edac: Convert debugfX to edac_dbg(X,
[linux-2.6/libata-dev.git] / drivers / edac / i5400_edac.c
blob90fad3a9c3f70506213d5ca3cd41a2404083852c
1 /*
2 * Intel 5400 class Memory Controllers kernel module (Seaburg)
4 * This file may be distributed under the terms of the
5 * GNU General Public License.
7 * Copyright (c) 2008 by:
8 * Ben Woodard <woodard@redhat.com>
9 * Mauro Carvalho Chehab <mchehab@redhat.com>
11 * Red Hat Inc. http://www.redhat.com
13 * Forked and adapted from the i5000_edac driver which was
14 * written by Douglas Thompson Linux Networx <norsk5@xmission.com>
16 * This module is based on the following document:
18 * Intel 5400 Chipset Memory Controller Hub (MCH) - Datasheet
19 * http://developer.intel.com/design/chipsets/datashts/313070.htm
21 * This Memory Controller manages DDR2 FB-DIMMs. It has 2 branches, each with
22 * 2 channels operating in lockstep no-mirror mode. Each channel can have up to
23 * 4 dimm's, each with up to 8GB.
27 #include <linux/module.h>
28 #include <linux/init.h>
29 #include <linux/pci.h>
30 #include <linux/pci_ids.h>
31 #include <linux/slab.h>
32 #include <linux/edac.h>
33 #include <linux/mmzone.h>
35 #include "edac_core.h"
38 * Alter this version for the I5400 module when modifications are made
40 #define I5400_REVISION " Ver: 1.0.0"
42 #define EDAC_MOD_STR "i5400_edac"
44 #define i5400_printk(level, fmt, arg...) \
45 edac_printk(level, "i5400", fmt, ##arg)
47 #define i5400_mc_printk(mci, level, fmt, arg...) \
48 edac_mc_chipset_printk(mci, level, "i5400", fmt, ##arg)
50 /* Limits for i5400 */
51 #define MAX_BRANCHES 2
52 #define CHANNELS_PER_BRANCH 2
53 #define DIMMS_PER_CHANNEL 4
54 #define MAX_CHANNELS (MAX_BRANCHES * CHANNELS_PER_BRANCH)
56 /* Device 16,
57 * Function 0: System Address
58 * Function 1: Memory Branch Map, Control, Errors Register
59 * Function 2: FSB Error Registers
61 * All 3 functions of Device 16 (0,1,2) share the SAME DID and
62 * uses PCI_DEVICE_ID_INTEL_5400_ERR for device 16 (0,1,2),
63 * PCI_DEVICE_ID_INTEL_5400_FBD0 and PCI_DEVICE_ID_INTEL_5400_FBD1
64 * for device 21 (0,1).
67 /* OFFSETS for Function 0 */
68 #define AMBASE 0x48 /* AMB Mem Mapped Reg Region Base */
69 #define MAXCH 0x56 /* Max Channel Number */
70 #define MAXDIMMPERCH 0x57 /* Max DIMM PER Channel Number */
72 /* OFFSETS for Function 1 */
73 #define TOLM 0x6C
74 #define REDMEMB 0x7C
75 #define REC_ECC_LOCATOR_ODD(x) ((x) & 0x3fe00) /* bits [17:9] indicate ODD, [8:0] indicate EVEN */
76 #define MIR0 0x80
77 #define MIR1 0x84
78 #define AMIR0 0x8c
79 #define AMIR1 0x90
81 /* Fatal error registers */
82 #define FERR_FAT_FBD 0x98 /* also called as FERR_FAT_FB_DIMM at datasheet */
83 #define FERR_FAT_FBDCHAN (3<<28) /* channel index where the highest-order error occurred */
85 #define NERR_FAT_FBD 0x9c
86 #define FERR_NF_FBD 0xa0 /* also called as FERR_NFAT_FB_DIMM at datasheet */
88 /* Non-fatal error register */
89 #define NERR_NF_FBD 0xa4
91 /* Enable error mask */
92 #define EMASK_FBD 0xa8
94 #define ERR0_FBD 0xac
95 #define ERR1_FBD 0xb0
96 #define ERR2_FBD 0xb4
97 #define MCERR_FBD 0xb8
99 /* No OFFSETS for Device 16 Function 2 */
102 * Device 21,
103 * Function 0: Memory Map Branch 0
105 * Device 22,
106 * Function 0: Memory Map Branch 1
109 /* OFFSETS for Function 0 */
110 #define AMBPRESENT_0 0x64
111 #define AMBPRESENT_1 0x66
112 #define MTR0 0x80
113 #define MTR1 0x82
114 #define MTR2 0x84
115 #define MTR3 0x86
117 /* OFFSETS for Function 1 */
118 #define NRECFGLOG 0x74
119 #define RECFGLOG 0x78
120 #define NRECMEMA 0xbe
121 #define NRECMEMB 0xc0
122 #define NRECFB_DIMMA 0xc4
123 #define NRECFB_DIMMB 0xc8
124 #define NRECFB_DIMMC 0xcc
125 #define NRECFB_DIMMD 0xd0
126 #define NRECFB_DIMME 0xd4
127 #define NRECFB_DIMMF 0xd8
128 #define REDMEMA 0xdC
129 #define RECMEMA 0xf0
130 #define RECMEMB 0xf4
131 #define RECFB_DIMMA 0xf8
132 #define RECFB_DIMMB 0xec
133 #define RECFB_DIMMC 0xf0
134 #define RECFB_DIMMD 0xf4
135 #define RECFB_DIMME 0xf8
136 #define RECFB_DIMMF 0xfC
139 * Error indicator bits and masks
140 * Error masks are according with Table 5-17 of i5400 datasheet
143 enum error_mask {
144 EMASK_M1 = 1<<0, /* Memory Write error on non-redundant retry */
145 EMASK_M2 = 1<<1, /* Memory or FB-DIMM configuration CRC read error */
146 EMASK_M3 = 1<<2, /* Reserved */
147 EMASK_M4 = 1<<3, /* Uncorrectable Data ECC on Replay */
148 EMASK_M5 = 1<<4, /* Aliased Uncorrectable Non-Mirrored Demand Data ECC */
149 EMASK_M6 = 1<<5, /* Unsupported on i5400 */
150 EMASK_M7 = 1<<6, /* Aliased Uncorrectable Resilver- or Spare-Copy Data ECC */
151 EMASK_M8 = 1<<7, /* Aliased Uncorrectable Patrol Data ECC */
152 EMASK_M9 = 1<<8, /* Non-Aliased Uncorrectable Non-Mirrored Demand Data ECC */
153 EMASK_M10 = 1<<9, /* Unsupported on i5400 */
154 EMASK_M11 = 1<<10, /* Non-Aliased Uncorrectable Resilver- or Spare-Copy Data ECC */
155 EMASK_M12 = 1<<11, /* Non-Aliased Uncorrectable Patrol Data ECC */
156 EMASK_M13 = 1<<12, /* Memory Write error on first attempt */
157 EMASK_M14 = 1<<13, /* FB-DIMM Configuration Write error on first attempt */
158 EMASK_M15 = 1<<14, /* Memory or FB-DIMM configuration CRC read error */
159 EMASK_M16 = 1<<15, /* Channel Failed-Over Occurred */
160 EMASK_M17 = 1<<16, /* Correctable Non-Mirrored Demand Data ECC */
161 EMASK_M18 = 1<<17, /* Unsupported on i5400 */
162 EMASK_M19 = 1<<18, /* Correctable Resilver- or Spare-Copy Data ECC */
