i5000_edac.c 42 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573
  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. * norsk5@xmission.com
  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. #include <linux/module.h>
  17. #include <linux/init.h>
  18. #include <linux/pci.h>
  19. #include <linux/pci_ids.h>
  20. #include <linux/slab.h>
  21. #include <linux/edac.h>
  22. #include <asm/mmzone.h>
  23. #include "edac_core.h"
  24. /*
  25. * Alter this version for the I5000 module when modifications are made
  26. */
  27. #define I5000_REVISION " Ver: 2.0.12 " __DATE__
  28. #define EDAC_MOD_STR "i5000_edac"
  29. #define i5000_printk(level, fmt, arg...) \
  30. edac_printk(level, "i5000", fmt, ##arg)
  31. #define i5000_mc_printk(mci, level, fmt, arg...) \
  32. edac_mc_chipset_printk(mci, level, "i5000", fmt, ##arg)
  33. #ifndef PCI_DEVICE_ID_INTEL_FBD_0
  34. #define PCI_DEVICE_ID_INTEL_FBD_0 0x25F5
  35. #endif
  36. #ifndef PCI_DEVICE_ID_INTEL_FBD_1
  37. #define PCI_DEVICE_ID_INTEL_FBD_1 0x25F6
  38. #endif
  39. /* Device 16,
  40. * Function 0: System Address
  41. * Function 1: Memory Branch Map, Control, Errors Register
  42. * Function 2: FSB Error Registers
  43. *
  44. * All 3 functions of Device 16 (0,1,2) share the SAME DID
  45. */
  46. #define PCI_DEVICE_ID_INTEL_I5000_DEV16 0x25F0
  47. /* OFFSETS for Function 0 */
  48. /* OFFSETS for Function 1 */
  49. #define AMBASE 0x48
  50. #define MAXCH 0x56
  51. #define MAXDIMMPERCH 0x57
  52. #define TOLM 0x6C
  53. #define REDMEMB 0x7C
  54. #define RED_ECC_LOCATOR(x) ((x) & 0x3FFFF)
  55. #define REC_ECC_LOCATOR_EVEN(x) ((x) & 0x001FF)
  56. #define REC_ECC_LOCATOR_ODD(x) ((x) & 0x3FE00)
  57. #define MIR0 0x80
  58. #define MIR1 0x84
  59. #define MIR2 0x88
  60. #define AMIR0 0x8C
  61. #define AMIR1 0x90
  62. #define AMIR2 0x94
  63. #define FERR_FAT_FBD 0x98
  64. #define NERR_FAT_FBD 0x9C
  65. #define EXTRACT_FBDCHAN_INDX(x) (((x)>>28) & 0x3)
  66. #define FERR_FAT_FBDCHAN 0x30000000
  67. #define FERR_FAT_M3ERR 0x00000004
  68. #define FERR_FAT_M2ERR 0x00000002
  69. #define FERR_FAT_M1ERR 0x00000001
  70. #define FERR_FAT_MASK (FERR_FAT_M1ERR | \
  71. FERR_FAT_M2ERR | \
  72. FERR_FAT_M3ERR)
  73. #define FERR_NF_FBD 0xA0
  74. /* Thermal and SPD or BFD errors */
  75. #define FERR_NF_M28ERR 0x01000000
  76. #define FERR_NF_M27ERR 0x00800000
  77. #define FERR_NF_M26ERR 0x00400000
  78. #define FERR_NF_M25ERR 0x00200000
  79. #define FERR_NF_M24ERR 0x00100000
  80. #define FERR_NF_M23ERR 0x00080000
  81. #define FERR_NF_M22ERR 0x00040000
  82. #define FERR_NF_M21ERR 0x00020000
  83. /* Correctable errors */
  84. #define FERR_NF_M20ERR 0x00010000
  85. #define FERR_NF_M19ERR 0x00008000
  86. #define FERR_NF_M18ERR 0x00004000
  87. #define FERR_NF_M17ERR 0x00002000
  88. /* Non-Retry or redundant Retry errors */
  89. #define FERR_NF_M16ERR 0x00001000
  90. #define FERR_NF_M15ERR 0x00000800
  91. #define FERR_NF_M14ERR 0x00000400
  92. #define FERR_NF_M13ERR 0x00000200
  93. /* Uncorrectable errors */
  94. #define FERR_NF_M12ERR 0x00000100
  95. #define FERR_NF_M11ERR 0x00000080
  96. #define FERR_NF_M10ERR 0x00000040
  97. #define FERR_NF_M9ERR 0x00000020
  98. #define FERR_NF_M8ERR 0x00000010
