2 * Intel e752x Memory Controller kernel module
3 * (C) 2004 Linux Networx (http://lnxi.com)
4 * This file may be distributed under the terms of the
5 * GNU General Public License.
7 * See "enum e752x_chips" below for supported chipsets
9 * Written by Tom Zimmerman
12 * Thayne Harbaugh at realmsys.com (?)
13 * Wang Zhenyu at intel.com
14 * Dave Jiang at mvista.com
16 * $Id: edac_e752x.c,v 1.5.2.11 2005/10/05 00:43:44 dsp_llnl Exp $
20 #include <linux/module.h>
21 #include <linux/init.h>
22 #include <linux/pci.h>
23 #include <linux/pci_ids.h>
24 #include <linux/slab.h>
25 #include <linux/edac.h>
26 #include "edac_core.h"
28 #define E752X_REVISION " Ver: 2.0.2 " __DATE__
29 #define EDAC_MOD_STR "e752x_edac"
31 static int report_non_memory_errors;
32 static int force_function_unhide;
33 static int sysbus_parity = -1;
35 static struct edac_pci_ctl_info *e752x_pci;
37 #define e752x_printk(level, fmt, arg...) \
38 edac_printk(level, "e752x", fmt, ##arg)
40 #define e752x_mc_printk(mci, level, fmt, arg...) \
41 edac_mc_chipset_printk(mci, level, "e752x", fmt, ##arg)
43 #ifndef PCI_DEVICE_ID_INTEL_7520_0
44 #define PCI_DEVICE_ID_INTEL_7520_0 0x3590
45 #endif /* PCI_DEVICE_ID_INTEL_7520_0 */
47 #ifndef PCI_DEVICE_ID_INTEL_7520_1_ERR
48 #define PCI_DEVICE_ID_INTEL_7520_1_ERR 0x3591
49 #endif /* PCI_DEVICE_ID_INTEL_7520_1_ERR */
51 #ifndef PCI_DEVICE_ID_INTEL_7525_0
52 #define PCI_DEVICE_ID_INTEL_7525_0 0x359E
53 #endif /* PCI_DEVICE_ID_INTEL_7525_0 */
55 #ifndef PCI_DEVICE_ID_INTEL_7525_1_ERR
56 #define PCI_DEVICE_ID_INTEL_7525_1_ERR 0x3593
57 #endif /* PCI_DEVICE_ID_INTEL_7525_1_ERR */
59 #ifndef PCI_DEVICE_ID_INTEL_7320_0
60 #define PCI_DEVICE_ID_INTEL_7320_0 0x3592
61 #endif /* PCI_DEVICE_ID_INTEL_7320_0 */
63 #ifndef PCI_DEVICE_ID_INTEL_7320_1_ERR
64 #define PCI_DEVICE_ID_INTEL_7320_1_ERR 0x3593
65 #endif /* PCI_DEVICE_ID_INTEL_7320_1_ERR */
67 #ifndef PCI_DEVICE_ID_INTEL_3100_0
68 #define PCI_DEVICE_ID_INTEL_3100_0 0x35B0
69 #endif /* PCI_DEVICE_ID_INTEL_3100_0 */
71 #ifndef PCI_DEVICE_ID_INTEL_3100_1_ERR
72 #define PCI_DEVICE_ID_INTEL_3100_1_ERR 0x35B1
73 #endif /* PCI_DEVICE_ID_INTEL_3100_1_ERR */
75 #define E752X_NR_CSROWS 8 /* number of csrows */
77 /* E752X register addresses - device 0 function 0 */
78 #define E752X_DRB 0x60 /* DRAM row boundary register (8b) */
79 #define E752X_DRA 0x70 /* DRAM row attribute register (8b) */
81 * 31:30 Device width row 7
82 * 01=x8 10=x4 11=x8 DDR2
83 * 27:26 Device width row 6
84 * 23:22 Device width row 5
85 * 19:20 Device width row 4
86 * 15:14 Device width row 3
87 * 11:10 Device width row 2
88 * 7:6 Device width row 1
89 * 3:2 Device width row 0
91 #define E752X_DRC 0x7C /* DRAM controller mode reg (32b) */
92 /* FIXME:IS THIS RIGHT? */
94 * 22 Number channels 0=1,1=2
95 * 19:18 DRB Granularity 32/64MB
97 #define E752X_DRM 0x80 /* Dimm mapping register */
98 #define E752X_DDRCSR 0x9A /* DDR control and status reg (16b) */
100 * 14:12 1 single A, 2 single B, 3 dual
102 #define E752X_TOLM 0xC4 /* DRAM top of low memory reg (16b) */
103 #define E752X_REMAPBASE 0xC6 /* DRAM remap base address reg (16b) */
104 #define E752X_REMAPLIMIT 0xC8 /* DRAM remap limit address reg (16b) */
105 #define E752X_REMAPOFFSET 0xCA /* DRAM remap limit offset reg (16b) */
107 /* E752X register addresses - device 0 function 1 */
108 #define E752X_FERR_GLOBAL 0x40 /* Global first error register (32b) */
109 #define E752X_NERR_GLOBAL 0x44 /* Global next error register (32b) */
110 #define E752X_HI_FERR 0x50 /* Hub interface first error reg (8b) */
111 #define E752X_HI_NERR 0x52 /* Hub interface next error reg (8b) */
112 #define E752X_HI_ERRMASK 0x54 /* Hub interface error mask reg (8b) */
113 #define E752X_HI_SMICMD 0x5A /* Hub interface SMI command reg (8b) */
114 #define E752X_SYSBUS_FERR 0x60 /* System buss first error reg (16b) */
115 #define E752X_SYSBUS_NERR 0x62 /* System buss next error reg (16b) */
116 #define E752X_SYSBUS_ERRMASK 0x64 /* System buss error mask reg (16b) */
117 #define E752X_SYSBUS_SMICMD 0x6A /* System buss SMI command reg (16b) */
118 #define E752X_BUF_FERR 0x70 /* Memory buffer first error reg (8b) */
119 #define E752X_BUF_NERR 0x72 /* Memory buffer next error reg (8b) */
120 #define E752X_BUF_ERRMASK 0x74 /* Memory buffer error mask reg (8b) */
121 #define E752X_BUF_SMICMD 0x7A /* Memory buffer SMI cmd reg (8b) */
122 #define E752X_DRAM_FERR 0x80 /* DRAM first error register (16b) */
123 #define E752X_DRAM_NERR 0x82 /* DRAM next error register (16b) */
124 #define E752X_DRAM_ERRMASK 0x84 /* DRAM error mask register (8b) */
125 #define E752X_DRAM_SMICMD 0x8A /* DRAM SMI command register (8b) */