163 EMASK_M20 = 1<<19, /* Correctable Patrol Data ECC */
164 EMASK_M21 = 1<<20, /* FB-DIMM Northbound parity error on FB-DIMM Sync Status */
165 EMASK_M22 = 1<<21, /* SPD protocol Error */
166 EMASK_M23 = 1<<22, /* Non-Redundant Fast Reset Timeout */
167 EMASK_M24 = 1<<23, /* Refresh error */
168 EMASK_M25 = 1<<24, /* Memory Write error on redundant retry */
169 EMASK_M26 = 1<<25, /* Redundant Fast Reset Timeout */
170 EMASK_M27 = 1<<26, /* Correctable Counter Threshold Exceeded */
171 EMASK_M28 = 1<<27, /* DIMM-Spare Copy Completed */
172 EMASK_M29 = 1<<28, /* DIMM-Isolation Completed */
176 * Names to translate bit error into something useful
178 static const char *error_name[] = {
179 [0] = "Memory Write error on non-redundant retry",
180 [1] = "Memory or FB-DIMM configuration CRC read error",
181 /* Reserved */
182 [3] = "Uncorrectable Data ECC on Replay",
183 [4] = "Aliased Uncorrectable Non-Mirrored Demand Data ECC",
184 /* M6 Unsupported on i5400 */
185 [6] = "Aliased Uncorrectable Resilver- or Spare-Copy Data ECC",
186 [7] = "Aliased Uncorrectable Patrol Data ECC",
187 [8] = "Non-Aliased Uncorrectable Non-Mirrored Demand Data ECC",
188 /* M10 Unsupported on i5400 */
189 [10] = "Non-Aliased Uncorrectable Resilver- or Spare-Copy Data ECC",
190 [11] = "Non-Aliased Uncorrectable Patrol Data ECC",
191 [12] = "Memory Write error on first attempt",
192 [13] = "FB-DIMM Configuration Write error on first attempt",
193 [14] = "Memory or FB-DIMM configuration CRC read error",
194 [15] = "Channel Failed-Over Occurred",
195 [16] = "Correctable Non-Mirrored Demand Data ECC",
196 /* M18 Unsupported on i5400 */
197 [18] = "Correctable Resilver- or Spare-Copy Data ECC",
198 [19] = "Correctable Patrol Data ECC",
199 [20] = "FB-DIMM Northbound parity error on FB-DIMM Sync Status",
200 [21] = "SPD protocol Error",
201 [22] = "Non-Redundant Fast Reset Timeout",
202 [23] = "Refresh error",
203 [24] = "Memory Write error on redundant retry",
204 [25] = "Redundant Fast Reset Timeout",
205 [26] = "Correctable Counter Threshold Exceeded",
206 [27] = "DIMM-Spare Copy Completed",
207 [28] = "DIMM-Isolation Completed",
210 /* Fatal errors */
211 #define ERROR_FAT_MASK (EMASK_M1 | \
212 EMASK_M2 | \
213 EMASK_M23)
215 /* Correctable errors */
216 #define ERROR_NF_CORRECTABLE (EMASK_M27 | \
217 EMASK_M20 | \
218 EMASK_M19 | \
219 EMASK_M18 | \
220 EMASK_M17 | \
221 EMASK_M16)
222 #define ERROR_NF_DIMM_SPARE (EMASK_M29 | \
223 EMASK_M28)
224 #define ERROR_NF_SPD_PROTOCOL (EMASK_M22)
225 #define ERROR_NF_NORTH_CRC (EMASK_M21)
227 /* Recoverable errors */
228 #define ERROR_NF_RECOVERABLE (EMASK_M26 | \
229 EMASK_M25 | \
230 EMASK_M24 | \
231 EMASK_M15 | \
232 EMASK_M14 | \
233 EMASK_M13 | \
234 EMASK_M12 | \
235 EMASK_M11 | \
236 EMASK_M9 | \
237 EMASK_M8 | \
238 EMASK_M7 | \
239 EMASK_M5)
241 /* uncorrectable errors */
242 #define ERROR_NF_UNCORRECTABLE (EMASK_M4)
244 /* mask to all non-fatal errors */
245 #define ERROR_NF_MASK (ERROR_NF_CORRECTABLE | \
246 ERROR_NF_UNCORRECTABLE | \
247 ERROR_NF_RECOVERABLE | \
248 ERROR_NF_DIMM_SPARE | \
249 ERROR_NF_SPD_PROTOCOL | \
250 ERROR_NF_NORTH_CRC)
253 * Define error masks for the several registers
256 /* Enable all fatal and non fatal errors */
257 #define ENABLE_EMASK_ALL (ERROR_FAT_MASK | ERROR_NF_MASK)
259 /* mask for fatal error registers */
260 #define FERR_FAT_MASK ERROR_FAT_MASK
262 /* masks for non-fatal error register */
263 static inline int to_nf_mask(unsigned int mask)
265 return (mask & EMASK_M29) | (mask >> 3);
268 static inline int from_nf_ferr(unsigned int mask)
270 return (mask & EMASK_M29) | /* Bit 28 */
271 (mask & ((1 << 28) - 1) << 3); /* Bits 0 to 27 */
274 #define FERR_NF_MASK to_nf_mask(ERROR_NF_MASK)
275 #define FERR_NF_CORRECTABLE to_nf_mask(ERROR_NF_CORRECTABLE)
276 #define FERR_NF_DIMM_SPARE to_nf_mask(ERROR_NF_DIMM_SPARE)
277 #define FERR_NF_SPD_PROTOCOL to_nf_mask(ERROR_NF_SPD_PROTOCOL)
278 #define FERR_NF_NORTH_CRC to_nf_mask(ERROR_NF_NORTH_CRC)
279 #define FERR_NF_RECOVERABLE to_nf_mask(ERROR_NF_RECOVERABLE)
280 #define FERR_NF_UNCORRECTABLE to_nf_mask(ERROR_NF_UNCORRECTABLE)
282 /* Defines to extract the vaious fields from the
283 * MTRx - Memory Technology Registers
285 #define MTR_DIMMS_PRESENT(mtr) ((mtr) & (1 << 10))
286 #define MTR_DIMMS_ETHROTTLE(mtr) ((mtr) & (1 << 9))
287 #define MTR_DRAM_WIDTH(mtr) (((mtr) & (1 << 8)) ? 8 : 4)
288 #define MTR_DRAM_BANKS(mtr) (((mtr) & (1 << 6)) ? 8 : 4)
289 #define MTR_DRAM_BANKS_ADDR_BITS(mtr) ((MTR_DRAM_BANKS(mtr) == 8) ? 3 : 2)
290 #define MTR_DIMM_RANK(mtr) (((mtr) >> 5) & 0x1)
291 #define MTR_DIMM_RANK_ADDR_BITS(mtr) (MTR_DIMM_RANK(mtr) ? 2 : 1)
292 #define MTR_DIMM_ROWS(mtr) (((mtr) >> 2) & 0x3)
293 #define MTR_DIMM_ROWS_ADDR_BITS(mtr) (MTR_DIMM_ROWS(mtr) + 13)
294 #define MTR_DIMM_COLS(mtr) ((mtr) & 0x3)
295 #define MTR_DIMM_COLS_ADDR_BITS(mtr) (MTR_DIMM_COLS(mtr) + 10)
297 /* This applies to FERR_NF_FB-DIMM as well as FERR_FAT_FB-DIMM */
298 static inline int extract_fbdchan_indx(u32 x)
300 return (x>>28) & 0x3;
303 /* Device name and register DID (Device ID) */
304 struct i5400_dev_info {
305 const char *ctl_name; /* name for this device */
306 u16 fsb_mapping_errors; /* DID for the branchmap,control */
309 /* Table of devices attributes supported by this driver */
310 static const struct i5400_dev_info i5400_devs[] = {
