5000P = 0,
311
I5000V = 1, /* future */
312
I5000X = 2 /* future */
313
};
314
315
/* Device name and register DID (Device ID) */
316
struct i5000_dev_info {
317
const char *ctl_name; /* name for this device */
318
u16 fsb_mapping_errors; /* DID for the branchmap,control */
319
};
320
321
/* Table of devices attributes supported by this driver */
322
static const struct i5000_dev_info i5000_devs[] = {
323
[I5000P] = {
324
.ctl_name = "I5000",
325
.fsb_mapping_errors = PCI_DEVICE_ID_INTEL_I5000_DEV16,
326
},
327
};
328
329
struct i5000_dimm_info {
330
int megabytes; /* size, 0 means not present */
331
int dual_rank;
332
};
333
334
#define MAX_CHANNELS 6 /* max possible channels */
335
#define MAX_CSROWS (8*2) /* max possible csrows per channel */
336
337
/* driver private data structure */
338
struct i5000_pvt {
339
struct pci_dev *system_address; /* 16.0 */
340
struct pci_dev *branchmap_werrors; /* 16.1 */
341
struct pci_dev *fsb_error_regs; /* 16.2 */
342
struct pci_dev *branch_0; /* 21.0 */
343
struct pci_dev *branch_1; /* 22.0 */
344
345
u16 tolm; /* top of low memory */
346
u64 ambase; /* AMB BAR */
347
348
u16 mir0, mir1, mir2;
349
350
u16 b0_mtr[NUM_MTRS]; /* Memory Technlogy Reg */
351
u16 b0_ambpresent0; /* Branch 0, Channel 0 */
352
u16 b0_ambpresent1; /* Brnach 0, Channel 1 */
353
354
u16 b1_mtr[NUM_MTRS]; /* Memory Technlogy Reg */
355
u16 b1_ambpresent0; /* Branch 1, Channel 8 */
356
u16 b1_ambpresent1; /* Branch 1, Channel 1 */
357
358
/* DIMM information matrix, allocating architecture maximums */
359
struct i5000_dimm_info dimm_info[MAX_CSROWS][MAX_CHANNELS];
360
361
/* Actual values for this controller */
362
int maxch; /* Max channels */
363
int maxdimmperch; /* Max DIMMs per channel */
364
};
365
366
/* I5000 MCH error information retrieved from Hardware */
367
struct i5000_error_info {
368
369
/* These registers are always read from the MC */
370
u32 ferr_fat_fbd; /* First Errors Fatal */
371
u32 nerr_fat_fbd; /* Next Errors Fatal */
372
u32 ferr_nf_fbd; /* First Errors Non-Fatal */
373
u32 nerr_nf_fbd; /* Next Errors Non-Fatal */
374
375
/* These registers are input ONLY if there was a Recoverable Error */
376
u32 redmemb; /* Recoverable Mem Data Error log B */
377
u16 recmema; /* Recoverable Mem Error log A */
378
u32 recmemb; /* Recoverable Mem Error log B */
379
380
/* These registers are input ONLY if there was a
381
* Non-Recoverable Error */
382
u16 nrecmema; /* Non-Recoverable Mem log A */
383
u16 nrecmemb; /* Non-Recoverable Mem log B */
384
385
};
386
387
static struct edac_pci_ctl_info *i5000_pci;
388
389
/*
390
* i5000_get_error_info Retrieve the hardware error information from
391
* the hardware and cache it in the 'info'
392
* structure
393
*/
394
static void i5000_get_error_info(struct mem_ctl_info *mci,
395
struct i5000_error_info *info)
396
{
397
struct i5000_pvt *pvt;
398
u32 value;
399
400
pvt = mci->pvt_info;
401
402
/* read in the 1st FATAL error register */
403
pci_read_config_dword(pvt->branchmap_werrors, FERR_FAT_FBD, &value);
404
405
/* Mask only the bits that the doc says are valid
406
*/
407
value &= (FERR_FAT_FBDCHAN | FERR_FAT_MASK);
408
409
/* If there is an error, then read in the */
410
/* NEXT FATAL error register and the Memory Error Log Register A */
411
if (value & FERR_FAT_MASK) {
412
info->ferr_fat_fbd = value;
413
414
/* harvest the various error data we need */
415
pci_read_config_dword(pvt->branchmap_werrors,
416
NERR_FAT_FBD, &info->nerr_fat_fbd);
417
pci_read_config_word(pvt->branchmap_werrors,
418
NRECMEMA, &info->nrecmema);
419
pci_read_config_word(pvt->branchmap_werrors,
420
NRECMEMB, &info->nrecmemb);
421
422
/* Clear the error bits, by writing them back */
423
pci_write_config_dword(pvt->branchmap_werrors,
424
FERR_FAT_FBD, value);
425
} else {
426
info->ferr_fat_fbd = 0;
427
info->nerr_fat_fbd = 0;
428
info->nrecmema = 0;
429
info->nrecmemb = 0;
430
}
431
432
/* read in the 1st NON-FATAL error register */
433
pci_read_config_dword(pvt->branchmap_werrors, FERR_NF_FBD, &value);
434
435
/* If there is an error, then read in the 1st NON-FATAL error
436
* register as well */
437
if (value & FERR_NF_MASK) {
438
info->ferr_nf_fbd = value;
439
440
/* harvest the various error data we need */
441
pci_read_config_dword(pvt->branchmap_werrors,
442
NERR_NF_FBD, &info->nerr_nf_fbd);
443
pci_read_config_word(pvt->branchmap_werrors,
444
RECMEMA, &info->recmema);
445
pci_read_config_dword(pvt->branchmap_werrors,
446
RECMEMB, &info->recmemb);
447
pci_read_config_dword(pvt->branchmap_werrors,
448
REDMEMB, &info->redmemb);
449
450
/* Clear the error bits, by writing them back */