  99. #define FERR_NF_M7ERR 0x00000008
  100. #define FERR_NF_M6ERR 0x00000004
  101. #define FERR_NF_M5ERR 0x00000002
  102. #define FERR_NF_M4ERR 0x00000001
  103. #define FERR_NF_UNCORRECTABLE (FERR_NF_M12ERR | \
  104. FERR_NF_M11ERR | \
  105. FERR_NF_M10ERR | \
  106. FERR_NF_M9ERR | \
  107. FERR_NF_M8ERR | \
  108. FERR_NF_M7ERR | \
  109. FERR_NF_M6ERR | \
  110. FERR_NF_M5ERR | \
  111. FERR_NF_M4ERR)
  112. #define FERR_NF_CORRECTABLE (FERR_NF_M20ERR | \
  113. FERR_NF_M19ERR | \
  114. FERR_NF_M18ERR | \
  115. FERR_NF_M17ERR)
  116. #define FERR_NF_DIMM_SPARE (FERR_NF_M27ERR | \
  117. FERR_NF_M28ERR)
  118. #define FERR_NF_THERMAL (FERR_NF_M26ERR | \
  119. FERR_NF_M25ERR | \
  120. FERR_NF_M24ERR | \
  121. FERR_NF_M23ERR)
  122. #define FERR_NF_SPD_PROTOCOL (FERR_NF_M22ERR)
  123. #define FERR_NF_NORTH_CRC (FERR_NF_M21ERR)
  124. #define FERR_NF_NON_RETRY (FERR_NF_M13ERR | \
  125. FERR_NF_M14ERR | \
  126. FERR_NF_M15ERR)
  127. #define NERR_NF_FBD 0xA4
  128. #define FERR_NF_MASK (FERR_NF_UNCORRECTABLE | \
  129. FERR_NF_CORRECTABLE | \
  130. FERR_NF_DIMM_SPARE | \
  131. FERR_NF_THERMAL | \
  132. FERR_NF_SPD_PROTOCOL | \
  133. FERR_NF_NORTH_CRC | \
  134. FERR_NF_NON_RETRY)
  135. #define EMASK_FBD 0xA8
  136. #define EMASK_FBD_M28ERR 0x08000000
  137. #define EMASK_FBD_M27ERR 0x04000000
  138. #define EMASK_FBD_M26ERR 0x02000000
  139. #define EMASK_FBD_M25ERR 0x01000000
  140. #define EMASK_FBD_M24ERR 0x00800000
  141. #define EMASK_FBD_M23ERR 0x00400000
  142. #define EMASK_FBD_M22ERR 0x00200000
  143. #define EMASK_FBD_M21ERR 0x00100000
  144. #define EMASK_FBD_M20ERR 0x00080000
  145. #define EMASK_FBD_M19ERR 0x00040000
  146. #define EMASK_FBD_M18ERR 0x00020000
  147. #define EMASK_FBD_M17ERR 0x00010000
  148. #define EMASK_FBD_M15ERR 0x00004000
  149. #define EMASK_FBD_M14ERR 0x00002000
  150. #define EMASK_FBD_M13ERR 0x00001000
  151. #define EMASK_FBD_M12ERR 0x00000800
  152. #define EMASK_FBD_M11ERR 0x00000400
  153. #define EMASK_FBD_M10ERR 0x00000200
  154. #define EMASK_FBD_M9ERR 0x00000100
  155. #define EMASK_FBD_M8ERR 0x00000080
  156. #define EMASK_FBD_M7ERR 0x00000040
  157. #define EMASK_FBD_M6ERR 0x00000020
  158. #define EMASK_FBD_M5ERR 0x00000010
  159. #define EMASK_FBD_M4ERR 0x00000008
  160. #define EMASK_FBD_M3ERR 0x00000004
  161. #define EMASK_FBD_M2ERR 0x00000002
  162. #define EMASK_FBD_M1ERR 0x00000001
  163. #define ENABLE_EMASK_FBD_FATAL_ERRORS (EMASK_FBD_M1ERR | \
  164. EMASK_FBD_M2ERR | \
  165. EMASK_FBD_M3ERR)
  166. #define ENABLE_EMASK_FBD_UNCORRECTABLE (EMASK_FBD_M4ERR | \
  167. EMASK_FBD_M5ERR | \
  168. EMASK_FBD_M6ERR | \
  169. EMASK_FBD_M7ERR | \
  170. EMASK_FBD_M8ERR | \
  171. EMASK_FBD_M9ERR | \
  172. EMASK_FBD_M10ERR | \
  173. EMASK_FBD_M11ERR | \
  174. EMASK_FBD_M12ERR)
  175. #define ENABLE_EMASK_FBD_CORRECTABLE (EMASK_FBD_M17ERR | \
  176. EMASK_FBD_M18ERR | \
  177. EMASK_FBD_M19ERR | \
  178. EMASK_FBD_M20ERR)
  179. #define ENABLE_EMASK_FBD_DIMM_SPARE (EMASK_FBD_M27ERR | \
  180. EMASK_FBD_M28ERR)
  181. #define ENABLE_EMASK_FBD_THERMALS (EMASK_FBD_M26ERR | \
  182. EMASK_FBD_M25ERR | \
  183. EMASK_FBD_M24ERR | \
  184. EMASK_FBD_M23ERR)
  185. #define ENABLE_EMASK_FBD_SPD_PROTOCOL (EMASK_FBD_M22ERR)