312 .ctl_name = "I5400",
313 .fsb_mapping_errors = PCI_DEVICE_ID_INTEL_5400_ERR,
317 struct i5400_dimm_info {
318 int megabytes; /* size, 0 means not present */
321 /* driver private data structure */
322 struct i5400_pvt {
323 struct pci_dev *system_address; /* 16.0 */
324 struct pci_dev *branchmap_werrors; /* 16.1 */
325 struct pci_dev *fsb_error_regs; /* 16.2 */
326 struct pci_dev *branch_0; /* 21.0 */
327 struct pci_dev *branch_1; /* 22.0 */
329 u16 tolm; /* top of low memory */
330 u64 ambase; /* AMB BAR */
332 u16 mir0, mir1;
334 u16 b0_mtr[DIMMS_PER_CHANNEL]; /* Memory Technlogy Reg */
335 u16 b0_ambpresent0; /* Branch 0, Channel 0 */
336 u16 b0_ambpresent1; /* Brnach 0, Channel 1 */
338 u16 b1_mtr[DIMMS_PER_CHANNEL]; /* Memory Technlogy Reg */
339 u16 b1_ambpresent0; /* Branch 1, Channel 8 */
340 u16 b1_ambpresent1; /* Branch 1, Channel 1 */
342 /* DIMM information matrix, allocating architecture maximums */
343 struct i5400_dimm_info dimm_info[DIMMS_PER_CHANNEL][MAX_CHANNELS];
345 /* Actual values for this controller */
346 int maxch; /* Max channels */
347 int maxdimmperch; /* Max DIMMs per channel */
350 /* I5400 MCH error information retrieved from Hardware */
351 struct i5400_error_info {
352 /* These registers are always read from the MC */
353 u32 ferr_fat_fbd; /* First Errors Fatal */
354 u32 nerr_fat_fbd; /* Next Errors Fatal */
355 u32 ferr_nf_fbd; /* First Errors Non-Fatal */
356 u32 nerr_nf_fbd; /* Next Errors Non-Fatal */
358 /* These registers are input ONLY if there was a Recoverable Error */
359 u32 redmemb; /* Recoverable Mem Data Error log B */
360 u16 recmema; /* Recoverable Mem Error log A */
361 u32 recmemb; /* Recoverable Mem Error log B */
363 /* These registers are input ONLY if there was a Non-Rec Error */
364 u16 nrecmema; /* Non-Recoverable Mem log A */
365 u16 nrecmemb; /* Non-Recoverable Mem log B */
369 /* note that nrec_rdwr changed from NRECMEMA to NRECMEMB between the 5000 and
370 5400 better to use an inline function than a macro in this case */
371 static inline int nrec_bank(struct i5400_error_info *info)
373 return ((info->nrecmema) >> 12) & 0x7;
375 static inline int nrec_rank(struct i5400_error_info *info)
377 return ((info->nrecmema) >> 8) & 0xf;
379 static inline int nrec_buf_id(struct i5400_error_info *info)
381 return ((info->nrecmema)) & 0xff;
383 static inline int nrec_rdwr(struct i5400_error_info *info)
385 return (info->nrecmemb) >> 31;
387 /* This applies to both NREC and REC string so it can be used with nrec_rdwr
388 and rec_rdwr */
389 static inline const char *rdwr_str(int rdwr)
391 return rdwr ? "Write" : "Read";
393 static inline int nrec_cas(struct i5400_error_info *info)
395 return ((info->nrecmemb) >> 16) & 0x1fff;
397 static inline int nrec_ras(struct i5400_error_info *info)
399 return (info->nrecmemb) & 0xffff;
401 static inline int rec_bank(struct i5400_error_info *info)
403 return ((info->recmema) >> 12) & 0x7;
405 static inline int rec_rank(struct i5400_error_info *info)
407 return ((info->recmema) >> 8) & 0xf;
409 static inline int rec_rdwr(struct i5400_error_info *info)
411 return (info->recmemb) >> 31;
413 static inline int rec_cas(struct i5400_error_info *info)
415 return ((info->recmemb) >> 16) & 0x1fff;
417 static inline int rec_ras(struct i5400_error_info *info)
419 return (info->recmemb) & 0xffff;
422 static struct edac_pci_ctl_info *i5400_pci;
425 * i5400_get_error_info Retrieve the hardware error information from
426 * the hardware and cache it in the 'info'
427 * structure
429 static void i5400_get_error_info(struct mem_ctl_info *mci,
430 struct i5400_error_info *info)
432 struct i5400_pvt *pvt;
433 u32 value;
435 pvt = mci->pvt_info;
437 /* read in the 1st FATAL error register */
438 pci_read_config_dword(pvt->branchmap_werrors, FERR_FAT_FBD, &value);
440 /* Mask only the bits that the doc says are valid
442 value &= (FERR_FAT_FBDCHAN | FERR_FAT_MASK);
444 /* If there is an error, then read in the
445 NEXT FATAL error register and the Memory Error Log Register A
447 if (value & FERR_FAT_MASK) {
448 info->ferr_fat_fbd = value;
450 /* harvest the various error data we need */
451 pci_read_config_dword(pvt->branchmap_werrors,
452 NERR_FAT_FBD, &info->nerr_fat_fbd);
453 pci_read_config_word(pvt->branchmap_werrors,
454 NRECMEMA, &info->nrecmema);
455 pci_read_config_word(pvt->branchmap_werrors,
456 NRECMEMB, &info->nrecmemb);
458 /* Clear the error bits, by writing them back */
459 pci_write_config_dword(pvt->branchmap_werrors,
460 FERR_FAT_FBD, value);
461 } else {
462 info->ferr_fat_fbd = 0;
463 info->nerr_fat_fbd = 0;
464 info->nrecmema = 0;
465 info->nrecmemb = 0;
468 /* read in the 1st NON-FATAL error register */
469 pci_read_config_dword(pvt->branchmap_werrors, FERR_NF_FBD, &value);
471 /* If there is an error, then read in the 1st NON-FATAL error
472 * register as well */
473 if (value & FERR_NF_MASK) {
474 info->ferr_nf_fbd = value;
476 /* harvest the various error data we need */
477 pci_read_config_dword(pvt->branchmap_werrors,
478 NERR_NF_FBD, &info->nerr_nf_fbd);
479 pci_read_config_word(pvt->branchmap_werrors,
480 RECMEMA, &info->recmema);
481 pci_read_config_dword(pvt->branchmap_werrors,
482 RECMEMB, &info->recmemb);
483 pci_read_config_dword(pvt->branchmap_werrors,
484 REDMEMB, &info->redmemb);
486 /* Clear the error bits, by writing them back */
487 pci_write_config_dword(pvt->branchmap_werrors,