451
pci_write_config_dword(pvt->branchmap_werrors,
452
FERR_NF_FBD, value);
453
} else {
454
info->ferr_nf_fbd = 0;
455
info->nerr_nf_fbd = 0;
456
info->recmema = 0;
457
info->recmemb = 0;
458
info->redmemb = 0;
459
}
460
}
461
462
/*
463
* i5000_process_fatal_error_info(struct mem_ctl_info *mci,
464
* struct i5000_error_info *info,
465
* int handle_errors);
466
*
467
* handle the Intel FATAL errors, if any
468
*/
469
static void i5000_process_fatal_error_info(struct mem_ctl_info *mci,
470
struct i5000_error_info *info,
471
int handle_errors)
472
{
473
char msg[EDAC_MC_LABEL_LEN + 1 + 160];
474
char *specific = NULL;
475
u32 allErrors;
476
int branch;
477
int channel;
478
int bank;
479
int rank;
480
int rdwr;
481
int ras, cas;
482
483
/* mask off the Error bits that are possible */
484
allErrors = (info->ferr_fat_fbd & FERR_FAT_MASK);
485
if (!allErrors)
486
return; /* if no error, return now */
487
488
branch = EXTRACT_FBDCHAN_INDX(info->ferr_fat_fbd);
489
channel = branch;
490
491
/* Use the NON-Recoverable macros to extract data */
492
bank = NREC_BANK(info->nrecmema);
493
rank = NREC_RANK(info->nrecmema);
494
rdwr = NREC_RDWR(info->nrecmema);
495
ras = NREC_RAS(info->nrecmemb);
496
cas = NREC_CAS(info->nrecmemb);
497
498
debugf0("\t\tCSROW= %d Channels= %d,%d (Branch= %d "
499
"DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
500
rank, channel, channel + 1, branch >> 1, bank,
501
rdwr ? "Write" : "Read", ras, cas);
502
503
/* Only 1 bit will be on */
504
switch (allErrors) {
505
case FERR_FAT_M1ERR:
506
specific = "Alert on non-redundant retry or fast "
507
"reset timeout";
508
break;
509
case FERR_FAT_M2ERR:
510
specific = "Northbound CRC error on non-redundant "
511
"retry";
512
break;
513
case FERR_FAT_M3ERR:
514
{
515
static int done;
516
517
/*
518
* This error is generated to inform that the intelligent
519
* throttling is disabled and the temperature passed the
520
* specified middle point. Since this is something the BIOS
521
* should take care of, we'll warn only once to avoid
522
* worthlessly flooding the log.
523
*/
524
if (done)
525
return;
526
done++;
527
528
specific = ">Tmid Thermal event with intelligent "
529
"throttling disabled";
530
}
531
break;
532
}
533
534
/* Form out message */
535
snprintf(msg, sizeof(msg),
536
"(Branch=%d DRAM-Bank=%d RDWR=%s RAS=%d CAS=%d "
537
"FATAL Err=0x%x (%s))",
538
branch >> 1, bank, rdwr ? "Write" : "Read", ras, cas,
539
allErrors, specific);
540
541
/* Call the helper to output message */
542
edac_mc_handle_fbd_ue(mci, rank, channel, channel + 1, msg);
543
}
544
545
/*
546
* i5000_process_fatal_error_info(struct mem_ctl_info *mci,
547
* struct i5000_error_info *info,
548
* int handle_errors);
549
*
550
* handle the Intel NON-FATAL errors, if any
551
*/
552
static void i5000_process_nonfatal_error_info(struct mem_ctl_info *mci,
553
struct i5000_error_info *info,
554
int handle_errors)
555
{
556
char msg[EDAC_MC_LABEL_LEN + 1 + 170];
557
char *specific = NULL;
558
u32 allErrors;
559
u32 ue_errors;
560
u32 ce_errors;
561
u32 misc_errors;
562
int branch;
563
int channel;
564
int bank;
565
int rank;
566
int rdwr;
567
int ras, cas;
568
569
/* mask off the Error bits that are possible */
570
allErrors = (info->ferr_nf_fbd & FERR_NF_MASK);
571
if (!allErrors)
572
return; /* if no error, return now */
573
574
/* ONLY ONE of the possible error bits will be set, as per the docs */
575
ue_errors = allErrors & FERR_NF_UNCORRECTABLE;
576
if (ue_errors) {
577
debugf0("\tUncorrected bits= 0x%x\n", ue_errors);
578
579
branch = EXTRACT_FBDCHAN_INDX(info->ferr_nf_fbd);
580
581
/*
582
* According with i5000 datasheet, bit 28 has no significance
583
* for errors M4Err-M12Err and M17Err-M21Err, on FERR_NF_FBD
584
*/
585
channel = branch & 2;
586
587
bank = NREC_BANK(info->nrecmema);
588
rank = NREC_RANK(info->nrecmema);
589
rdwr = NREC_RDWR(info->nrecmema);
590
ras = NREC_RAS(info->nrecmemb);
591
cas = NREC_CAS(info->nrecmemb);
592
593
debugf0
594
("\t\tCSROW= %d Channels= %d,%d (Branch= %d "
595
"DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
596
rank, channel, channel + 1, branch >> 1, bank,
597
rdwr ? "Write" : "Read", ras, cas);
598
599
switch (ue_errors) {
600
case FERR_NF_M12ERR:
601
specific = "Non-Aliased Uncorrectable Patrol Data ECC";
602
break;
603
case FERR_NF_M11ERR:
604
specific = "Non-Aliased Uncorrectable Spare-Copy "
605
"Data ECC";
606
break;
607
case FERR_NF_M10ERR:
608
specific = "Non-Aliased Uncorrectable Mirrored Demand "
609
"Data ECC";
610
break;
611
case FERR_NF_M9ERR:
612
specific = "Non-Aliased Uncorrectable Non-Mirrored "
613