  186. #define ENABLE_EMASK_FBD_NORTH_CRC (EMASK_FBD_M21ERR)
  187. #define ENABLE_EMASK_FBD_NON_RETRY (EMASK_FBD_M15ERR | \
  188. EMASK_FBD_M14ERR | \
  189. EMASK_FBD_M13ERR)
  190. #define ENABLE_EMASK_ALL (ENABLE_EMASK_FBD_NON_RETRY | \
  191. ENABLE_EMASK_FBD_NORTH_CRC | \
  192. ENABLE_EMASK_FBD_SPD_PROTOCOL | \
  193. ENABLE_EMASK_FBD_THERMALS | \
  194. ENABLE_EMASK_FBD_DIMM_SPARE | \
  195. ENABLE_EMASK_FBD_FATAL_ERRORS | \
  196. ENABLE_EMASK_FBD_CORRECTABLE | \
  197. ENABLE_EMASK_FBD_UNCORRECTABLE)
  198. #define ERR0_FBD 0xAC
  199. #define ERR1_FBD 0xB0
  200. #define ERR2_FBD 0xB4
  201. #define MCERR_FBD 0xB8
  202. #define NRECMEMA 0xBE
  203. #define NREC_BANK(x) (((x)>>12) & 0x7)
  204. #define NREC_RDWR(x) (((x)>>11) & 1)
  205. #define NREC_RANK(x) (((x)>>8) & 0x7)
  206. #define NRECMEMB 0xC0
  207. #define NREC_CAS(x) (((x)>>16) & 0xFFFFFF)
  208. #define NREC_RAS(x) ((x) & 0x7FFF)
  209. #define NRECFGLOG 0xC4
  210. #define NREEECFBDA 0xC8
  211. #define NREEECFBDB 0xCC
  212. #define NREEECFBDC 0xD0
  213. #define NREEECFBDD 0xD4
  214. #define NREEECFBDE 0xD8
  215. #define REDMEMA 0xDC
  216. #define RECMEMA 0xE2
  217. #define REC_BANK(x) (((x)>>12) & 0x7)
  218. #define REC_RDWR(x) (((x)>>11) & 1)
  219. #define REC_RANK(x) (((x)>>8) & 0x7)
  220. #define RECMEMB 0xE4
  221. #define REC_CAS(x) (((x)>>16) & 0xFFFFFF)
  222. #define REC_RAS(x) ((x) & 0x7FFF)
  223. #define RECFGLOG 0xE8
  224. #define RECFBDA 0xEC
  225. #define RECFBDB 0xF0
  226. #define RECFBDC 0xF4
  227. #define RECFBDD 0xF8
  228. #define RECFBDE 0xFC
  229. /* OFFSETS for Function 2 */
  230. /*
  231. * Device 21,
  232. * Function 0: Memory Map Branch 0
  233. *
  234. * Device 22,
  235. * Function 0: Memory Map Branch 1
  236. */
  237. #define PCI_DEVICE_ID_I5000_BRANCH_0 0x25F5
  238. #define PCI_DEVICE_ID_I5000_BRANCH_1 0x25F6
  239. #define AMB_PRESENT_0 0x64
  240. #define AMB_PRESENT_1 0x66
  241. #define MTR0 0x80
  242. #define MTR1 0x84
  243. #define MTR2 0x88
  244. #define MTR3 0x8C
  245. #define NUM_MTRS 4
  246. #define CHANNELS_PER_BRANCH (2)
  247. /* Defines to extract the vaious fields from the
  248. * MTRx - Memory Technology Registers
  249. */
  250. #define MTR_DIMMS_PRESENT(mtr) ((mtr) & (0x1 << 8))
  251. #define MTR_DRAM_WIDTH(mtr) ((((mtr) >> 6) & 0x1) ? 8 : 4)
  252. #define MTR_DRAM_BANKS(mtr) ((((mtr) >> 5) & 0x1) ? 8 : 4)
  253. #define MTR_DRAM_BANKS_ADDR_BITS(mtr) ((MTR_DRAM_BANKS(mtr) == 8) ? 3 : 2)
  254. #define MTR_DIMM_RANK(mtr) (((mtr) >> 4) & 0x1)
  255. #define MTR_DIMM_RANK_ADDR_BITS(mtr) (MTR_DIMM_RANK(mtr) ? 2 : 1)
  256. #define MTR_DIMM_ROWS(mtr) (((mtr) >> 2) & 0x3)
  257. #define MTR_DIMM_ROWS_ADDR_BITS(mtr) (MTR_DIMM_ROWS(mtr) + 13)
  258. #define MTR_DIMM_COLS(mtr) ((mtr) & 0x3)
  259. #define MTR_DIMM_COLS_ADDR_BITS(mtr) (MTR_DIMM_COLS(mtr) + 10)
  260. #ifdef CONFIG_EDAC_DEBUG
  261. static char *numrow_toString[] = {
  262. "8,192 - 13 rows",
  263. "16,384 - 14 rows",
  264. "32,768 - 15 rows",
  265. "reserved"
  266. };
  267. static char *numcol_toString[] = {
  268. "1,024 - 10 columns",
  269. "2,048 - 11 columns",
  270. "4,096 - 12 columns",
  271. "reserved"
  272. };
  273. #endif