488 FERR_NF_FBD, value);
489 } else {
490 info->ferr_nf_fbd = 0;
491 info->nerr_nf_fbd = 0;
492 info->recmema = 0;
493 info->recmemb = 0;
494 info->redmemb = 0;
499 * i5400_proccess_non_recoverable_info(struct mem_ctl_info *mci,
500 * struct i5400_error_info *info,
501 * int handle_errors);
503 * handle the Intel FATAL and unrecoverable errors, if any
505 static void i5400_proccess_non_recoverable_info(struct mem_ctl_info *mci,
506 struct i5400_error_info *info,
507 unsigned long allErrors)
509 char msg[EDAC_MC_LABEL_LEN + 1 + 90 + 80];
510 int branch;
511 int channel;
512 int bank;
513 int buf_id;
514 int rank;
515 int rdwr;
516 int ras, cas;
517 int errnum;
518 char *type = NULL;
519 enum hw_event_mc_err_type tp_event = HW_EVENT_ERR_UNCORRECTED;
521 if (!allErrors)
522 return; /* if no error, return now */
524 if (allErrors & ERROR_FAT_MASK) {
525 type = "FATAL";
526 tp_event = HW_EVENT_ERR_FATAL;
527 } else if (allErrors & FERR_NF_UNCORRECTABLE)
528 type = "NON-FATAL uncorrected";
529 else
530 type = "NON-FATAL recoverable";
532 /* ONLY ONE of the possible error bits will be set, as per the docs */
534 branch = extract_fbdchan_indx(info->ferr_fat_fbd);
535 channel = branch;
537 /* Use the NON-Recoverable macros to extract data */
538 bank = nrec_bank(info);
539 rank = nrec_rank(info);
540 buf_id = nrec_buf_id(info);
541 rdwr = nrec_rdwr(info);
542 ras = nrec_ras(info);
543 cas = nrec_cas(info);
545 edac_dbg(0, "\t\tDIMM= %d Channels= %d,%d (Branch= %d DRAM Bank= %d Buffer ID = %d rdwr= %s ras= %d cas= %d)\n",
546 rank, channel, channel + 1, branch >> 1, bank,
547 buf_id, rdwr_str(rdwr), ras, cas);
549 /* Only 1 bit will be on */
550 errnum = find_first_bit(&allErrors, ARRAY_SIZE(error_name));
552 /* Form out message */
553 snprintf(msg, sizeof(msg),
554 "Bank=%d Buffer ID = %d RAS=%d CAS=%d Err=0x%lx (%s)",
555 bank, buf_id, ras, cas, allErrors, error_name[errnum]);
557 edac_mc_handle_error(tp_event, mci, 0, 0, 0,
558 branch >> 1, -1, rank,
559 rdwr ? "Write error" : "Read error",
560 msg, NULL);
564 * i5400_process_fatal_error_info(struct mem_ctl_info *mci,
565 * struct i5400_error_info *info,
566 * int handle_errors);
568 * handle the Intel NON-FATAL errors, if any
570 static void i5400_process_nonfatal_error_info(struct mem_ctl_info *mci,
571 struct i5400_error_info *info)
573 char msg[EDAC_MC_LABEL_LEN + 1 + 90 + 80];
574 unsigned long allErrors;
575 int branch;
576 int channel;
577 int bank;
578 int rank;
579 int rdwr;
580 int ras, cas;
581 int errnum;
583 /* mask off the Error bits that are possible */
584 allErrors = from_nf_ferr(info->ferr_nf_fbd & FERR_NF_MASK);
585 if (!allErrors)
586 return; /* if no error, return now */
588 /* ONLY ONE of the possible error bits will be set, as per the docs */
590 if (allErrors & (ERROR_NF_UNCORRECTABLE | ERROR_NF_RECOVERABLE)) {
591 i5400_proccess_non_recoverable_info(mci, info, allErrors);
592 return;
595 /* Correctable errors */
596 if (allErrors & ERROR_NF_CORRECTABLE) {
597 edac_dbg(0, "\tCorrected bits= 0x%lx\n", allErrors);
599 branch = extract_fbdchan_indx(info->ferr_nf_fbd);
601 channel = 0;
602 if (REC_ECC_LOCATOR_ODD(info->redmemb))
603 channel = 1;
605 /* Convert channel to be based from zero, instead of
606 * from branch base of 0 */
607 channel += branch;
609 bank = rec_bank(info);
610 rank = rec_rank(info);
611 rdwr = rec_rdwr(info);
612 ras = rec_ras(info);
613 cas = rec_cas(info);
615 /* Only 1 bit will be on */
616 errnum = find_first_bit(&allErrors, ARRAY_SIZE(error_name));
618 edac_dbg(0, "\t\tDIMM= %d Channel= %d (Branch %d DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
619 rank, channel, branch >> 1, bank,
620 rdwr_str(rdwr), ras, cas);
622 /* Form out message */
623 snprintf(msg, sizeof(msg),
624 "Corrected error (Branch=%d DRAM-Bank=%d RDWR=%s "
625 "RAS=%d CAS=%d, CE Err=0x%lx (%s))",
626 branch >> 1, bank, rdwr_str(rdwr), ras, cas,
627 allErrors, error_name[errnum]);
629 edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 0, 0, 0,
630 branch >> 1, channel % 2, rank,
631 rdwr ? "Write error" : "Read error",
632 msg, NULL);
634 return;
637 /* Miscellaneous errors */
638 errnum = find_first_bit(&allErrors, ARRAY_SIZE(error_name));
640 branch = extract_fbdchan_indx(info->ferr_nf_fbd);
642 i5400_mc_printk(mci, KERN_EMERG,
643 "Non-Fatal misc error (Branch=%d Err=%#lx (%s))",
644 branch >> 1, allErrors, error_name[errnum]);
648 * i5400_process_error_info Process the error info that is
649 * in the 'info' structure, previously retrieved from hardware
651 static void i5400_process_error_info(struct mem_ctl_info *mci,
652 struct i5400_error_info *info)
653 { u32 allErrors;
655 /* First handle any fatal errors that occurred */
656 allErrors = (info->ferr_fat_fbd & FERR_FAT_MASK);
657 i5400_proccess_non_recoverable_info(mci, info, allErrors);
659 /* now handle any non-fatal errors that occurred */
660 i5400_process_nonfatal_error_info(mci, info);
664 * i5400_clear_error Retrieve any error from the hardware
665 * but do NOT process that error.
666 * Used for 'clearing' out of previous errors
667 * Called by the Core module.
669 static void i5400_clear_error(struct mem_ctl_info *mci)
671 struct i5400_error_info info;
673 i5400_get_error_info(mci, &info);
677 * i5400_check_error Retrieve and process errors reported by the
678 * hardware. Called by the Core module.