"Demand Data ECC";
614
break;
615
case FERR_NF_M8ERR:
616
specific = "Aliased Uncorrectable Patrol Data ECC";
617
break;
618
case FERR_NF_M7ERR:
619
specific = "Aliased Uncorrectable Spare-Copy Data ECC";
620
break;
621
case FERR_NF_M6ERR:
622
specific = "Aliased Uncorrectable Mirrored Demand "
623
"Data ECC";
624
break;
625
case FERR_NF_M5ERR:
626
specific = "Aliased Uncorrectable Non-Mirrored Demand "
627
"Data ECC";
628
break;
629
case FERR_NF_M4ERR:
630
specific = "Uncorrectable Data ECC on Replay";
631
break;
632
}
633
634
/* Form out message */
635
snprintf(msg, sizeof(msg),
636
"(Branch=%d DRAM-Bank=%d RDWR=%s RAS=%d "
637
"CAS=%d, UE Err=0x%x (%s))",
638
branch >> 1, bank, rdwr ? "Write" : "Read", ras, cas,
639
ue_errors, specific);
640
641
/* Call the helper to output message */
642
edac_mc_handle_fbd_ue(mci, rank, channel, channel + 1, msg);
643
}
644
645
/* Check correctable errors */
646
ce_errors = allErrors & FERR_NF_CORRECTABLE;
647
if (ce_errors) {
648
debugf0("\tCorrected bits= 0x%x\n", ce_errors);
649
650
branch = EXTRACT_FBDCHAN_INDX(info->ferr_nf_fbd);
651
652
channel = 0;
653
if (REC_ECC_LOCATOR_ODD(info->redmemb))
654
channel = 1;
655
656
/* Convert channel to be based from zero, instead of
657
* from branch base of 0 */
658
channel += branch;
659
660
bank = REC_BANK(info->recmema);
661
rank = REC_RANK(info->recmema);
662
rdwr = REC_RDWR(info->recmema);
663
ras = REC_RAS(info->recmemb);
664
cas = REC_CAS(info->recmemb);
665
666
debugf0("\t\tCSROW= %d Channel= %d (Branch %d "
667
"DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
668
rank, channel, branch >> 1, bank,
669
rdwr ? "Write" : "Read", ras, cas);
670
671
switch (ce_errors) {
672
case FERR_NF_M17ERR:
673
specific = "Correctable Non-Mirrored Demand Data ECC";
674
break;
675
case FERR_NF_M18ERR:
676
specific = "Correctable Mirrored Demand Data ECC";
677
break;
678
case FERR_NF_M19ERR:
679
specific = "Correctable Spare-Copy Data ECC";
680
break;
681
case FERR_NF_M20ERR:
682
specific = "Correctable Patrol Data ECC";
683
break;
684
}
685
686
/* Form out message */
687
snprintf(msg, sizeof(msg),
688
"(Branch=%d DRAM-Bank=%d RDWR=%s RAS=%d "
689
"CAS=%d, CE Err=0x%x (%s))", branch >> 1, bank,
690
rdwr ? "Write" : "Read", ras, cas, ce_errors,
691
specific);
692
693
/* Call the helper to output message */
694
edac_mc_handle_fbd_ce(mci, rank, channel, msg);
695
}
696
697
if (!misc_messages)
698
return;
699
700
misc_errors = allErrors & (FERR_NF_NON_RETRY | FERR_NF_NORTH_CRC |
701
FERR_NF_SPD_PROTOCOL | FERR_NF_DIMM_SPARE);
702
if (misc_errors) {
703
switch (misc_errors) {
704
case FERR_NF_M13ERR:
705
specific = "Non-Retry or Redundant Retry FBD Memory "
706
"Alert or Redundant Fast Reset Timeout";
707
break;
708
case FERR_NF_M14ERR:
709
specific = "Non-Retry or Redundant Retry FBD "
710
"Configuration Alert";
711
break;
712
case FERR_NF_M15ERR:
713
specific = "Non-Retry or Redundant Retry FBD "
714
"Northbound CRC error on read data";
715
break;
716
case FERR_NF_M21ERR:
717
specific = "FBD Northbound CRC error on "
718
"FBD Sync Status";
719
break;
720
case FERR_NF_M22ERR:
721
specific = "SPD protocol error";
722
break;
723
case FERR_NF_M27ERR:
724
specific = "DIMM-spare copy started";
725
break;
726
case FERR_NF_M28ERR:
727
specific = "DIMM-spare copy completed";
728
break;
729
}
730
branch = EXTRACT_FBDCHAN_INDX(info->ferr_nf_fbd);
731
732
/* Form out message */
733
snprintf(msg, sizeof(msg),
734
"(Branch=%d Err=%#x (%s))", branch >> 1,
735
misc_errors, specific);
736
737
/* Call the helper to output message */
738
edac_mc_handle_fbd_ce(mci, 0, 0, msg);
739
}
740
}
741
742
/*
743
* i5000_process_error_info Process the error info that is
744
* in the 'info' structure, previously retrieved from hardware
745
*/
746
static void i5000_process_error_info(struct mem_ctl_info *mci,
747
struct i5000_error_info *info,
748
int handle_errors)
749
{
750
/* First handle any fatal errors that occurred */
751
i5000_process_fatal_error_info(mci, info, handle_errors);
752
753
/* now handle any non-fatal errors that occurred */
754
i5000_process_nonfatal_error_info(mci, info, handle_errors);
755
}
756
757
/*
758
* i5000_clear_error Retrieve any error from the hardware
759
* but do NOT process that error.
760
* Used for 'clearing' out of previous errors
761
* Called by the Core module.
762
*/
763
static void i5000_clear_error(struct mem_ctl_info *mci)
764
{
765
struct i5000_error_info info;
766
767
i5000_get_error_info(mci, &info);
768
}
769
770
/*
771
* i5000_check_error Retrieve and process errors reported by the
772
* hardware. Called by the Core module.