  274. /* enables the report of miscellaneous messages as CE errors - default off */
  275. static int misc_messages;
  276. /* Enumeration of supported devices */
  277. enum i5000_chips {
  278. I5000P = 0,
  279. I5000V = 1, /* future */
  280. I5000X = 2 /* future */
  281. };
  282. /* Device name and register DID (Device ID) */
  283. struct i5000_dev_info {
  284. const char *ctl_name; /* name for this device */
  285. u16 fsb_mapping_errors; /* DID for the branchmap,control */
  286. };
  287. /* Table of devices attributes supported by this driver */
  288. static const struct i5000_dev_info i5000_devs[] = {
  289. [I5000P] = {
  290. .ctl_name = "I5000",
  291. .fsb_mapping_errors = PCI_DEVICE_ID_INTEL_I5000_DEV16,
  292. },
  293. };
  294. struct i5000_dimm_info {
  295. int megabytes; /* size, 0 means not present */
  296. int dual_rank;
  297. };
  298. #define MAX_CHANNELS 6 /* max possible channels */
  299. #define MAX_CSROWS (8*2) /* max possible csrows per channel */
  300. /* driver private data structure */
  301. struct i5000_pvt {
  302. struct pci_dev *system_address; /* 16.0 */
  303. struct pci_dev *branchmap_werrors; /* 16.1 */
  304. struct pci_dev *fsb_error_regs; /* 16.2 */
  305. struct pci_dev *branch_0; /* 21.0 */
  306. struct pci_dev *branch_1; /* 22.0 */
  307. u16 tolm; /* top of low memory */
  308. u64 ambase; /* AMB BAR */
  309. u16 mir0, mir1, mir2;
  310. u16 b0_mtr[NUM_MTRS]; /* Memory Technlogy Reg */
  311. u16 b0_ambpresent0; /* Branch 0, Channel 0 */
  312. u16 b0_ambpresent1; /* Brnach 0, Channel 1 */
  313. u16 b1_mtr[NUM_MTRS]; /* Memory Technlogy Reg */
  314. u16 b1_ambpresent0; /* Branch 1, Channel 8 */
  315. u16 b1_ambpresent1; /* Branch 1, Channel 1 */
  316. /* DIMM information matrix, allocating architecture maximums */
  317. struct i5000_dimm_info dimm_info[MAX_CSROWS][MAX_CHANNELS];
  318. /* Actual values for this controller */
  319. int maxch; /* Max channels */
  320. int maxdimmperch; /* Max DIMMs per channel */
  321. };
  322. /* I5000 MCH error information retrieved from Hardware */
  323. struct i5000_error_info {
  324. /* These registers are always read from the MC */
  325. u32 ferr_fat_fbd; /* First Errors Fatal */
  326. u32 nerr_fat_fbd; /* Next Errors Fatal */
  327. u32 ferr_nf_fbd; /* First Errors Non-Fatal */
  328. u32 nerr_nf_fbd; /* Next Errors Non-Fatal */
  329. /* These registers are input ONLY if there was a Recoverable Error */
  330. u32 redmemb; /* Recoverable Mem Data Error log B */
  331. u16 recmema; /* Recoverable Mem Error log A */
  332. u32 recmemb; /* Recoverable Mem Error log B */
  333. /* These registers are input ONLY if there was a
  334. * Non-Recoverable Error */
  335. u16 nrecmema; /* Non-Recoverable Mem log A */
  336. u16 nrecmemb; /* Non-Recoverable Mem log B */
  337. };
  338. static struct edac_pci_ctl_info *i5000_pci;
  339. /*
  340. * i5000_get_error_info Retrieve the hardware error information from
  341. * the hardware and cache it in the 'info'
  342. * structure
  343. */
  344. static void i5000_get_error_info(struct mem_ctl_info *mci,
  345. struct i5000_error_info *info)
  346. {
  347. struct i5000_pvt *pvt;
  348. u32 value;
  349. pvt = mci->pvt_info;