126 #define E752X_DRAM_RETR_ADD 0xAC /* DRAM Retry address register (32b) */
127 #define E752X_DRAM_SEC1_ADD 0xA0 /* DRAM first correctable memory */
128 /* error address register (32b) */
131 * 30:2 CE address (64 byte block 34:6
135 #define E752X_DRAM_SEC2_ADD 0xC8 /* DRAM first correctable memory */
136 /* error address register (32b) */
139 * 30:2 CE address (64 byte block 34:6)
143 #define E752X_DRAM_DED_ADD 0xA4 /* DRAM first uncorrectable memory */
144 /* error address register (32b) */
147 * 30:2 CE address (64 byte block 34:6)
151 #define E752X_DRAM_SCRB_ADD 0xA8 /* DRAM 1st uncorrectable scrub mem */
152 /* error address register (32b) */
155 * 30:2 CE address (64 byte block 34:6
159 #define E752X_DRAM_SEC1_SYNDROME 0xC4 /* DRAM first correctable memory */
160 /* error syndrome register (16b) */
161 #define E752X_DRAM_SEC2_SYNDROME 0xC6 /* DRAM second correctable memory */
162 /* error syndrome register (16b) */
163 #define E752X_DEVPRES1 0xF4 /* Device Present 1 register (8b) */
165 /* 3100 IMCH specific register addresses - device 0 function 1 */
166 #define I3100_NSI_FERR 0x48 /* NSI first error reg (32b) */
167 #define I3100_NSI_NERR 0x4C /* NSI next error reg (32b) */
168 #define I3100_NSI_SMICMD 0x54 /* NSI SMI command register (32b) */
169 #define I3100_NSI_EMASK 0x90 /* NSI error mask register (32b) */
171 /* ICH5R register addresses - device 30 function 0 */
172 #define ICH5R_PCI_STAT 0x06 /* PCI status register (16b) */
173 #define ICH5R_PCI_2ND_STAT 0x1E /* PCI status secondary reg (16b) */
174 #define ICH5R_PCI_BRIDGE_CTL 0x3E /* PCI bridge control register (16b) */
184 struct pci_dev *bridge_ck;
185 struct pci_dev *dev_d0f0;
186 struct pci_dev *dev_d0f1;
193 const struct e752x_dev_info *dev_info;
196 struct e752x_dev_info {
199 const char *ctl_name;
202 struct e752x_error_info {
205 u32 nsi_ferr; /* 3100 only */
206 u32 nsi_nerr; /* 3100 only */
207 u8 hi_ferr; /* all but 3100 */
208 u8 hi_nerr; /* all but 3100 */
217 u16 dram_sec1_syndrome;
218 u16 dram_sec2_syndrome;
224 static const struct e752x_dev_info e752x_devs[] = {
226 .err_dev = PCI_DEVICE_ID_INTEL_7520_1_ERR,
227 .ctl_dev = PCI_DEVICE_ID_INTEL_7520_0,
228 .ctl_name = "E7520"},
230 .err_dev = PCI_DEVICE_ID_INTEL_7525_1_ERR,
231 .ctl_dev = PCI_DEVICE_ID_INTEL_7525_0,
232 .ctl_name = "E7525"},
234 .err_dev = PCI_DEVICE_ID_INTEL_7320_1_ERR,
235 .ctl_dev = PCI_DEVICE_ID_INTEL_7320_0,
236 .ctl_name = "E7320"},
238 .err_dev = PCI_DEVICE_ID_INTEL_3100_1_ERR,
239 .ctl_dev = PCI_DEVICE_ID_INTEL_3100_0,
243 static unsigned long ctl_page_to_phys(struct mem_ctl_info *mci,
247 struct e752x_pvt *pvt = (struct e752x_pvt *)mci->pvt_info;
249 debugf3("%s()\n", __func__);
251 if (page < pvt->tolm)
254 if ((page >= 0x100000) && (page < pvt->remapbase))
257 remap = (page - pvt->tolm) + pvt->remapbase;
259 if (remap < pvt->remaplimit)
262 e752x_printk(KERN_ERR, "Invalid page %lx - out of range\n", page);
263 return pvt->tolm - 1;
266 static void do_process_ce(struct mem_ctl_info *mci, u16 error_one,
267 u32 sec1_add, u16 sec1_syndrome)
273 struct e752x_pvt *pvt = (struct e752x_pvt *)mci->pvt_info;
275 debugf3("%s()\n", __func__);
277 /* convert the addr to 4k page */
278 page = sec1_add >> (PAGE_SHIFT - 4);
280 /* FIXME - check for -1 */
281 if (pvt->mc_symmetric) {
282 /* chip select are bits 14 & 13 */
283 row = ((page >> 1) & 3);
284 e752x_printk(KERN_WARNING,
285 "Test row %d Table %d %d %d %d %d %d %d %d\n", row,
286 pvt->map[0], pvt->map[1], pvt->map[2], pvt->map[3],
287 pvt->map[4], pvt->map[5], pvt->map[6],
290 /* test for channel remapping */
291 for (i = 0; i < 8; i++) {
292 if (pvt->map[i] == row)
296 e752x_printk(KERN_WARNING, "Test computed row %d\n", i);
301 e752x_mc_printk(mci, KERN_WARNING,
302 "row %d not found in remap table\n",
305 row = edac_mc_find_csrow_by_page(mci, page);
307 /* 0 = channel A, 1 = channel B */
308 channel = !(error_one & 1);
310 /* e752x mc reads 34:6 of the DRAM linear address */
311 edac_mc_handle_ce(mci, page, offset_in_page(sec1_add << 4),
312 sec1_syndrome, row, channel, "e752x CE");
315 static inline void process_ce(struct mem_ctl_info *mci, u16 error_one,
316 u32 sec1_add, u16 sec1_syndrome, int *error_found,
322 do_process_ce(mci, error_one, sec1_add, sec1_syndrome);
325 static void do_process_ue(struct mem_ctl_info *mci, u16 error_one,
326 u32 ded_add, u32 scrb_add)
328 u32 error_2b, block_page;
330 struct e752x_pvt *pvt = (struct e752x_pvt *)mci->pvt_info;
332 debugf3("%s()\n", __func__);
334 if (error_one & 0x0202) {
337 /* convert to 4k address */
338 block_page = error_2b >> (PAGE_SHIFT - 4);
340 row = pvt->mc_symmetric ?