773
*/
774
static void i5000_check_error(struct mem_ctl_info *mci)
775
{
776
struct i5000_error_info info;
777
debugf4("MC%d: %s: %s()\n", mci->mc_idx, __FILE__, __func__);
778
i5000_get_error_info(mci, &info);
779
i5000_process_error_info(mci, &info, 1);
780
}
781
782
/*
783
* i5000_get_devices Find and perform 'get' operation on the MCH's
784
* device/functions we want to reference for this driver
785
*
786
* Need to 'get' device 16 func 1 and func 2
787
*/
788
static int i5000_get_devices(struct mem_ctl_info *mci, int dev_idx)
789
{
790
//const struct i5000_dev_info *i5000_dev = &i5000_devs[dev_idx];
791
struct i5000_pvt *pvt;
792
struct pci_dev *pdev;
793
794
pvt = mci->pvt_info;
795
796
/* Attempt to 'get' the MCH register we want */
797
pdev = NULL;
798
while (1) {
799
pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
800
PCI_DEVICE_ID_INTEL_I5000_DEV16, pdev);
801
802
/* End of list, leave */
803
if (pdev == NULL) {
804
i5000_printk(KERN_ERR,
805
"'system address,Process Bus' "
806
"device not found:"
807
"vendor 0x%x device 0x%x FUNC 1 "
808
"(broken BIOS?)\n",
809
PCI_VENDOR_ID_INTEL,
810
PCI_DEVICE_ID_INTEL_I5000_DEV16);
811
812
return 1;
813
}
814
815
/* Scan for device 16 func 1 */
816
if (PCI_FUNC(pdev->devfn) == 1)
817
break;
818
}
819
820
pvt->branchmap_werrors = pdev;
821
822
/* Attempt to 'get' the MCH register we want */
823
pdev = NULL;
824
while (1) {
825
pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
826
PCI_DEVICE_ID_INTEL_I5000_DEV16, pdev);
827
828
if (pdev == NULL) {
829
i5000_printk(KERN_ERR,
830
"MC: 'branchmap,control,errors' "
831
"device not found:"
832
"vendor 0x%x device 0x%x Func 2 "
833
"(broken BIOS?)\n",
834
PCI_VENDOR_ID_INTEL,
835
PCI_DEVICE_ID_INTEL_I5000_DEV16);
836
837
pci_dev_put(pvt->branchmap_werrors);
838
return 1;
839
}
840
841
/* Scan for device 16 func 1 */
842
if (PCI_FUNC(pdev->devfn) == 2)
843
break;
844
}
845
846
pvt->fsb_error_regs = pdev;
847
848
debugf1("System Address, processor bus- PCI Bus ID: %s %x:%x\n",
849
pci_name(pvt->system_address),
850
pvt->system_address->vendor, pvt->system_address->device);
851
debugf1("Branchmap, control and errors - PCI Bus ID: %s %x:%x\n",
852
pci_name(pvt->branchmap_werrors),
853
pvt->branchmap_werrors->vendor, pvt->branchmap_werrors->device);
854
debugf1("FSB Error Regs - PCI Bus ID: %s %x:%x\n",
855
pci_name(pvt->fsb_error_regs),
856
pvt->fsb_error_regs->vendor, pvt->fsb_error_regs->device);
857
858
pdev = NULL;
859
pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
860
PCI_DEVICE_ID_I5000_BRANCH_0, pdev);
861
862
if (pdev == NULL) {
863
i5000_printk(KERN_ERR,
864
"MC: 'BRANCH 0' device not found:"
865
"vendor 0x%x device 0x%x Func 0 (broken BIOS?)\n",
866
PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_I5000_BRANCH_0);
867
868
pci_dev_put(pvt->branchmap_werrors);
869
pci_dev_put(pvt->fsb_error_regs);
870
return 1;
871
}
872
873
pvt->branch_0 = pdev;
874
875
/* If this device claims to have more than 2 channels then
876
* fetch Branch 1's information
877
*/
878
if (pvt->maxch >= CHANNELS_PER_BRANCH) {
879
pdev = NULL;
880
pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
881
PCI_DEVICE_ID_I5000_BRANCH_1, pdev);
882
883
if (pdev == NULL) {
884
i5000_printk(KERN_ERR,
885
"MC: 'BRANCH 1' device not found:"
886
"vendor 0x%x device 0x%x Func 0 "
887
"(broken BIOS?)\n",
888
PCI_VENDOR_ID_INTEL,
889
PCI_DEVICE_ID_I5000_BRANCH_1);
890
891
pci_dev_put(pvt->branchmap_werrors);
892
pci_dev_put(pvt->fsb_error_regs);
893
pci_dev_put(pvt->branch_0);
894
return 1;
895
}
896
897
pvt->branch_1 = pdev;
898
}
899
900
return 0;
901
}
902
903
/*
904
* i5000_put_devices 'put' all the devices that we have
905
* reserved via 'get'
906
*/
907
static void i5000_put_devices(struct mem_ctl_info *mci)
908
{
909
struct i5000_pvt *pvt;
910
911
pvt = mci->pvt_info;
912
913
pci_dev_put(pvt->branchmap_werrors); /* FUNC 1 */
914
pci_dev_put(pvt->fsb_error_regs); /* FUNC 2 */
915
pci_dev_put(pvt->branch_0); /* DEV 21 */
916
917
/* Only if more than 2 channels do we release the second branch */
918
if (pvt->maxch >= CHANNELS_PER_BRANCH)
919
pci_dev_put(pvt->branch_1); /* DEV 22 */
920
}
921
922
/*
923
* determine_amb_resent
924
*
925
* the information is contained in NUM_MTRS different registers
926
* determineing which of the NUM_MTRS requires knowing
927
* which channel is in question
928
*
929
* 2 branches, each with 2 channels
930
* b0_ambpresent0 for channel '0'
931
* b0_ambpresent1 for channel '1'
932
* b1_ambpresent0 for channel '2'