  350. /* read in the 1st FATAL error register */
  351. pci_read_config_dword(pvt->branchmap_werrors, FERR_FAT_FBD, &value);
  352. /* Mask only the bits that the doc says are valid
  353. */
  354. value &= (FERR_FAT_FBDCHAN | FERR_FAT_MASK);
  355. /* If there is an error, then read in the */
  356. /* NEXT FATAL error register and the Memory Error Log Register A */
  357. if (value & FERR_FAT_MASK) {
  358. info->ferr_fat_fbd = value;
  359. /* harvest the various error data we need */
  360. pci_read_config_dword(pvt->branchmap_werrors,
  361. NERR_FAT_FBD, &info->nerr_fat_fbd);
  362. pci_read_config_word(pvt->branchmap_werrors,
  363. NRECMEMA, &info->nrecmema);
  364. pci_read_config_word(pvt->branchmap_werrors,
  365. NRECMEMB, &info->nrecmemb);
  366. /* Clear the error bits, by writing them back */
  367. pci_write_config_dword(pvt->branchmap_werrors,
  368. FERR_FAT_FBD, value);
  369. } else {
  370. info->ferr_fat_fbd = 0;
  371. info->nerr_fat_fbd = 0;
  372. info->nrecmema = 0;
  373. info->nrecmemb = 0;
  374. }
  375. /* read in the 1st NON-FATAL error register */
  376. pci_read_config_dword(pvt->branchmap_werrors, FERR_NF_FBD, &value);
  377. /* If there is an error, then read in the 1st NON-FATAL error
  378. * register as well */
  379. if (value & FERR_NF_MASK) {
  380. info->ferr_nf_fbd = value;
  381. /* harvest the various error data we need */
  382. pci_read_config_dword(pvt->branchmap_werrors,
  383. NERR_NF_FBD, &info->nerr_nf_fbd);
  384. pci_read_config_word(pvt->branchmap_werrors,
  385. RECMEMA, &info->recmema);
  386. pci_read_config_dword(pvt->branchmap_werrors,
  387. RECMEMB, &info->recmemb);
  388. pci_read_config_dword(pvt->branchmap_werrors,
  389. REDMEMB, &info->redmemb);
  390. /* Clear the error bits, by writing them back */
  391. pci_write_config_dword(pvt->branchmap_werrors,
  392. FERR_NF_FBD, value);
  393. } else {
  394. info->ferr_nf_fbd = 0;
  395. info->nerr_nf_fbd = 0;
  396. info->recmema = 0;
  397. info->recmemb = 0;
  398. info->redmemb = 0;
  399. }
  400. }
  401. /*
  402. * i5000_process_fatal_error_info(struct mem_ctl_info *mci,
  403. * struct i5000_error_info *info,
  404. * int handle_errors);
  405. *
  406. * handle the Intel FATAL errors, if any
  407. */
  408. static void i5000_process_fatal_error_info(struct mem_ctl_info *mci,
  409. struct i5000_error_info *info,
  410. int handle_errors)
  411. {
  412. char msg[EDAC_MC_LABEL_LEN + 1 + 160];
  413. char *specific = NULL;
  414. u32 allErrors;
  415. int branch;
  416. int channel;
  417. int bank;
  418. int rank;
  419. int rdwr;
  420. int ras, cas;
  421. /* mask off the Error bits that are possible */
  422. allErrors = (info->ferr_fat_fbd & FERR_FAT_MASK);
  423. if (!allErrors)
  424. return; /* if no error, return now */
  425. branch = EXTRACT_FBDCHAN_INDX(info->ferr_fat_fbd);
  426. channel = branch;
  427. /* Use the NON-Recoverable macros to extract data */
  428. bank = NREC_BANK(info->nrecmema);
  429. rank = NREC_RANK(info->nrecmema);
  430. rdwr = NREC_RDWR(info->nrecmema);
  431. ras = NREC_RAS(info->nrecmemb);
  432. cas = NREC_CAS(info->nrecmemb);
  433. debugf0("\t\tCSROW= %d Channels= %d,%d (Branch= %d "