680 static void i5400_check_error(struct mem_ctl_info *mci)
682 struct i5400_error_info info;
683 edac_dbg(4, "MC%d\n", mci->mc_idx);
684 i5400_get_error_info(mci, &info);
685 i5400_process_error_info(mci, &info);
689 * i5400_put_devices 'put' all the devices that we have
690 * reserved via 'get'
692 static void i5400_put_devices(struct mem_ctl_info *mci)
694 struct i5400_pvt *pvt;
696 pvt = mci->pvt_info;
698 /* Decrement usage count for devices */
699 pci_dev_put(pvt->branch_1);
700 pci_dev_put(pvt->branch_0);
701 pci_dev_put(pvt->fsb_error_regs);
702 pci_dev_put(pvt->branchmap_werrors);
706 * i5400_get_devices Find and perform 'get' operation on the MCH's
707 * device/functions we want to reference for this driver
709 * Need to 'get' device 16 func 1 and func 2
711 static int i5400_get_devices(struct mem_ctl_info *mci, int dev_idx)
713 struct i5400_pvt *pvt;
714 struct pci_dev *pdev;
716 pvt = mci->pvt_info;
717 pvt->branchmap_werrors = NULL;
718 pvt->fsb_error_regs = NULL;
719 pvt->branch_0 = NULL;
720 pvt->branch_1 = NULL;
722 /* Attempt to 'get' the MCH register we want */
723 pdev = NULL;
724 while (1) {
725 pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
726 PCI_DEVICE_ID_INTEL_5400_ERR, pdev);
727 if (!pdev) {
728 /* End of list, leave */
729 i5400_printk(KERN_ERR,
730 "'system address,Process Bus' "
731 "device not found:"
732 "vendor 0x%x device 0x%x ERR func 1 "
733 "(broken BIOS?)\n",
734 PCI_VENDOR_ID_INTEL,
735 PCI_DEVICE_ID_INTEL_5400_ERR);
736 return -ENODEV;
739 /* Store device 16 func 1 */
740 if (PCI_FUNC(pdev->devfn) == 1)
741 break;
743 pvt->branchmap_werrors = pdev;
745 pdev = NULL;
746 while (1) {
747 pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
748 PCI_DEVICE_ID_INTEL_5400_ERR, pdev);
749 if (!pdev) {
750 /* End of list, leave */
751 i5400_printk(KERN_ERR,
752 "'system address,Process Bus' "
753 "device not found:"
754 "vendor 0x%x device 0x%x ERR func 2 "
755 "(broken BIOS?)\n",
756 PCI_VENDOR_ID_INTEL,
757 PCI_DEVICE_ID_INTEL_5400_ERR);
759 pci_dev_put(pvt->branchmap_werrors);
760 return -ENODEV;
763 /* Store device 16 func 2 */
764 if (PCI_FUNC(pdev->devfn) == 2)
765 break;
767 pvt->fsb_error_regs = pdev;
769 edac_dbg(1, "System Address, processor bus- PCI Bus ID: %s %x:%x\n",
770 pci_name(pvt->system_address),
771 pvt->system_address->vendor, pvt->system_address->device);
772 edac_dbg(1, "Branchmap, control and errors - PCI Bus ID: %s %x:%x\n",
773 pci_name(pvt->branchmap_werrors),
774 pvt->branchmap_werrors->vendor,
775 pvt->branchmap_werrors->device);
776 edac_dbg(1, "FSB Error Regs - PCI Bus ID: %s %x:%x\n",
777 pci_name(pvt->fsb_error_regs),
778 pvt->fsb_error_regs->vendor, pvt->fsb_error_regs->device);
780 pvt->branch_0 = pci_get_device(PCI_VENDOR_ID_INTEL,
781 PCI_DEVICE_ID_INTEL_5400_FBD0, NULL);
782 if (!pvt->branch_0) {
783 i5400_printk(KERN_ERR,
784 "MC: 'BRANCH 0' device not found:"
785 "vendor 0x%x device 0x%x Func 0 (broken BIOS?)\n",
786 PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_5400_FBD0);
788 pci_dev_put(pvt->fsb_error_regs);
789 pci_dev_put(pvt->branchmap_werrors);
790 return -ENODEV;
793 /* If this device claims to have more than 2 channels then
794 * fetch Branch 1's information
796 if (pvt->maxch < CHANNELS_PER_BRANCH)
797 return 0;
799 pvt->branch_1 = pci_get_device(PCI_VENDOR_ID_INTEL,
800 PCI_DEVICE_ID_INTEL_5400_FBD1, NULL);
801 if (!pvt->branch_1) {
802 i5400_printk(KERN_ERR,
803 "MC: 'BRANCH 1' device not found:"
804 "vendor 0x%x device 0x%x Func 0 "
805 "(broken BIOS?)\n",
806 PCI_VENDOR_ID_INTEL,
807 PCI_DEVICE_ID_INTEL_5400_FBD1);
809 pci_dev_put(pvt->branch_0);
810 pci_dev_put(pvt->fsb_error_regs);
811 pci_dev_put(pvt->branchmap_werrors);
812 return -ENODEV;
815 return 0;
819 * determine_amb_present
821 * the information is contained in DIMMS_PER_CHANNEL different
822 * registers determining which of the DIMMS_PER_CHANNEL requires
823 * knowing which channel is in question
825 * 2 branches, each with 2 channels
826 * b0_ambpresent0 for channel '0'
827 * b0_ambpresent1 for channel '1'
828 * b1_ambpresent0 for channel '2'
829 * b1_ambpresent1 for channel '3'
831 static int determine_amb_present_reg(struct i5400_pvt *pvt, int channel)
833 int amb_present;
835 if (channel < CHANNELS_PER_BRANCH) {
836 if (channel & 0x1)
837 amb_present = pvt->b0_ambpresent1;
838 else
839 amb_present = pvt->b0_ambpresent0;
840 } else {
841 if (channel & 0x1)
842 amb_present = pvt->b1_ambpresent1;
843 else
844 amb_present = pvt->b1_ambpresent0;
847 return amb_present;
851 * determine_mtr(pvt, dimm, channel)
853 * return the proper MTR register as determine by the dimm and desired channel
855 static int determine_mtr(struct i5400_pvt *pvt, int dimm, int channel)
857 int mtr;
858 int n;