933
* b1_ambpresent1 for channel '3'
934
*/
935
static int determine_amb_present_reg(struct i5000_pvt *pvt, int channel)
936
{
937
int amb_present;
938
939
if (channel < CHANNELS_PER_BRANCH) {
940
if (channel & 0x1)
941
amb_present = pvt->b0_ambpresent1;
942
else
943
amb_present = pvt->b0_ambpresent0;
944
} else {
945
if (channel & 0x1)
946
amb_present = pvt->b1_ambpresent1;
947
else
948
amb_present = pvt->b1_ambpresent0;
949
}
950
951
return amb_present;
952
}
953
954
/*
955
* determine_mtr(pvt, csrow, channel)
956
*
957
* return the proper MTR register as determine by the csrow and channel desired
958
*/
959
static int determine_mtr(struct i5000_pvt *pvt, int csrow, int channel)
960
{
961
int mtr;
962
963
if (channel < CHANNELS_PER_BRANCH)
964
mtr = pvt->b0_mtr[csrow >> 1];
965
else
966
mtr = pvt->b1_mtr[csrow >> 1];
967
968
return mtr;
969
}
970
971
/*
972
*/
973
static void decode_mtr(int slot_row, u16 mtr)
974
{
975
int ans;
976
977
ans = MTR_DIMMS_PRESENT(mtr);
978
979
debugf2("\tMTR%d=0x%x: DIMMs are %s\n", slot_row, mtr,
980
ans ? "Present" : "NOT Present");
981
if (!ans)
982
return;
983
984
debugf2("\t\tWIDTH: x%d\n", MTR_DRAM_WIDTH(mtr));
985
debugf2("\t\tNUMBANK: %d bank(s)\n", MTR_DRAM_BANKS(mtr));
986
debugf2("\t\tNUMRANK: %s\n", MTR_DIMM_RANK(mtr) ? "double" : "single");
987
debugf2("\t\tNUMROW: %s\n", numrow_toString[MTR_DIMM_ROWS(mtr)]);
988
debugf2("\t\tNUMCOL: %s\n", numcol_toString[MTR_DIMM_COLS(mtr)]);
989
}
990
991
static void handle_channel(struct i5000_pvt *pvt, int csrow, int channel,
992
struct i5000_dimm_info *dinfo)
993
{
994
int mtr;
995
int amb_present_reg;
996
int addrBits;
997
998
mtr = determine_mtr(pvt, csrow, channel);
999
if (MTR_DIMMS_PRESENT(mtr)) {
1000
amb_present_reg = determine_amb_present_reg(pvt, channel);
1001
1002
/* Determine if there is a DIMM present in this DIMM slot */
1003
if (amb_present_reg & (1 << (csrow >> 1))) {
1004
dinfo->dual_rank = MTR_DIMM_RANK(mtr);
1005
1006
if (!((dinfo->dual_rank == 0) &&
1007
((csrow & 0x1) == 0x1))) {
1008
/* Start with the number of bits for a Bank
1009
* on the DRAM */
1010
addrBits = MTR_DRAM_BANKS_ADDR_BITS(mtr);
1011
/* Add thenumber of ROW bits */
1012
addrBits += MTR_DIMM_ROWS_ADDR_BITS(mtr);
1013
/* add the number of COLUMN bits */
1014
addrBits += MTR_DIMM_COLS_ADDR_BITS(mtr);
1015
1016
addrBits += 6; /* add 64 bits per DIMM */
1017
addrBits -= 20; /* divide by 2^^20 */
1018
addrBits -= 3; /* 8 bits per bytes */
1019
1020
dinfo->megabytes = 1 << addrBits;
1021
}
1022
}
1023
}
1024
}
1025
1026
/*
1027
* calculate_dimm_size
1028
*
1029
* also will output a DIMM matrix map, if debug is enabled, for viewing
1030
* how the DIMMs are populated
1031
*/
1032
static void calculate_dimm_size(struct i5000_pvt *pvt)
1033
{
1034
struct i5000_dimm_info *dinfo;
1035
int csrow, max_csrows;
1036
char *p, *mem_buffer;
1037
int space, n;
1038
int channel;
1039
1040
/* ================= Generate some debug output ================= */
1041
space = PAGE_SIZE;
1042
mem_buffer = p = kmalloc(space, GFP_KERNEL);
1043
if (p == NULL) {
1044
i5000_printk(KERN_ERR, "MC: %s:%s() kmalloc() failed\n",
1045
__FILE__, __func__);
1046
return;
1047
}
1048
1049
n = snprintf(p, space, "\n");
1050
p += n;
1051
space -= n;
1052
1053
/* Scan all the actual CSROWS (which is # of DIMMS * 2)
1054
* and calculate the information for each DIMM
1055
* Start with the highest csrow first, to display it first
1056
* and work toward the 0th csrow
1057
*/
1058
max_csrows = pvt->maxdimmperch * 2;
1059
for (csrow = max_csrows - 1; csrow >= 0; csrow--) {
1060
1061
/* on an odd csrow, first output a 'boundary' marker,
1062
* then reset the message buffer */
1063
if (csrow & 0x1) {
1064
n = snprintf(p, space, "---------------------------"
1065
"--------------------------------");
1066
p += n;
1067
space -= n;
1068
debugf2("%s\n", mem_buffer);
1069
p = mem_buffer;
1070
space = PAGE_SIZE;
1071
}
1072
n = snprintf(p, space, "csrow %2d ", csrow);
1073
p += n;
1074
space -= n;
1075
1076
for (channel = 0; channel < pvt->maxch; channel++) {
1077
dinfo = &pvt->dimm_info[csrow][channel];
1078
handle_channel(pvt, csrow, channel, dinfo);
1079
n = snprintf(p, space, "%4d MB | ", dinfo->megabytes);
1080
p += n;
1081
space -= n;
1082
}
1083
n = snprintf(p, space, "\n");
1084
p += n;
1085
space -= n;
1086
}
1087
1088
/* Output the last bottom 'boundary' marker */
1089
n = snprintf(p, space, "---------------------------"
1090
"--------------------------------\n");
1091
p += n;
1092
space -= n;