  434. "DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
  435. rank, channel, channel + 1, branch >> 1, bank,
  436. rdwr ? "Write" : "Read", ras, cas);
  437. /* Only 1 bit will be on */
  438. switch (allErrors) {
  439. case FERR_FAT_M1ERR:
  440. specific = "Alert on non-redundant retry or fast "
  441. "reset timeout";
  442. break;
  443. case FERR_FAT_M2ERR:
  444. specific = "Northbound CRC error on non-redundant "
  445. "retry";
  446. break;
  447. case FERR_FAT_M3ERR:
  448. {
  449. static int done;
  450. /*
  451. * This error is generated to inform that the intelligent
  452. * throttling is disabled and the temperature passed the
  453. * specified middle point. Since this is something the BIOS
  454. * should take care of, we'll warn only once to avoid
  455. * worthlessly flooding the log.
  456. */
  457. if (done)
  458. return;
  459. done++;
  460. specific = ">Tmid Thermal event with intelligent "
  461. "throttling disabled";
  462. }
  463. break;
  464. }
  465. /* Form out message */
  466. snprintf(msg, sizeof(msg),
  467. "(Branch=%d DRAM-Bank=%d RDWR=%s RAS=%d CAS=%d "
  468. "FATAL Err=0x%x (%s))",
  469. branch >> 1, bank, rdwr ? "Write" : "Read", ras, cas,
  470. allErrors, specific);
  471. /* Call the helper to output message */
  472. edac_mc_handle_fbd_ue(mci, rank, channel, channel + 1, msg);
  473. }
  474. /*
  475. * i5000_process_fatal_error_info(struct mem_ctl_info *mci,
  476. * struct i5000_error_info *info,
  477. * int handle_errors);
  478. *
  479. * handle the Intel NON-FATAL errors, if any
  480. */
  481. static void i5000_process_nonfatal_error_info(struct mem_ctl_info *mci,
  482. struct i5000_error_info *info,
  483. int handle_errors)
  484. {
  485. char msg[EDAC_MC_LABEL_LEN + 1 + 170];
  486. char *specific = NULL;
  487. u32 allErrors;
  488. u32 ue_errors;
  489. u32 ce_errors;
  490. u32 misc_errors;
  491. int branch;
  492. int channel;
  493. int bank;
  494. int rank;
  495. int rdwr;
  496. int ras, cas;
  497. /* mask off the Error bits that are possible */
  498. allErrors = (info->ferr_nf_fbd & FERR_NF_MASK);
  499. if (!allErrors)
  500. return; /* if no error, return now */
  501. /* ONLY ONE of the possible error bits will be set, as per the docs */
  502. ue_errors = allErrors & FERR_NF_UNCORRECTABLE;
  503. if (ue_errors) {
  504. debugf0("\tUncorrected bits= 0x%x\n", ue_errors);
  505. branch = EXTRACT_FBDCHAN_INDX(info->ferr_nf_fbd);
  506. channel = branch;
  507. bank = NREC_BANK(info->nrecmema);
  508. rank = NREC_RANK(info->nrecmema);
  509. rdwr = NREC_RDWR(info->nrecmema);
  510. ras = NREC_RAS(info->nrecmemb);
  511. cas = NREC_CAS(info->nrecmemb);
  512. debugf0
  513. ("\t\tCSROW= %d Channels= %d,%d (Branch= %d "
  514. "DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
  515. rank, channel, channel + 1, branch >> 1, bank,
  516. rdwr ? "Write" : "Read", ras, cas);
  517. switch (ue_errors) {
  518. case FERR_NF_M12ERR:
  519. specific = "Non-Aliased Uncorrectable Patrol Data ECC";
  520. break;
  521. case FERR_NF_M11ERR:
  522. specific = "Non-Aliased Uncorrectable Spare-Copy "
  523. "Data ECC";
  524. break;
  525. case FERR_NF_M10ERR:
  526. specific = "Non-Aliased Uncorrectable Mirrored Demand "