341 /* chip select are bits 14 & 13 */
342 ((block_page >> 1) & 3) :
343 edac_mc_find_csrow_by_page(mci, block_page);
345 /* e752x mc reads 34:6 of the DRAM linear address */
346 edac_mc_handle_ue(mci, block_page,
347 offset_in_page(error_2b << 4),
348 row, "e752x UE from Read");
350 if (error_one & 0x0404) {
353 /* convert to 4k address */
354 block_page = error_2b >> (PAGE_SHIFT - 4);
356 row = pvt->mc_symmetric ?
357 /* chip select are bits 14 & 13 */
358 ((block_page >> 1) & 3) :
359 edac_mc_find_csrow_by_page(mci, block_page);
361 /* e752x mc reads 34:6 of the DRAM linear address */
362 edac_mc_handle_ue(mci, block_page,
363 offset_in_page(error_2b << 4),
364 row, "e752x UE from Scruber");
368 static inline void process_ue(struct mem_ctl_info *mci, u16 error_one,
369 u32 ded_add, u32 scrb_add, int *error_found,
375 do_process_ue(mci, error_one, ded_add, scrb_add);
378 static inline void process_ue_no_info_wr(struct mem_ctl_info *mci,
379 int *error_found, int handle_error)
386 debugf3("%s()\n", __func__);
387 edac_mc_handle_ue_no_info(mci, "e752x UE log memory write");
390 static void do_process_ded_retry(struct mem_ctl_info *mci, u16 error,
395 struct e752x_pvt *pvt = (struct e752x_pvt *)mci->pvt_info;
397 error_1b = retry_add;
398 page = error_1b >> (PAGE_SHIFT - 4); /* convert the addr to 4k page */
400 /* chip select are bits 14 & 13 */
401 row = pvt->mc_symmetric ? ((page >> 1) & 3) :
402 edac_mc_find_csrow_by_page(mci, page);
404 e752x_mc_printk(mci, KERN_WARNING,
405 "CE page 0x%lx, row %d : Memory read retry\n",
406 (long unsigned int)page, row);
409 static inline void process_ded_retry(struct mem_ctl_info *mci, u16 error,
410 u32 retry_add, int *error_found,
416 do_process_ded_retry(mci, error, retry_add);
419 static inline void process_threshold_ce(struct mem_ctl_info *mci, u16 error,
420 int *error_found, int handle_error)
425 e752x_mc_printk(mci, KERN_WARNING, "Memory threshold CE\n");
428 static char *global_message[11] = {
436 "HUB or NS Interface",
438 "DRAM Controller", /* 9th entry */
444 static char *fatal_message[2] = { "Non-Fatal ", "Fatal " };
446 static void do_global_error(int fatal, u32 errors)
450 for (i = 0; i < 11; i++) {
451 if (errors & (1 << i)) {
452 /* If the error is from DRAM Controller OR
453 * we are to report ALL errors, then
456 if ((i == DRAM_ENTRY) || report_non_memory_errors)
457 e752x_printk(KERN_WARNING, "%sError %s\n",
458 fatal_message[fatal],
464 static inline void global_error(int fatal, u32 errors, int *error_found,
470 do_global_error(fatal, errors);
473 static char *hub_message[7] = {
474 "HI Address or Command Parity", "HI Illegal Access",
475 "HI Internal Parity", "Out of Range Access",
476 "HI Data Parity", "Enhanced Config Access",
477 "Hub Interface Target Abort"
480 static void do_hub_error(int fatal, u8 errors)
484 for (i = 0; i < 7; i++) {
485 if (errors & (1 << i))
486 e752x_printk(KERN_WARNING, "%sError %s\n",
487 fatal_message[fatal], hub_message[i]);
491 static inline void hub_error(int fatal, u8 errors, int *error_found,
497 do_hub_error(fatal, errors);
500 #define NSI_FATAL_MASK 0x0c080081
501 #define NSI_NON_FATAL_MASK 0x23a0ba64
502 #define NSI_ERR_MASK (NSI_FATAL_MASK | NSI_NON_FATAL_MASK)
504 static char *nsi_message[30] = {
505 "NSI Link Down", /* NSI_FERR/NSI_NERR bit 0, fatal error */
507 "NSI Parity Error", /* bit 2, non-fatal */
510 "Correctable Error Message", /* bit 5, non-fatal */
511 "Non-Fatal Error Message", /* bit 6, non-fatal */
512 "Fatal Error Message", /* bit 7, fatal */
514 "Receiver Error", /* bit 9, non-fatal */
516 "Bad TLP", /* bit 11, non-fatal */
517 "Bad DLLP", /* bit 12, non-fatal */
518 "REPLAY_NUM Rollover", /* bit 13, non-fatal */
520 "Replay Timer Timeout", /* bit 15, non-fatal */
524 "Data Link Protocol Error", /* bit 19, fatal */
526 "Poisoned TLP", /* bit 21, non-fatal */
528 "Completion Timeout", /* bit 23, non-fatal */
529 "Completer Abort", /* bit 24, non-fatal */
530 "Unexpected Completion", /* bit 25, non-fatal */
531 "Receiver Overflow", /* bit 26, fatal */
532 "Malformed TLP", /* bit 27, fatal */
534 "Unsupported Request" /* bit 29, non-fatal */
537 static void do_nsi_error(int fatal, u32 errors)