860 /* There is one MTR for each slot pair of FB-DIMMs,
861 Each slot pair may be at branch 0 or branch 1.
863 n = dimm;
865 if (n >= DIMMS_PER_CHANNEL) {
866 edac_dbg(0, "ERROR: trying to access an invalid dimm: %d\n",
867 dimm);
868 return 0;
871 if (channel < CHANNELS_PER_BRANCH)
872 mtr = pvt->b0_mtr[n];
873 else
874 mtr = pvt->b1_mtr[n];
876 return mtr;
881 static void decode_mtr(int slot_row, u16 mtr)
883 int ans;
885 ans = MTR_DIMMS_PRESENT(mtr);
887 edac_dbg(2, "\tMTR%d=0x%x: DIMMs are %sPresent\n",
888 slot_row, mtr, ans ? "" : "NOT ");
889 if (!ans)
890 return;
892 edac_dbg(2, "\t\tWIDTH: x%d\n", MTR_DRAM_WIDTH(mtr));
894 edac_dbg(2, "\t\tELECTRICAL THROTTLING is %s\n",
895 MTR_DIMMS_ETHROTTLE(mtr) ? "enabled" : "disabled");
897 edac_dbg(2, "\t\tNUMBANK: %d bank(s)\n", MTR_DRAM_BANKS(mtr));
898 edac_dbg(2, "\t\tNUMRANK: %s\n",
899 MTR_DIMM_RANK(mtr) ? "double" : "single");
900 edac_dbg(2, "\t\tNUMROW: %s\n",
901 MTR_DIMM_ROWS(mtr) == 0 ? "8,192 - 13 rows" :
902 MTR_DIMM_ROWS(mtr) == 1 ? "16,384 - 14 rows" :
903 MTR_DIMM_ROWS(mtr) == 2 ? "32,768 - 15 rows" :
904 "65,536 - 16 rows");
905 edac_dbg(2, "\t\tNUMCOL: %s\n",
906 MTR_DIMM_COLS(mtr) == 0 ? "1,024 - 10 columns" :
907 MTR_DIMM_COLS(mtr) == 1 ? "2,048 - 11 columns" :
908 MTR_DIMM_COLS(mtr) == 2 ? "4,096 - 12 columns" :
909 "reserved");
912 static void handle_channel(struct i5400_pvt *pvt, int dimm, int channel,
913 struct i5400_dimm_info *dinfo)
915 int mtr;
916 int amb_present_reg;
917 int addrBits;
919 mtr = determine_mtr(pvt, dimm, channel);
920 if (MTR_DIMMS_PRESENT(mtr)) {
921 amb_present_reg = determine_amb_present_reg(pvt, channel);
923 /* Determine if there is a DIMM present in this DIMM slot */
924 if (amb_present_reg & (1 << dimm)) {
925 /* Start with the number of bits for a Bank
926 * on the DRAM */
927 addrBits = MTR_DRAM_BANKS_ADDR_BITS(mtr);
928 /* Add thenumber of ROW bits */
929 addrBits += MTR_DIMM_ROWS_ADDR_BITS(mtr);
930 /* add the number of COLUMN bits */
931 addrBits += MTR_DIMM_COLS_ADDR_BITS(mtr);
932 /* add the number of RANK bits */
933 addrBits += MTR_DIMM_RANK(mtr);
935 addrBits += 6; /* add 64 bits per DIMM */
936 addrBits -= 20; /* divide by 2^^20 */
937 addrBits -= 3; /* 8 bits per bytes */
939 dinfo->megabytes = 1 << addrBits;
945 * calculate_dimm_size
947 * also will output a DIMM matrix map, if debug is enabled, for viewing
948 * how the DIMMs are populated
950 static void calculate_dimm_size(struct i5400_pvt *pvt)
952 struct i5400_dimm_info *dinfo;
953 int dimm, max_dimms;
954 char *p, *mem_buffer;
955 int space, n;
956 int channel, branch;
958 /* ================= Generate some debug output ================= */
959 space = PAGE_SIZE;
960 mem_buffer = p = kmalloc(space, GFP_KERNEL);
961 if (p == NULL) {
962 i5400_printk(KERN_ERR, "MC: %s:%s() kmalloc() failed\n",
963 __FILE__, __func__);
964 return;
967 /* Scan all the actual DIMMS
968 * and calculate the information for each DIMM
969 * Start with the highest dimm first, to display it first
970 * and work toward the 0th dimm
972 max_dimms = pvt->maxdimmperch;
973 for (dimm = max_dimms - 1; dimm >= 0; dimm--) {
975 /* on an odd dimm, first output a 'boundary' marker,
976 * then reset the message buffer */
977 if (dimm & 0x1) {
978 n = snprintf(p, space, "---------------------------"
979 "-------------------------------");
980 p += n;
981 space -= n;
982 edac_dbg(2, "%s\n", mem_buffer);
983 p = mem_buffer;
984 space = PAGE_SIZE;
986 n = snprintf(p, space, "dimm %2d ", dimm);
987 p += n;
988 space -= n;
990 for (channel = 0; channel < pvt->maxch; channel++) {
991 dinfo = &pvt->dimm_info[dimm][channel];
992 handle_channel(pvt, dimm, channel, dinfo);
993 n = snprintf(p, space, "%4d MB | ", dinfo->megabytes);
994 p += n;
995 space -= n;
997 edac_dbg(2, "%s\n", mem_buffer);
998 p = mem_buffer;
999 space = PAGE_SIZE;
1002 /* Output the last bottom 'boundary' marker */
1003 n = snprintf(p, space, "---------------------------"
1004 "-------------------------------");
1005 p += n;
1006 space -= n;
1007 edac_dbg(2, "%s\n", mem_buffer);
1008 p = mem_buffer;
1009 space = PAGE_SIZE;
1011 /* now output the 'channel' labels */
1012 n = snprintf(p, space, " ");
1013 p += n;
1014 space -= n;
1015 for (channel = 0; channel < pvt->maxch; channel++) {
1016 n = snprintf(p, space, "channel %d | ", channel);
1017 p += n;
1018 space -= n;
1021 space -= n;
1022 edac_dbg(2, "%s\n", mem_buffer);
1023 p = mem_buffer;
1024 space = PAGE_SIZE;
1026 n = snprintf(p, space, " ");
1027 p += n;
1028 for (branch = 0; branch < MAX_BRANCHES; branch++) {
1029 n = snprintf(p, space, " branch %d | ", branch);
1030 p += n;
1031 space -= n;
1034 /* output the last message and free buffer */
1035 edac_dbg(2, "%s\n", mem_buffer);
1036 kfree(mem_buffer);
1040 * i5400_get_mc_regs read in the necessary registers and
1041 * cache locally
1043 * Fills in the private data members
1045 static void i5400_get_mc_regs(struct mem_ctl_info *mci)
1047 struct i5400_pvt *pvt;
1048 u32 actual_tolm;
1049 u16 limit;
1050 int slot_row;
1051 int maxch;
1052 int maxdimmperch;
1053 int way0, way1;
1055 pvt = mci->pvt_info;
1057 pci_read_config_dword(pvt->system_address, AMBASE,
1058 (u32 *) &pvt->ambase);
1059 pci_read_config_dword(pvt->system_address, AMBASE + sizeof(u32),
1060 ((u32 *) &pvt->ambase) + sizeof(u32));
1062 maxdimmperch = pvt->maxdimmperch;
1063 maxch = pvt->maxch;
1065 edac_dbg(2, "AMBASE= 0x%lx MAXCH= %d MAX-DIMM-Per-CH= %d\n",