1093
1094
/* now output the 'channel' labels */
1095
n = snprintf(p, space, " ");
1096
p += n;
1097
space -= n;
1098
for (channel = 0; channel < pvt->maxch; channel++) {
1099
n = snprintf(p, space, "channel %d | ", channel);
1100
p += n;
1101
space -= n;
1102
}
1103
n = snprintf(p, space, "\n");
1104
p += n;
1105
space -= n;
1106
1107
/* output the last message and free buffer */
1108
debugf2("%s\n", mem_buffer);
1109
kfree(mem_buffer);
1110
}
1111
1112
/*
1113
* i5000_get_mc_regs read in the necessary registers and
1114
* cache locally
1115
*
1116
* Fills in the private data members
1117
*/
1118
static void i5000_get_mc_regs(struct mem_ctl_info *mci)
1119
{
1120
struct i5000_pvt *pvt;
1121
u32 actual_tolm;
1122
u16 limit;
1123
int slot_row;
1124
int maxch;
1125
int maxdimmperch;
1126
int way0, way1;
1127
1128
pvt = mci->pvt_info;
1129
1130
pci_read_config_dword(pvt->system_address, AMBASE,
1131
(u32 *) & pvt->ambase);
1132
pci_read_config_dword(pvt->system_address, AMBASE + sizeof(u32),
1133
((u32 *) & pvt->ambase) + sizeof(u32));
1134
1135
maxdimmperch = pvt->maxdimmperch;
1136
maxch = pvt->maxch;
1137
1138
debugf2("AMBASE= 0x%lx MAXCH= %d MAX-DIMM-Per-CH= %d\n",
1139
(long unsigned int)pvt->ambase, pvt->maxch, pvt->maxdimmperch);
1140
1141
/* Get the Branch Map regs */
1142
pci_read_config_word(pvt->branchmap_werrors, TOLM, &pvt->tolm);
1143
pvt->tolm >>= 12;
1144
debugf2("\nTOLM (number of 256M regions) =%u (0x%x)\n", pvt->tolm,
1145
pvt->tolm);
1146
1147
actual_tolm = pvt->tolm << 28;
1148
debugf2("Actual TOLM byte addr=%u (0x%x)\n", actual_tolm, actual_tolm);
1149
1150
pci_read_config_word(pvt->branchmap_werrors, MIR0, &pvt->mir0);
1151
pci_read_config_word(pvt->branchmap_werrors, MIR1, &pvt->mir1);
1152
pci_read_config_word(pvt->branchmap_werrors, MIR2, &pvt->mir2);
1153
1154
/* Get the MIR[0-2] regs */
1155
limit = (pvt->mir0 >> 4) & 0x0FFF;
1156
way0 = pvt->mir0 & 0x1;
1157
way1 = pvt->mir0 & 0x2;
1158
debugf2("MIR0: limit= 0x%x WAY1= %u WAY0= %x\n", limit, way1, way0);
1159
limit = (pvt->mir1 >> 4) & 0x0FFF;
1160
way0 = pvt->mir1 & 0x1;
1161
way1 = pvt->mir1 & 0x2;
1162
debugf2("MIR1: limit= 0x%x WAY1= %u WAY0= %x\n", limit, way1, way0);
1163
limit = (pvt->mir2 >> 4) & 0x0FFF;
1164
way0 = pvt->mir2 & 0x1;
1165
way1 = pvt->mir2 & 0x2;
1166
debugf2("MIR2: limit= 0x%x WAY1= %u WAY0= %x\n", limit, way1, way0);
1167
1168
/* Get the MTR[0-3] regs */
1169
for (slot_row = 0; slot_row < NUM_MTRS; slot_row++) {
1170
int where = MTR0 + (slot_row * sizeof(u32));
1171
1172
pci_read_config_word(pvt->branch_0, where,
1173
&pvt->b0_mtr[slot_row]);
1174
1175
debugf2("MTR%d where=0x%x B0 value=0x%x\n", slot_row, where,
1176
pvt->b0_mtr[slot_row]);
1177
1178
if (pvt->maxch >= CHANNELS_PER_BRANCH) {
1179
pci_read_config_word(pvt->branch_1, where,
1180
&pvt->b1_mtr[slot_row]);
1181
debugf2("MTR%d where=0x%x B1 value=0x%x\n", slot_row,
1182
where, pvt->b1_mtr[slot_row]);
1183
} else {
1184
pvt->b1_mtr[slot_row] = 0;
1185
}
1186
}
1187
1188
/* Read and dump branch 0's MTRs */
1189
debugf2("\nMemory Technology Registers:\n");
1190
debugf2(" Branch 0:\n");
1191
for (slot_row = 0; slot_row < NUM_MTRS; slot_row++) {
1192
decode_mtr(slot_row, pvt->b0_mtr[slot_row]);
1193
}
1194
pci_read_config_word(pvt->branch_0, AMB_PRESENT_0,
1195
&pvt->b0_ambpresent0);
1196
debugf2("\t\tAMB-Branch 0-present0 0x%x:\n", pvt->b0_ambpresent0);
1197
pci_read_config_word(pvt->branch_0, AMB_PRESENT_1,
1198
&pvt->b0_ambpresent1);
1199
debugf2("\t\tAMB-Branch 0-present1 0x%x:\n", pvt->b0_ambpresent1);
1200
1201
/* Only if we have 2 branchs (4 channels) */
1202
if (pvt->maxch < CHANNELS_PER_BRANCH) {
1203
pvt->b1_ambpresent0 = 0;
1204
pvt->b1_ambpresent1 = 0;
1205
} else {
1206
/* Read and dump branch 1's MTRs */
1207
debugf2(" Branch 1:\n");
1208
for (slot_row = 0; slot_row < NUM_MTRS; slot_row++) {
1209
decode_mtr(slot_row, pvt->b1_mtr[slot_row]);
1210
}
1211
pci_read_config_word(pvt->branch_1, AMB_PRESENT_0,
1212
&pvt->b1_ambpresent0);
1213
debugf2("\t\tAMB-Branch 1-present0 0x%x:\n",
1214
pvt->b1_ambpresent0);
1215
pci_read_config_word(pvt->branch_1, AMB_PRESENT_1,
1216
&pvt->b1_ambpresent1);
1217
debugf2("\t\tAMB-Branch 1-present1 0x%x:\n",
1218
pvt->b1_ambpresent1);
1219
}
1220
1221
/* Go and determine the size of each DIMM and place in an
1222
* orderly matrix */
1223
calculate_dimm_size(pvt);
1224
}
1225
1226
/*
1227
* i5000_init_csrows Initialize the 'csrows' table within
1228
* the mci control structure with the
1229
* addressing of memory.