541 for (i = 0; i < 30; i++) {
542 if (errors & (1 << i))
543 printk(KERN_WARNING "%sError %s\n",
544 fatal_message[fatal], nsi_message[i]);
548 static inline void nsi_error(int fatal, u32 errors, int *error_found,
554 do_nsi_error(fatal, errors);
557 static char *membuf_message[4] = {
558 "Internal PMWB to DRAM parity",
559 "Internal PMWB to System Bus Parity",
560 "Internal System Bus or IO to PMWB Parity",
561 "Internal DRAM to PMWB Parity"
564 static void do_membuf_error(u8 errors)
568 for (i = 0; i < 4; i++) {
569 if (errors & (1 << i))
570 e752x_printk(KERN_WARNING, "Non-Fatal Error %s\n",
575 static inline void membuf_error(u8 errors, int *error_found, int handle_error)
580 do_membuf_error(errors);
583 static char *sysbus_message[10] = {
584 "Addr or Request Parity",
585 "Data Strobe Glitch",
586 "Addr Strobe Glitch",
589 "Non DRAM Lock Error",
592 "IO Subsystem Parity"
595 static void do_sysbus_error(int fatal, u32 errors)
599 for (i = 0; i < 10; i++) {
600 if (errors & (1 << i))
601 e752x_printk(KERN_WARNING, "%sError System Bus %s\n",
602 fatal_message[fatal], sysbus_message[i]);
606 static inline void sysbus_error(int fatal, u32 errors, int *error_found,
612 do_sysbus_error(fatal, errors);
615 static void e752x_check_hub_interface(struct e752x_error_info *info,
616 int *error_found, int handle_error)
620 //pci_read_config_byte(dev,E752X_HI_FERR,&stat8);
622 stat8 = info->hi_ferr;
624 if (stat8 & 0x7f) { /* Error, so process */
628 hub_error(1, stat8 & 0x2b, error_found, handle_error);
631 hub_error(0, stat8 & 0x54, error_found, handle_error);
633 //pci_read_config_byte(dev,E752X_HI_NERR,&stat8);
635 stat8 = info->hi_nerr;
637 if (stat8 & 0x7f) { /* Error, so process */
641 hub_error(1, stat8 & 0x2b, error_found, handle_error);
644 hub_error(0, stat8 & 0x54, error_found, handle_error);
648 static void e752x_check_ns_interface(struct e752x_error_info *info,
649 int *error_found, int handle_error)
653 stat32 = info->nsi_ferr;
654 if (stat32 & NSI_ERR_MASK) { /* Error, so process */
655 if (stat32 & NSI_FATAL_MASK) /* check for fatal errors */
656 nsi_error(1, stat32 & NSI_FATAL_MASK, error_found,
658 if (stat32 & NSI_NON_FATAL_MASK) /* check for non-fatal ones */
659 nsi_error(0, stat32 & NSI_NON_FATAL_MASK, error_found,
662 stat32 = info->nsi_nerr;
663 if (stat32 & NSI_ERR_MASK) {
664 if (stat32 & NSI_FATAL_MASK)
665 nsi_error(1, stat32 & NSI_FATAL_MASK, error_found,
667 if (stat32 & NSI_NON_FATAL_MASK)
668 nsi_error(0, stat32 & NSI_NON_FATAL_MASK, error_found,
673 static void e752x_check_sysbus(struct e752x_error_info *info,
674 int *error_found, int handle_error)
678 //pci_read_config_dword(dev,E752X_SYSBUS_FERR,&stat32);
679 stat32 = info->sysbus_ferr + (info->sysbus_nerr << 16);
682 return; /* no errors */
684 error32 = (stat32 >> 16) & 0x3ff;
685 stat32 = stat32 & 0x3ff;
688 sysbus_error(1, stat32 & 0x087, error_found, handle_error);
691 sysbus_error(0, stat32 & 0x378, error_found, handle_error);
694 sysbus_error(1, error32 & 0x087, error_found, handle_error);
697 sysbus_error(0, error32 & 0x378, error_found, handle_error);
700 static void e752x_check_membuf(struct e752x_error_info *info,
701 int *error_found, int handle_error)
705 stat8 = info->buf_ferr;
707 if (stat8 & 0x0f) { /* Error, so process */
709 membuf_error(stat8, error_found, handle_error);
712 stat8 = info->buf_nerr;
714 if (stat8 & 0x0f) { /* Error, so process */
716 membuf_error(stat8, error_found, handle_error);
720 static void e752x_check_dram(struct mem_ctl_info *mci,
721 struct e752x_error_info *info, int *error_found,
724 u16 error_one, error_next;
726 error_one = info->dram_ferr;
727 error_next = info->dram_nerr;
729 /* decode and report errors */
730 if (error_one & 0x0101) /* check first error correctable */
731 process_ce(mci, error_one, info->dram_sec1_add,
732 info->dram_sec1_syndrome, error_found, handle_error);
734 if (error_next & 0x0101) /* check next error correctable */
735 process_ce(mci, error_next, info->dram_sec2_add,
736 info->dram_sec2_syndrome, error_found, handle_error);
738 if (error_one & 0x4040)
739 process_ue_no_info_wr(mci, error_found, handle_error);
741 if (error_next & 0x4040)
742 process_ue_no_info_wr(mci, error_found, handle_error);