  527. "Data ECC";
  528. break;
  529. case FERR_NF_M9ERR:
  530. specific = "Non-Aliased Uncorrectable Non-Mirrored "
  531. "Demand Data ECC";
  532. break;
  533. case FERR_NF_M8ERR:
  534. specific = "Aliased Uncorrectable Patrol Data ECC";
  535. break;
  536. case FERR_NF_M7ERR:
  537. specific = "Aliased Uncorrectable Spare-Copy Data ECC";
  538. break;
  539. case FERR_NF_M6ERR:
  540. specific = "Aliased Uncorrectable Mirrored Demand "
  541. "Data ECC";
  542. break;
  543. case FERR_NF_M5ERR:
  544. specific = "Aliased Uncorrectable Non-Mirrored Demand "
  545. "Data ECC";
  546. break;
  547. case FERR_NF_M4ERR:
  548. specific = "Uncorrectable Data ECC on Replay";
  549. break;
  550. }
  551. /* Form out message */
  552. snprintf(msg, sizeof(msg),
  553. "(Branch=%d DRAM-Bank=%d RDWR=%s RAS=%d "
  554. "CAS=%d, UE Err=0x%x (%s))",
  555. branch >> 1, bank, rdwr ? "Write" : "Read", ras, cas,
  556. ue_errors, specific);
  557. /* Call the helper to output message */
  558. edac_mc_handle_fbd_ue(mci, rank, channel, channel + 1, msg);
  559. }
  560. /* Check correctable errors */
  561. ce_errors = allErrors & FERR_NF_CORRECTABLE;
  562. if (ce_errors) {
  563. debugf0("\tCorrected bits= 0x%x\n", ce_errors);
  564. branch = EXTRACT_FBDCHAN_INDX(info->ferr_nf_fbd);
  565. channel = 0;
  566. if (REC_ECC_LOCATOR_ODD(info->redmemb))
  567. channel = 1;
  568. /* Convert channel to be based from zero, instead of
  569. * from branch base of 0 */
  570. channel += branch;
  571. bank = REC_BANK(info->recmema);
  572. rank = REC_RANK(info->recmema);
  573. rdwr = REC_RDWR(info->recmema);
  574. ras = REC_RAS(info->recmemb);
  575. cas = REC_CAS(info->recmemb);
  576. debugf0("\t\tCSROW= %d Channel= %d (Branch %d "
  577. "DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
  578. rank, channel, branch >> 1, bank,
  579. rdwr ? "Write" : "Read", ras, cas);
  580. switch (ce_errors) {
  581. case FERR_NF_M17ERR:
  582. specific = "Correctable Non-Mirrored Demand Data ECC";
  583. break;
  584. case FERR_NF_M18ERR:
  585. specific = "Correctable Mirrored Demand Data ECC";
  586. break;
  587. case FERR_NF_M19ERR:
  588. specific = "Correctable Spare-Copy Data ECC";
  589. break;
  590. case FERR_NF_M20ERR:
  591. specific = "Correctable Patrol Data ECC";
  592. break;
  593. }
  594. /* Form out message */
  595. snprintf(msg, sizeof(msg),
  596. "(Branch=%d DRAM-Bank=%d RDWR=%s RAS=%d "
  597. "CAS=%d, CE Err=0x%x (%s))", branch >> 1, bank,
  598. rdwr ? "Write" : "Read", ras, cas, ce_errors,
  599. specific);
  600. /* Call the helper to output message */
  601. edac_mc_handle_fbd_ce(mci, rank, channel, msg);
  602. }
  603. if (!misc_messages)
  604. return;
  605. misc_errors = allErrors & (FERR_NF_NON_RETRY | FERR_NF_NORTH_CRC |
  606. FERR_NF_SPD_PROTOCOL | FERR_NF_DIMM_SPARE);
  607. if (misc_errors) {
  608. switch (misc_errors) {
  609. case FERR_NF_M13ERR:
  610. specific = "Non-Retry or Redundant Retry FBD Memory "
  611. "Alert or Redundant Fast Reset Timeout";
  612. break;
  613. case FERR_NF_M14ERR:
  614. specific = "Non-Retry or Redundant Retry FBD "
  615. "Configuration Alert";
  616. break;
  617. case FERR_NF_M15ERR:
  618. specific = "Non-Retry or Redundant Retry FBD "