1230
*
1231
* return:
1232
* 0 success
1233
* 1 no actual memory found on this MC
1234
*/
1235
static int i5000_init_csrows(struct mem_ctl_info *mci)
1236
{
1237
struct i5000_pvt *pvt;
1238
struct csrow_info *p_csrow;
1239
int empty, channel_count;
1240
int max_csrows;
1241
int mtr, mtr1;
1242
int csrow_megs;
1243
int channel;
1244
int csrow;
1245
1246
pvt = mci->pvt_info;
1247
1248
channel_count = pvt->maxch;
1249
max_csrows = pvt->maxdimmperch * 2;
1250
1251
empty = 1; /* Assume NO memory */
1252
1253
for (csrow = 0; csrow < max_csrows; csrow++) {
1254
p_csrow = &mci->csrows[csrow];
1255
1256
p_csrow->csrow_idx = csrow;
1257
1258
/* use branch 0 for the basis */
1259
mtr = pvt->b0_mtr[csrow >> 1];
1260
mtr1 = pvt->b1_mtr[csrow >> 1];
1261
1262
/* if no DIMMS on this row, continue */
1263
if (!MTR_DIMMS_PRESENT(mtr) && !MTR_DIMMS_PRESENT(mtr1))
1264
continue;
1265
1266
/* FAKE OUT VALUES, FIXME */
1267
p_csrow->first_page = 0 + csrow * 20;
1268
p_csrow->last_page = 9 + csrow * 20;
1269
p_csrow->page_mask = 0xFFF;
1270
1271
p_csrow->grain = 8;
1272
1273
csrow_megs = 0;
1274
for (channel = 0; channel < pvt->maxch; channel++) {
1275
csrow_megs += pvt->dimm_info[csrow][channel].megabytes;
1276
}
1277
1278
p_csrow->nr_pages = csrow_megs << 8;
1279
1280
/* Assume DDR2 for now */
1281
p_csrow->mtype = MEM_FB_DDR2;
1282
1283
/* ask what device type on this row */
1284
if (MTR_DRAM_WIDTH(mtr))
1285
p_csrow->dtype = DEV_X8;
1286
else
1287
p_csrow->dtype = DEV_X4;
1288
1289
p_csrow->edac_mode = EDAC_S8ECD8ED;
1290
1291
empty = 0;
1292
}
1293
1294
return empty;
1295
}
1296
1297
/*
1298
* i5000_enable_error_reporting
1299
* Turn on the memory reporting features of the hardware
1300
*/
1301
static void i5000_enable_error_reporting(struct mem_ctl_info *mci)
1302
{
1303
struct i5000_pvt *pvt;
1304
u32 fbd_error_mask;
1305
1306
pvt = mci->pvt_info;
1307
1308
/* Read the FBD Error Mask Register */
1309
pci_read_config_dword(pvt->branchmap_werrors, EMASK_FBD,
1310
&fbd_error_mask);
1311
1312
/* Enable with a '0' */
1313
fbd_error_mask &= ~(ENABLE_EMASK_ALL);
1314
1315
pci_write_config_dword(pvt->branchmap_werrors, EMASK_FBD,
1316
fbd_error_mask);
1317
}
1318
1319
/*
1320
* i5000_get_dimm_and_channel_counts(pdev, &num_csrows, &num_channels)
1321
*
1322
* ask the device how many channels are present and how many CSROWS
1323
* as well
1324
*/
1325
static void i5000_get_dimm_and_channel_counts(struct pci_dev *pdev,
1326
int *num_dimms_per_channel,
1327
int *num_channels)
1328
{
1329
u8 value;
1330
1331
/* Need to retrieve just how many channels and dimms per channel are
1332
* supported on this memory controller
1333
*/
1334
pci_read_config_byte(pdev, MAXDIMMPERCH, &value);
1335
*num_dimms_per_channel = (int)value *2;
1336
1337
pci_read_config_byte(pdev, MAXCH, &value);
1338
*num_channels = (int)value;
1339
}
1340
1341
/*
1342
* i5000_probe1 Probe for ONE instance of device to see if it is
1343
* present.
1344
* return:
1345
* 0 for FOUND a device
1346
* < 0 for error code
1347
*/
1348
static int i5000_probe1(struct pci_dev *pdev, int dev_idx)
1349
{
1350
struct mem_ctl_info *mci;
1351
struct i5000_pvt *pvt;
1352
int num_channels;
1353
int num_dimms_per_channel;
1354
int num_csrows;
1355
1356
debugf0("MC: %s: %s(), pdev bus %u dev=0x%x fn=0x%x\n",
1357
__FILE__, __func__,
1358
pdev->bus->number,
1359
PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
1360
1361
/* We only are looking for func 0 of the set */
1362
if (PCI_FUNC(pdev->devfn) != 0)
1363
return -ENODEV;
1364
1365
/* Ask the devices for the number of CSROWS and CHANNELS so
1366
* that we can calculate the memory resources, etc
1367
*
1368
* The Chipset will report what it can handle which will be greater
1369
* or equal to what the motherboard manufacturer will implement.
1370
*
1371
* As we don't have a motherboard identification routine to determine
1372
* actual number of slots/dimms per channel, we thus utilize the
1373
* resource as specified by the chipset. Thus, we might have
1374
* have more DIMMs per channel than actually on the mobo, but this
1375
* allows the driver to support up to the chipset max, without
1376
* some fancy mobo determination.