744 if (error_one & 0x2020)
745 process_ded_retry(mci, error_one, info->dram_retr_add,
746 error_found, handle_error);
748 if (error_next & 0x2020)
749 process_ded_retry(mci, error_next, info->dram_retr_add,
750 error_found, handle_error);
752 if (error_one & 0x0808)
753 process_threshold_ce(mci, error_one, error_found, handle_error);
755 if (error_next & 0x0808)
756 process_threshold_ce(mci, error_next, error_found,
759 if (error_one & 0x0606)
760 process_ue(mci, error_one, info->dram_ded_add,
761 info->dram_scrb_add, error_found, handle_error);
763 if (error_next & 0x0606)
764 process_ue(mci, error_next, info->dram_ded_add,
765 info->dram_scrb_add, error_found, handle_error);
768 static void e752x_get_error_info(struct mem_ctl_info *mci,
769 struct e752x_error_info *info)
772 struct e752x_pvt *pvt;
774 memset(info, 0, sizeof(*info));
775 pvt = (struct e752x_pvt *)mci->pvt_info;
777 pci_read_config_dword(dev, E752X_FERR_GLOBAL, &info->ferr_global);
779 if (info->ferr_global) {
780 if (pvt->dev_info->err_dev == PCI_DEVICE_ID_INTEL_3100_1_ERR) {
781 pci_read_config_dword(dev, I3100_NSI_FERR,
785 pci_read_config_byte(dev, E752X_HI_FERR,
789 pci_read_config_word(dev, E752X_SYSBUS_FERR,
791 pci_read_config_byte(dev, E752X_BUF_FERR, &info->buf_ferr);
792 pci_read_config_word(dev, E752X_DRAM_FERR, &info->dram_ferr);
793 pci_read_config_dword(dev, E752X_DRAM_SEC1_ADD,
794 &info->dram_sec1_add);
795 pci_read_config_word(dev, E752X_DRAM_SEC1_SYNDROME,
796 &info->dram_sec1_syndrome);
797 pci_read_config_dword(dev, E752X_DRAM_DED_ADD,
798 &info->dram_ded_add);
799 pci_read_config_dword(dev, E752X_DRAM_SCRB_ADD,
800 &info->dram_scrb_add);
801 pci_read_config_dword(dev, E752X_DRAM_RETR_ADD,
802 &info->dram_retr_add);
804 /* ignore the reserved bits just in case */
805 if (info->hi_ferr & 0x7f)
806 pci_write_config_byte(dev, E752X_HI_FERR,
809 if (info->nsi_ferr & NSI_ERR_MASK)
810 pci_write_config_dword(dev, I3100_NSI_FERR,
813 if (info->sysbus_ferr)
814 pci_write_config_word(dev, E752X_SYSBUS_FERR,
817 if (info->buf_ferr & 0x0f)
818 pci_write_config_byte(dev, E752X_BUF_FERR,
822 pci_write_bits16(pvt->bridge_ck, E752X_DRAM_FERR,
823 info->dram_ferr, info->dram_ferr);
825 pci_write_config_dword(dev, E752X_FERR_GLOBAL,
829 pci_read_config_dword(dev, E752X_NERR_GLOBAL, &info->nerr_global);
831 if (info->nerr_global) {
832 if (pvt->dev_info->err_dev == PCI_DEVICE_ID_INTEL_3100_1_ERR) {
833 pci_read_config_dword(dev, I3100_NSI_NERR,
837 pci_read_config_byte(dev, E752X_HI_NERR,
841 pci_read_config_word(dev, E752X_SYSBUS_NERR,
843 pci_read_config_byte(dev, E752X_BUF_NERR, &info->buf_nerr);
844 pci_read_config_word(dev, E752X_DRAM_NERR, &info->dram_nerr);
845 pci_read_config_dword(dev, E752X_DRAM_SEC2_ADD,
846 &info->dram_sec2_add);
847 pci_read_config_word(dev, E752X_DRAM_SEC2_SYNDROME,
848 &info->dram_sec2_syndrome);
850 if (info->hi_nerr & 0x7f)
851 pci_write_config_byte(dev, E752X_HI_NERR,
854 if (info->nsi_nerr & NSI_ERR_MASK)
855 pci_write_config_dword(dev, I3100_NSI_NERR,
858 if (info->sysbus_nerr)
859 pci_write_config_word(dev, E752X_SYSBUS_NERR,
862 if (info->buf_nerr & 0x0f)
863 pci_write_config_byte(dev, E752X_BUF_NERR,
867 pci_write_bits16(pvt->bridge_ck, E752X_DRAM_NERR,
868 info->dram_nerr, info->dram_nerr);
870 pci_write_config_dword(dev, E752X_NERR_GLOBAL,
875 static int e752x_process_error_info(struct mem_ctl_info *mci,
876 struct e752x_error_info *info,
883 error32 = (info->ferr_global >> 18) & 0x3ff;
884 stat32 = (info->ferr_global >> 4) & 0x7ff;
887 global_error(1, error32, &error_found, handle_errors);
890 global_error(0, stat32, &error_found, handle_errors);
892 error32 = (info->nerr_global >> 18) & 0x3ff;
893 stat32 = (info->nerr_global >> 4) & 0x7ff;
896 global_error(1, error32, &error_found, handle_errors);
899 global_error(0, stat32, &error_found, handle_errors);
901 e752x_check_hub_interface(info, &error_found, handle_errors);
902 e752x_check_ns_interface(info, &error_found, handle_errors);
903 e752x_check_sysbus(info, &error_found, handle_errors);
904 e752x_check_membuf(info, &error_found, handle_errors);
905 e752x_check_dram(mci, info, &error_found, handle_errors);
909 static void e752x_check(struct mem_ctl_info *mci)