1066 (long unsigned int)pvt->ambase, pvt->maxch, pvt->maxdimmperch);
1068 /* Get the Branch Map regs */
1069 pci_read_config_word(pvt->branchmap_werrors, TOLM, &pvt->tolm);
1070 pvt->tolm >>= 12;
1071 edac_dbg(2, "\nTOLM (number of 256M regions) =%u (0x%x)\n",
1072 pvt->tolm, pvt->tolm);
1074 actual_tolm = (u32) ((1000l * pvt->tolm) >> (30 - 28));
1075 edac_dbg(2, "Actual TOLM byte addr=%u.%03u GB (0x%x)\n",
1076 actual_tolm/1000, actual_tolm % 1000, pvt->tolm << 28);
1078 pci_read_config_word(pvt->branchmap_werrors, MIR0, &pvt->mir0);
1079 pci_read_config_word(pvt->branchmap_werrors, MIR1, &pvt->mir1);
1081 /* Get the MIR[0-1] regs */
1082 limit = (pvt->mir0 >> 4) & 0x0fff;
1083 way0 = pvt->mir0 & 0x1;
1084 way1 = pvt->mir0 & 0x2;
1085 edac_dbg(2, "MIR0: limit= 0x%x WAY1= %u WAY0= %x\n",
1086 limit, way1, way0);
1087 limit = (pvt->mir1 >> 4) & 0xfff;
1088 way0 = pvt->mir1 & 0x1;
1089 way1 = pvt->mir1 & 0x2;
1090 edac_dbg(2, "MIR1: limit= 0x%x WAY1= %u WAY0= %x\n",
1091 limit, way1, way0);
1093 /* Get the set of MTR[0-3] regs by each branch */
1094 for (slot_row = 0; slot_row < DIMMS_PER_CHANNEL; slot_row++) {
1095 int where = MTR0 + (slot_row * sizeof(u16));
1097 /* Branch 0 set of MTR registers */
1098 pci_read_config_word(pvt->branch_0, where,
1099 &pvt->b0_mtr[slot_row]);
1101 edac_dbg(2, "MTR%d where=0x%x B0 value=0x%x\n",
1102 slot_row, where, pvt->b0_mtr[slot_row]);
1104 if (pvt->maxch < CHANNELS_PER_BRANCH) {
1105 pvt->b1_mtr[slot_row] = 0;
1106 continue;
1109 /* Branch 1 set of MTR registers */
1110 pci_read_config_word(pvt->branch_1, where,
1111 &pvt->b1_mtr[slot_row]);
1112 edac_dbg(2, "MTR%d where=0x%x B1 value=0x%x\n",
1113 slot_row, where, pvt->b1_mtr[slot_row]);
1116 /* Read and dump branch 0's MTRs */
1117 edac_dbg(2, "Memory Technology Registers:\n");
1118 edac_dbg(2, " Branch 0:\n");
1119 for (slot_row = 0; slot_row < DIMMS_PER_CHANNEL; slot_row++)
1120 decode_mtr(slot_row, pvt->b0_mtr[slot_row]);
1122 pci_read_config_word(pvt->branch_0, AMBPRESENT_0,
1123 &pvt->b0_ambpresent0);
1124 edac_dbg(2, "\t\tAMB-Branch 0-present0 0x%x:\n", pvt->b0_ambpresent0);
1125 pci_read_config_word(pvt->branch_0, AMBPRESENT_1,
1126 &pvt->b0_ambpresent1);
1127 edac_dbg(2, "\t\tAMB-Branch 0-present1 0x%x:\n", pvt->b0_ambpresent1);
1129 /* Only if we have 2 branchs (4 channels) */
1130 if (pvt->maxch < CHANNELS_PER_BRANCH) {
1131 pvt->b1_ambpresent0 = 0;
1132 pvt->b1_ambpresent1 = 0;
1133 } else {
1134 /* Read and dump branch 1's MTRs */
1135 edac_dbg(2, " Branch 1:\n");
1136 for (slot_row = 0; slot_row < DIMMS_PER_CHANNEL; slot_row++)
1137 decode_mtr(slot_row, pvt->b1_mtr[slot_row]);
1139 pci_read_config_word(pvt->branch_1, AMBPRESENT_0,
1140 &pvt->b1_ambpresent0);
1141 edac_dbg(2, "\t\tAMB-Branch 1-present0 0x%x:\n",
1142 pvt->b1_ambpresent0);
1143 pci_read_config_word(pvt->branch_1, AMBPRESENT_1,
1144 &pvt->b1_ambpresent1);
1145 edac_dbg(2, "\t\tAMB-Branch 1-present1 0x%x:\n",
1146 pvt->b1_ambpresent1);
1149 /* Go and determine the size of each DIMM and place in an
1150 * orderly matrix */
1151 calculate_dimm_size(pvt);
1155 * i5400_init_dimms Initialize the 'dimms' table within
1156 * the mci control structure with the
1157 * addressing of memory.
1159 * return:
1160 * 0 success
1161 * 1 no actual memory found on this MC
1163 static int i5400_init_dimms(struct mem_ctl_info *mci)
1165 struct i5400_pvt *pvt;
1166 struct dimm_info *dimm;
1167 int ndimms, channel_count;
1168 int max_dimms;
1169 int mtr;
1170 int size_mb;
1171 int channel, slot;
1173 pvt = mci->pvt_info;
1175 channel_count = pvt->maxch;
1176 max_dimms = pvt->maxdimmperch;
1178 ndimms = 0;
1181 * FIXME: remove pvt->dimm_info[slot][channel] and use the 3
1182 * layers here.
1184 for (channel = 0; channel < mci->layers[0].size * mci->layers[1].size;
1185 channel++) {
1186 for (slot = 0; slot < mci->layers[2].size; slot++) {
1187 mtr = determine_mtr(pvt, slot, channel);
1189 /* if no DIMMS on this slot, continue */
1190 if (!MTR_DIMMS_PRESENT(mtr))
1191 continue;
1193 dimm = EDAC_DIMM_PTR(mci->layers, mci->dimms, mci->n_layers,
1194 channel / 2, channel % 2, slot);
1196 size_mb = pvt->dimm_info[slot][channel].megabytes;
1198 edac_dbg(2, "dimm (branch %d channel %d slot %d): %d.%03d GB\n",
1199 channel / 2, channel % 2, slot,
1200 size_mb / 1000, size_mb % 1000);
1202 dimm->nr_pages = size_mb << 8;
1203 dimm->grain = 8;
1204 dimm->dtype = MTR_DRAM_WIDTH(mtr) ? DEV_X8 : DEV_X4;
1205 dimm->mtype = MEM_FB_DDR2;
1207 * The eccc mechanism is SDDC (aka SECC), with
1208 * is similar to Chipkill.
1210 dimm->edac_mode = MTR_DRAM_WIDTH(mtr) ?
1211 EDAC_S8ECD8ED : EDAC_S4ECD4ED;
1212 ndimms++;
1217 * When just one memory is provided, it should be at location (0,0,0).
1218 * With such single-DIMM mode, the SDCC algorithm degrades to SECDEC+.
1220 if (ndimms == 1)
1221 mci->dimms[0]->edac_mode = EDAC_SECDED;
1223 return (ndimms == 0);
1227 * i5400_enable_error_reporting
1228 * Turn on the memory reporting features of the hardware
1230 static void i5400_enable_error_reporting(struct mem_ctl_info *mci)
1232 struct i5400_pvt *pvt;
1233 u32 fbd_error_mask;
1235 pvt = mci->pvt_info;
1237 /* Read the FBD Error Mask Register */
1238 pci_read_config_dword(pvt->branchmap_werrors, EMASK_FBD,
1239 &fbd_error_mask);
1241 /* Enable with a '0' */
1242 fbd_error_mask &= ~(ENABLE_EMASK_ALL);
1244 pci_write_config_dword(pvt->branchmap_werrors, EMASK_FBD,
1245 fbd_error_mask);
1249 * i5400_probe1 Probe for ONE instance of device to see if it is
1250 * present.