1377
*/
1378
i5000_get_dimm_and_channel_counts(pdev, &num_dimms_per_channel,
1379
&num_channels);
1380
num_csrows = num_dimms_per_channel * 2;
1381
1382
debugf0("MC: %s(): Number of - Channels= %d DIMMS= %d CSROWS= %d\n",
1383
__func__, num_channels, num_dimms_per_channel, num_csrows);
1384
1385
/* allocate a new MC control structure */
1386
mci = edac_mc_alloc(sizeof(*pvt), num_csrows, num_channels, 0);
1387
1388
if (mci == NULL)
1389
return -ENOMEM;
1390
1391
kobject_get(&mci->edac_mci_kobj);
1392
debugf0("MC: %s: %s(): mci = %p\n", __FILE__, __func__, mci);
1393
1394
mci->dev = &pdev->dev; /* record ptr to the generic device */
1395
1396
pvt = mci->pvt_info;
1397
pvt->system_address = pdev; /* Record this device in our private */
1398
pvt->maxch = num_channels;
1399
pvt->maxdimmperch = num_dimms_per_channel;
1400
1401
/* 'get' the pci devices we want to reserve for our use */
1402
if (i5000_get_devices(mci, dev_idx))
1403
goto fail0;
1404
1405
/* Time to get serious */
1406
i5000_get_mc_regs(mci); /* retrieve the hardware registers */
1407
1408
mci->mc_idx = 0;
1409
mci->mtype_cap = MEM_FLAG_FB_DDR2;
1410
mci->edac_ctl_cap = EDAC_FLAG_NONE;
1411
mci->edac_cap = EDAC_FLAG_NONE;
1412
mci->mod_name = "i5000_edac.c";
1413
mci->mod_ver = I5000_REVISION;
1414
mci->ctl_name = i5000_devs[dev_idx].ctl_name;
1415
mci->dev_name = pci_name(pdev);
1416
mci->ctl_page_to_phys = NULL;
1417
1418
/* Set the function pointer to an actual operation function */
1419
mci->edac_check = i5000_check_error;
1420
1421
/* initialize the MC control structure 'csrows' table
1422
* with the mapping and control information */
1423
if (i5000_init_csrows(mci)) {
1424
debugf0("MC: Setting mci->edac_cap to EDAC_FLAG_NONE\n"
1425
" because i5000_init_csrows() returned nonzero "
1426
"value\n");
1427
mci->edac_cap = EDAC_FLAG_NONE; /* no csrows found */
1428
} else {
1429
debugf1("MC: Enable error reporting now\n");
1430
i5000_enable_error_reporting(mci);
1431
}
1432
1433
/* add this new MC control structure to EDAC's list of MCs */
1434
if (edac_mc_add_mc(mci)) {
1435
debugf0("MC: %s: %s(): failed edac_mc_add_mc()\n",
1436
__FILE__, __func__);
1437
/* FIXME: perhaps some code should go here that disables error
1438
* reporting if we just enabled it
1439
*/
1440
goto fail1;
1441
}
1442
1443
i5000_clear_error(mci);
1444
1445
/* allocating generic PCI control info */
1446
i5000_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
1447
if (!i5000_pci) {
1448
printk(KERN_WARNING
1449
"%s(): Unable to create PCI control\n",
1450
__func__);
1451
printk(KERN_WARNING
1452
"%s(): PCI error report via EDAC not setup\n",
1453
__func__);
1454
}
1455
1456
return 0;
1457
1458
/* Error exit unwinding stack */
1459
fail1:
1460
1461
i5000_put_devices(mci);
1462
1463
fail0:
1464
kobject_put(&mci->edac_mci_kobj);
1465
edac_mc_free(mci);
1466
return -ENODEV;
1467
}
1468
1469
/*
1470
* i5000_init_one constructor for one instance of device
1471
*
1472
* returns:
1473
* negative on error
1474
* count (>= 0)
1475
*/
1476
static int __devinit i5000_init_one(struct pci_dev *pdev,
1477
const struct pci_device_id *id)
1478
{
1479
int rc;
1480
1481
debugf0("MC: %s: %s()\n", __FILE__, __func__);
1482
1483
/* wake up device */
1484
rc = pci_enable_device(pdev);
1485
if (rc)
1486
return rc;
1487
1488
/* now probe and enable the device */
1489
return i5000_probe1(pdev, id->driver_data);
1490
}
1491
1492
/*
1493
* i5000_remove_one destructor for one instance of device
1494
*
1495
*/
1496
static void __devexit i5000_remove_one(struct pci_dev *pdev)
1497
{
1498
struct mem_ctl_info *mci;
1499
1500
debugf0("%s: %s()\n", __FILE__, __func__);
1501
1502
if (i5000_pci)
1503
edac_pci_release_generic_ctl(i5000_pci);
1504
1505
if ((mci = edac_mc_del_mc(&pdev->dev)) == NULL)
1506
return;
1507
1508
/* retrieve references to resources, and free those resources */
1509
i5000_put_devices(mci);
1510
kobject_put(&mci->edac_mci_kobj);
1511
edac_mc_free(mci);
1512
}
1513
1514
/*
1515
* pci_device_id table for which devices we are looking for
1516
*
1517
* The "E500P" device is the first device supported.
1518
*/
1519
static const struct pci_device_id i5000_pci_tbl[] __devinitdata = {
1520
{PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_I5000_DEV16),
1521
.driver_data = I5000P},
1522
1523
{0,} /* 0 terminated list. */
1524
};
1525
1526
MODULE_DEVICE_TABLE(pci, i5000_pci_tbl);
1527
1528
/*
1529
* i5000_driver pci_driver structure for this module
1530
*
1531
*/
1532
static struct pci_driver i5000_driver = {
1533
.name = KBUILD_BASENAME,
1534
.probe = i5000_init_one,
1535
.remove = __devexit_p(i5000_remove_one),
1536
.id_table = i5000_pci_tbl,
1537
};
1538
1539
/*
1540
* i5000_init Module entry function
1541
* Try to initialize this module for its devices
1542
*/
1543
static int __init i5000_init(void)
1544
{
1545
int pci_rc;
1546
1547
debugf2("MC: %s: %s()\n", __FILE__, __func__);
1548
1549
/* Ensure that the OPSTATE is set correctly for POLL or NMI */
1550
opstate_init();
1551
1552
pci_rc = pci_register_driver(&i5000_driver);
1553
1554
return (pci_rc < 0) ? pci_rc : 0;
1555
}
1556
1557
/*
1558
* i5000_exit() Module exit function
1559
* Unregister the driver
1560
*/
1561
static void __exit i5000_exit(void)
1562
{
1563
debugf2("MC: %s: %s()\n", __FILE__, __func__);
1564
pci_unregister_driver(&i5000_driver);
1565
}
1566
1567
module_init(i5000_init);
1568
module_exit(i5000_exit);
1569
1570
MODULE_LICENSE("GPL");
1571
MODULE_AUTHOR
1572
("Linux Networx (http://lnxi.com) Doug Thompson <[email protected]>");
1573
MODULE_DESCRIPTION("MC Driver for Intel I5000 memory controllers - "
1574
I5000_REVISION);
1575
1576
module_param(edac_op_state, int, 0444);
1577
MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
1578
module_param(misc_messages, int, 0444);
1579
MODULE_PARM_DESC(misc_messages, "Log miscellaneous non fatal messages");
1580
1581
1582