911 struct e752x_error_info info;
913 debugf3("%s()\n", __func__);
914 e752x_get_error_info(mci, &info);
915 e752x_process_error_info(mci, &info, 1);
918 /* Return 1 if dual channel mode is active. Else return 0. */
919 static inline int dual_channel_active(u16 ddrcsr)
921 return (((ddrcsr >> 12) & 3) == 3);
924 /* Remap csrow index numbers if map_type is "reverse"
926 static inline int remap_csrow_index(struct mem_ctl_info *mci, int index)
928 struct e752x_pvt *pvt = mci->pvt_info;
936 static void e752x_init_csrows(struct mem_ctl_info *mci, struct pci_dev *pdev,
939 struct csrow_info *csrow;
940 unsigned long last_cumul_size;
941 int index, mem_dev, drc_chan;
942 int drc_drbg; /* DRB granularity 0=64mb, 1=128mb */
943 int drc_ddim; /* DRAM Data Integrity Mode 0=none, 2=edac */
945 u32 dra, drc, cumul_size;
948 for (index = 0; index < 4; index++) {
950 pci_read_config_byte(pdev, E752X_DRA + index, &dra_reg);
951 dra |= dra_reg << (index * 8);
953 pci_read_config_dword(pdev, E752X_DRC, &drc);
954 drc_chan = dual_channel_active(ddrcsr);
955 drc_drbg = drc_chan + 1; /* 128 in dual mode, 64 in single */
956 drc_ddim = (drc >> 20) & 0x3;
958 /* The dram row boundary (DRB) reg values are boundary address for
959 * each DRAM row with a granularity of 64 or 128MB (single/dual
960 * channel operation). DRB regs are cumulative; therefore DRB7 will
961 * contain the total memory contained in all eight rows.
963 for (last_cumul_size = index = 0; index < mci->nr_csrows; index++) {
964 /* mem_dev 0=x8, 1=x4 */
965 mem_dev = (dra >> (index * 4 + 2)) & 0x3;
966 csrow = &mci->csrows[remap_csrow_index(mci, index)];
968 mem_dev = (mem_dev == 2);
969 pci_read_config_byte(pdev, E752X_DRB + index, &value);
970 /* convert a 128 or 64 MiB DRB to a page size. */
971 cumul_size = value << (25 + drc_drbg - PAGE_SHIFT);
972 debugf3("%s(): (%d) cumul_size 0x%x\n", __func__, index,
974 if (cumul_size == last_cumul_size)
975 continue; /* not populated */
977 csrow->first_page = last_cumul_size;
978 csrow->last_page = cumul_size - 1;
979 csrow->nr_pages = cumul_size - last_cumul_size;
980 last_cumul_size = cumul_size;
981 csrow->grain = 1 << 12; /* 4KiB - resolution of CELOG */
982 csrow->mtype = MEM_RDDR; /* only one type supported */
983 csrow->dtype = mem_dev ? DEV_X4 : DEV_X8;
986 * if single channel or x8 devices then SECDED
987 * if dual channel and x4 then S4ECD4ED
990 if (drc_chan && mem_dev) {
991 csrow->edac_mode = EDAC_S4ECD4ED;
992 mci->edac_cap |= EDAC_FLAG_S4ECD4ED;
994 csrow->edac_mode = EDAC_SECDED;
995 mci->edac_cap |= EDAC_FLAG_SECDED;
998 csrow->edac_mode = EDAC_NONE;
1002 static void e752x_init_mem_map_table(struct pci_dev *pdev,
1003 struct e752x_pvt *pvt)
1006 u8 value, last, row;
1011 for (index = 0; index < 8; index += 2) {
1012 pci_read_config_byte(pdev, E752X_DRB + index, &value);
1013 /* test if there is a dimm in this slot */
1014 if (value == last) {
1015 /* no dimm in the slot, so flag it as empty */
1016 pvt->map[index] = 0xff;
1017 pvt->map[index + 1] = 0xff;
1018 } else { /* there is a dimm in the slot */
1019 pvt->map[index] = row;
1022 /* test the next value to see if the dimm is double
1025 pci_read_config_byte(pdev, E752X_DRB + index + 1,
1028 /* the dimm is single sided, so flag as empty */
1029 /* this is a double sided dimm to save the next row #*/
1030 pvt->map[index + 1] = (value == last) ? 0xff : row;
1037 /* Return 0 on success or 1 on failure. */
1038 static int e752x_get_devs(struct pci_dev *pdev, int dev_idx,
1039 struct e752x_pvt *pvt)
1041 struct pci_dev *dev;
1043 pvt->bridge_ck = pci_get_device(PCI_VENDOR_ID_INTEL,
1044 pvt->dev_info->err_dev, pvt->bridge_ck);
1046 if (pvt->bridge_ck == NULL)
1047 pvt->bridge_ck = pci_scan_single_device(pdev->bus,
1050 if (pvt->bridge_ck == NULL) {
1051 e752x_printk(KERN_ERR, "error reporting device not found:"
1052 "vendor %x device 0x%x (broken BIOS?)\n",
1053 PCI_VENDOR_ID_INTEL, e752x_devs[dev_idx].err_dev);
1057 dev = pci_get_device(PCI_VENDOR_ID_INTEL,
1058 e752x_devs[dev_idx].ctl_dev,
1064 pvt->dev_d0f0 = dev;
1065 pvt->dev_d0f1 = pci_dev_get(pvt->bridge_ck);
1070 pci_dev_put(pvt->bridge_ck);
1074 /* Setup system bus parity mask register.