1251 * return:
1252 * 0 for FOUND a device
1253 * < 0 for error code
1255 static int i5400_probe1(struct pci_dev *pdev, int dev_idx)
1257 struct mem_ctl_info *mci;
1258 struct i5400_pvt *pvt;
1259 struct edac_mc_layer layers[3];
1261 if (dev_idx >= ARRAY_SIZE(i5400_devs))
1262 return -EINVAL;
1264 edac_dbg(0, "MC: pdev bus %u dev=0x%x fn=0x%x\n",
1265 pdev->bus->number,
1266 PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
1268 /* We only are looking for func 0 of the set */
1269 if (PCI_FUNC(pdev->devfn) != 0)
1270 return -ENODEV;
1273 * allocate a new MC control structure
1275 * This drivers uses the DIMM slot as "csrow" and the rest as "channel".
1277 layers[0].type = EDAC_MC_LAYER_BRANCH;
1278 layers[0].size = MAX_BRANCHES;
1279 layers[0].is_virt_csrow = false;
1280 layers[1].type = EDAC_MC_LAYER_CHANNEL;
1281 layers[1].size = CHANNELS_PER_BRANCH;
1282 layers[1].is_virt_csrow = false;
1283 layers[2].type = EDAC_MC_LAYER_SLOT;
1284 layers[2].size = DIMMS_PER_CHANNEL;
1285 layers[2].is_virt_csrow = true;
1286 mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, sizeof(*pvt));
1287 if (mci == NULL)
1288 return -ENOMEM;
1290 edac_dbg(0, "MC: mci = %p\n", mci);
1292 mci->pdev = &pdev->dev; /* record ptr to the generic device */
1294 pvt = mci->pvt_info;
1295 pvt->system_address = pdev; /* Record this device in our private */
1296 pvt->maxch = MAX_CHANNELS;
1297 pvt->maxdimmperch = DIMMS_PER_CHANNEL;
1299 /* 'get' the pci devices we want to reserve for our use */
1300 if (i5400_get_devices(mci, dev_idx))
1301 goto fail0;
1303 /* Time to get serious */
1304 i5400_get_mc_regs(mci); /* retrieve the hardware registers */
1306 mci->mc_idx = 0;
1307 mci->mtype_cap = MEM_FLAG_FB_DDR2;
1308 mci->edac_ctl_cap = EDAC_FLAG_NONE;
1309 mci->edac_cap = EDAC_FLAG_NONE;
1310 mci->mod_name = "i5400_edac.c";
1311 mci->mod_ver = I5400_REVISION;
1312 mci->ctl_name = i5400_devs[dev_idx].ctl_name;
1313 mci->dev_name = pci_name(pdev);
1314 mci->ctl_page_to_phys = NULL;
1316 /* Set the function pointer to an actual operation function */
1317 mci->edac_check = i5400_check_error;
1319 /* initialize the MC control structure 'dimms' table
1320 * with the mapping and control information */
1321 if (i5400_init_dimms(mci)) {
1322 edac_dbg(0, "MC: Setting mci->edac_cap to EDAC_FLAG_NONE because i5400_init_dimms() returned nonzero value\n");
1323 mci->edac_cap = EDAC_FLAG_NONE; /* no dimms found */
1324 } else {
1325 edac_dbg(1, "MC: Enable error reporting now\n");
1326 i5400_enable_error_reporting(mci);
1329 /* add this new MC control structure to EDAC's list of MCs */
1330 if (edac_mc_add_mc(mci)) {
1331 edac_dbg(0, "MC: failed edac_mc_add_mc()\n");
1332 /* FIXME: perhaps some code should go here that disables error
1333 * reporting if we just enabled it
1335 goto fail1;
1338 i5400_clear_error(mci);
1340 /* allocating generic PCI control info */
1341 i5400_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
1342 if (!i5400_pci) {
1343 printk(KERN_WARNING
1344 "%s(): Unable to create PCI control\n",
1345 __func__);
1346 printk(KERN_WARNING
1347 "%s(): PCI error report via EDAC not setup\n",
1348 __func__);
1351 return 0;
1353 /* Error exit unwinding stack */
1354 fail1:
1356 i5400_put_devices(mci);
1358 fail0:
1359 edac_mc_free(mci);
1360 return -ENODEV;
1364 * i5400_init_one constructor for one instance of device
1366 * returns:
1367 * negative on error
1368 * count (>= 0)
1370 static int __devinit i5400_init_one(struct pci_dev *pdev,
1371 const struct pci_device_id *id)
1373 int rc;
1375 edac_dbg(0, "MC:\n");
1377 /* wake up device */
1378 rc = pci_enable_device(pdev);
1379 if (rc)
1380 return rc;
1382 /* now probe and enable the device */
1383 return i5400_probe1(pdev, id->driver_data);
1387 * i5400_remove_one destructor for one instance of device
1390 static void __devexit i5400_remove_one(struct pci_dev *pdev)
1392 struct mem_ctl_info *mci;
1394 edac_dbg(0, "\n");
1396 if (i5400_pci)
1397 edac_pci_release_generic_ctl(i5400_pci);
1399 mci = edac_mc_del_mc(&pdev->dev);
1400 if (!mci)
1401 return;
1403 /* retrieve references to resources, and free those resources */
1404 i5400_put_devices(mci);
1406 edac_mc_free(mci);
1410 * pci_device_id table for which devices we are looking for
1412 * The "E500P" device is the first device supported.
1414 static DEFINE_PCI_DEVICE_TABLE(i5400_pci_tbl) = {
1415 {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_5400_ERR)},
1416 {0,} /* 0 terminated list. */
1419 MODULE_DEVICE_TABLE(pci, i5400_pci_tbl);
1422 * i5400_driver pci_driver structure for this module
1425 static struct pci_driver i5400_driver = {
1426 .name = "i5400_edac",
1427 .probe = i5400_init_one,
1428 .remove = __devexit_p(i5400_remove_one),
1429 .id_table = i5400_pci_tbl,
1433 * i5400_init Module entry function
1434 * Try to initialize this module for its devices
1436 static int __init i5400_init(void)
1438 int pci_rc;
1440 edac_dbg(2, "MC:\n");
1442 /* Ensure that the OPSTATE is set correctly for POLL or NMI */
1443 opstate_init();
1445 pci_rc = pci_register_driver(&i5400_driver);
1447 return (pci_rc < 0) ? pci_rc : 0;
1451 * i5400_exit() Module exit function
1452 * Unregister the driver
1454 static void __exit i5400_exit(void)
1456 edac_dbg(2, "MC:\n");
1457 pci_unregister_driver(&i5400_driver);
1460 module_init(i5400_init);
1461 module_exit(i5400_exit);
1463 MODULE_LICENSE("GPL");
1464 MODULE_AUTHOR("Ben Woodard <woodard@redhat.com>");
1465 MODULE_AUTHOR("Mauro Carvalho Chehab <mchehab@redhat.com>");
1466 MODULE_AUTHOR("Red Hat Inc. (http://www.redhat.com)");
1467 MODULE_DESCRIPTION("MC Driver for Intel I5400 memory controllers - "
1468 I5400_REVISION);
1470 module_param(edac_op_state, int, 0444);
1471 MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");