1075 * Sysbus parity supported on:
1076 * e7320/e7520/e7525 + Xeon
1077 * i3100 + Xeon/Celeron
1078 * Sysbus parity not supported on:
1079 * i3100 + Pentium M/Celeron M/Core Duo/Core2 Duo
1081 static void e752x_init_sysbus_parity_mask(struct e752x_pvt *pvt)
1083 char *cpu_id = cpu_data(0).x86_model_id;
1084 struct pci_dev *dev = pvt->dev_d0f1;
1087 /* Allow module paramter override, else see if CPU supports parity */
1088 if (sysbus_parity != -1) {
1089 enable = sysbus_parity;
1090 } else if (cpu_id[0] &&
1091 ((strstr(cpu_id, "Pentium") && strstr(cpu_id, " M ")) ||
1092 (strstr(cpu_id, "Celeron") && strstr(cpu_id, " M ")) ||
1093 (strstr(cpu_id, "Core") && strstr(cpu_id, "Duo")))) {
1094 e752x_printk(KERN_INFO, "System Bus Parity not "
1095 "supported by CPU, disabling\n");
1100 pci_write_config_word(dev, E752X_SYSBUS_ERRMASK, 0x0000);
1102 pci_write_config_word(dev, E752X_SYSBUS_ERRMASK, 0x0309);
1105 static void e752x_init_error_reporting_regs(struct e752x_pvt *pvt)
1107 struct pci_dev *dev;
1109 dev = pvt->dev_d0f1;
1110 /* Turn off error disable & SMI in case the BIOS turned it on */
1111 if (pvt->dev_info->err_dev == PCI_DEVICE_ID_INTEL_3100_1_ERR) {
1112 pci_write_config_dword(dev, I3100_NSI_EMASK, 0);
1113 pci_write_config_dword(dev, I3100_NSI_SMICMD, 0);
1115 pci_write_config_byte(dev, E752X_HI_ERRMASK, 0x00);
1116 pci_write_config_byte(dev, E752X_HI_SMICMD, 0x00);
1119 e752x_init_sysbus_parity_mask(pvt);
1121 pci_write_config_word(dev, E752X_SYSBUS_SMICMD, 0x00);
1122 pci_write_config_byte(dev, E752X_BUF_ERRMASK, 0x00);
1123 pci_write_config_byte(dev, E752X_BUF_SMICMD, 0x00);
1124 pci_write_config_byte(dev, E752X_DRAM_ERRMASK, 0x00);
1125 pci_write_config_byte(dev, E752X_DRAM_SMICMD, 0x00);
1128 static int e752x_probe1(struct pci_dev *pdev, int dev_idx)
1132 struct mem_ctl_info *mci;
1133 struct e752x_pvt *pvt;
1135 int drc_chan; /* Number of channels 0=1chan,1=2chan */
1136 struct e752x_error_info discard;
1138 debugf0("%s(): mci\n", __func__);
1139 debugf0("Starting Probe1\n");
1141 /* check to see if device 0 function 1 is enabled; if it isn't, we
1142 * assume the BIOS has reserved it for a reason and is expecting
1143 * exclusive access, we take care not to violate that assumption and
1144 * fail the probe. */
1145 pci_read_config_byte(pdev, E752X_DEVPRES1, &stat8);
1146 if (!force_function_unhide && !(stat8 & (1 << 5))) {
1147 printk(KERN_INFO "Contact your BIOS vendor to see if the "
1148 "E752x error registers can be safely un-hidden\n");
1152 pci_write_config_byte(pdev, E752X_DEVPRES1, stat8);
1154 pci_read_config_word(pdev, E752X_DDRCSR, &ddrcsr);
1155 /* FIXME: should check >>12 or 0xf, true for all? */
1156 /* Dual channel = 1, Single channel = 0 */
1157 drc_chan = dual_channel_active(ddrcsr);
1159 mci = edac_mc_alloc(sizeof(*pvt), E752X_NR_CSROWS, drc_chan + 1, 0);
1165 debugf3("%s(): init mci\n", __func__);
1166 mci->mtype_cap = MEM_FLAG_RDDR;
1167 /* 3100 IMCH supports SECDEC only */
1168 mci->edac_ctl_cap = (dev_idx == I3100) ? EDAC_FLAG_SECDED :
1169 (EDAC_FLAG_NONE | EDAC_FLAG_SECDED | EDAC_FLAG_S4ECD4ED);
1170 /* FIXME - what if different memory types are in different csrows? */
1171 mci->mod_name = EDAC_MOD_STR;
1172 mci->mod_ver = E752X_REVISION;
1173 mci->dev = &pdev->dev;
1175 debugf3("%s(): init pvt\n", __func__);
1176 pvt = (struct e752x_pvt *)mci->pvt_info;
1177 pvt->dev_info = &e752x_devs[dev_idx];
1178 pvt->mc_symmetric = ((ddrcsr & 0x10) != 0);
1180 if (e752x_get_devs(pdev, dev_idx, pvt)) {
1185 debugf3("%s(): more mci init\n", __func__);
1186 mci->ctl_name = pvt->dev_info->ctl_name;
1187 mci->dev_name = pci_name(pdev);
1188 mci->edac_check = e752x_check;
1189 mci->ctl_page_to_phys = ctl_page_to_phys;
1191 /* set the map type. 1 = normal, 0 = reversed
1192 * Must be set before e752x_init_csrows in case csrow mapping
1195 pci_read_config_byte(pdev, E752X_DRM, &stat8);
1196 pvt->map_type = ((stat8 & 0x0f) > ((stat8 >> 4) & 0x0f));
1198 e752x_init_csrows(mci, pdev, ddrcsr);
1199 e752x_init_mem_map_table(pdev, pvt);
1201 if (dev_idx == I3100)
1202 mci->edac_cap = EDAC_FLAG_SECDED; /* the only mode supported */
1204 mci->edac_cap |= EDAC_FLAG_NONE;
1205 debugf3("%s(): tolm, remapbase, remaplimit\n", __func__);
1207 /* load the top of low memory, remap base, and remap limit vars */
1208 pci_read_config_word(pdev, E752X_TOLM, &pci_data);
1209 pvt->tolm = ((u32) pci_data) << 4;
1210 pci_read_config_word(pdev, E752X_REMAPBASE, &pci_data);
1211 pvt->remapbase = ((u32) pci_data) << 14;
1212 pci_read_config_word(pdev, E752X_REMAPLIMIT, &pci_data);
1213 pvt->remaplimit = ((u32) pci_data) << 14;
1214 e752x_printk(KERN_INFO,
1215 "tolm = %x, remapbase = %x, remaplimit = %x\n",
1216 pvt->tolm, pvt->remapbase, pvt->remaplimit);
1218 /* Here we assume that we will never see multiple instances of this
1219 * type of memory controller. The ID is therefore hardcoded to 0.
1221 if (edac_mc_add_mc(mci)) {
1222 debugf3("%s(): failed edac_mc_add_mc()\n", __func__);
1226 e752x_init_error_reporting_regs(pvt);
1227 e752x_get_error_info(mci, &discard); /* clear other MCH errors */
1229 /* allocating generic PCI control info */
1230 e752x_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
1233 "%s(): Unable to create PCI control\n", __func__);
1235 "%s(): PCI error report via EDAC not setup\n",
1239 /* get this far and it's successful */
1240 debugf3("%s(): success\n", __func__);
1244 pci_dev_put(pvt->dev_d0f0);
1245 pci_dev_put(pvt->dev_d0f1);
1246 pci_dev_put(pvt->bridge_ck);
1252 /* returns count (>= 0), or negative on error */
1253 static int __devinit e752x_init_one(struct pci_dev *pdev,
1254 const struct pci_device_id *ent)
1256 debugf0("%s()\n", __func__);
1258 /* wake up and enable device */
1259 if (pci_enable_device(pdev) < 0)
1262 return e752x_probe1(pdev, ent->driver_data);
1265 static void __devexit e752x_remove_one(struct pci_dev *pdev)
1267 struct mem_ctl_info *mci;
1268 struct e752x_pvt *pvt;
1270 debugf0("%s()\n", __func__);
1273 edac_pci_release_generic_ctl(e752x_pci);
1275 if ((mci = edac_mc_del_mc(&pdev->dev)) == NULL)
1278 pvt = (struct e752x_pvt *)mci->pvt_info;
1279 pci_dev_put(pvt->dev_d0f0);
1280 pci_dev_put(pvt->dev_d0f1);
1281 pci_dev_put(pvt->bridge_ck);
1285 static const struct pci_device_id e752x_pci_tbl[] __devinitdata = {
1287 PCI_VEND_DEV(INTEL, 7520_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
1290 PCI_VEND_DEV(INTEL, 7525_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
1293 PCI_VEND_DEV(INTEL, 7320_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
1296 PCI_VEND_DEV(INTEL, 3100_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
1300 } /* 0 terminated list. */
1303 MODULE_DEVICE_TABLE(pci, e752x_pci_tbl);
1305 static struct pci_driver e752x_driver = {
1306 .name = EDAC_MOD_STR,
1307 .probe = e752x_init_one,
1308 .remove = __devexit_p(e752x_remove_one),
1309 .id_table = e752x_pci_tbl,
1312 static int __init e752x_init(void)
1316 debugf3("%s()\n", __func__);
1318 /* Ensure that the OPSTATE is set correctly for POLL or NMI */
1321 pci_rc = pci_register_driver(&e752x_driver);
1322 return (pci_rc < 0) ? pci_rc : 0;
1325 static void __exit e752x_exit(void)
1327 debugf3("%s()\n", __func__);
1328 pci_unregister_driver(&e752x_driver);
1331 module_init(e752x_init);
1332 module_exit(e752x_exit);
1334 MODULE_LICENSE("GPL");
1335 MODULE_AUTHOR("Linux Networx (http://lnxi.com) Tom Zimmerman\n");
1336 MODULE_DESCRIPTION("MC support for Intel e752x/3100 memory controllers");
1338 module_param(force_function_unhide, int, 0444);
1339 MODULE_PARM_DESC(force_function_unhide, "if BIOS sets Dev0:Fun1 up as hidden:"
1340 " 1=force unhide and hope BIOS doesn't fight driver for "
1341 "Dev0:Fun1 access");
1343 module_param(edac_op_state, int, 0444);
1344 MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
1346 module_param(sysbus_parity, int, 0444);
1347 MODULE_PARM_DESC(sysbus_parity, "0=disable system bus parity checking,"
1348 " 1=enable system bus parity checking, default=auto-detect");
1349 module_param(report_non_memory_errors, int, 0644);
1350 MODULE_PARM_DESC(report_non_memory_errors, "0=disable non-memory error "
1351 "reporting, 1=enable non-memory error reporting");