  619. "Northbound CRC error on read data";
  620. break;
  621. case FERR_NF_M21ERR:
  622. specific = "FBD Northbound CRC error on "
  623. "FBD Sync Status";
  624. break;
  625. case FERR_NF_M22ERR:
  626. specific = "SPD protocol error";
  627. break;
  628. case FERR_NF_M27ERR:
  629. specific = "DIMM-spare copy started";
  630. break;
  631. case FERR_NF_M28ERR:
  632. specific = "DIMM-spare copy completed";
  633. break;
  634. }
  635. branch = EXTRACT_FBDCHAN_INDX(info->ferr_nf_fbd);
  636. /* Form out message */
  637. snprintf(msg, sizeof(msg),
  638. "(Branch=%d Err=%#x (%s))", branch >> 1,
  639. misc_errors, specific);
  640. /* Call the helper to output message */
  641. edac_mc_handle_fbd_ce(mci, 0, 0, msg);
  642. }
  643. }
  644. /*
  645. * i5000_process_error_info Process the error info that is
  646. * in the 'info' structure, previously retrieved from hardware
  647. */
  648. static void i5000_process_error_info(struct mem_ctl_info *mci,
  649. struct i5000_error_info *info,
  650. int handle_errors)
  651. {
  652. /* First handle any fatal errors that occurred */
  653. i5000_process_fatal_error_info(mci, info, handle_errors);
  654. /* now handle any non-fatal errors that occurred */
  655. i5000_process_nonfatal_error_info(mci, info, handle_errors);
  656. }
  657. /*
  658. * i5000_clear_error Retrieve any error from the hardware
  659. * but do NOT process that error.
  660. * Used for 'clearing' out of previous errors
  661. * Called by the Core module.
  662. */
  663. static void i5000_clear_error(struct mem_ctl_info *mci)
  664. {
  665. struct i5000_error_info info;
  666. i5000_get_error_info(mci, &info);
  667. }
  668. /*
  669. * i5000_check_error Retrieve and process errors reported by the
  670. * hardware. Called by the Core module.
  671. */
  672. static void i5000_check_error(struct mem_ctl_info *mci)
  673. {
  674. struct i5000_error_info info;
  675. debugf4("MC%d: " __FILE__ ": %s()\n", mci->mc_idx, __func__);
  676. i5000_get_error_info(mci, &info);
  677. i5000_process_error_info(mci, &info, 1);
  678. }
  679. /*
  680. * i5000_get_devices Find and perform 'get' operation on the MCH's
  681. * device/functions we want to reference for this driver
  682. *
  683. * Need to 'get' device 16 func 1 and func 2
  684. */
  685. static int i5000_get_devices(struct mem_ctl_info *mci, int dev_idx)
  686. {
  687. //const struct i5000_dev_info *i5000_dev = &i5000_devs[dev_idx];
  688. struct i5000_pvt *pvt;
  689. struct pci_dev *pdev;
  690. pvt = mci->pvt_info;
  691. /* Attempt to 'get' the MCH register we want */
  692. pdev = NULL;
  693. while (1) {
  694. pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
  695. PCI_DEVICE_ID_INTEL_I5000_DEV16, pdev);
  696. /* End of list, leave */
  697. if (pdev == NULL) {
  698. i5000_printk(KERN_ERR,
  699. "'system address,Process Bus' "
  700. "device not found:"
  701. "vendor 0x%x device 0x%x FUNC 1 "
  702. "(broken BIOS?)\n",
  703. PCI_VENDOR_ID_INTEL,
  704. PCI_DEVICE_ID_INTEL_I5000_DEV16);
  705. return 1;
  706. }
  707. /* Scan for device 16 func 1 */
  708. if (PCI_FUNC(pdev->devfn) == 1)
  709. break;
  710. }
  711. pvt->branchmap_werrors = pdev;
  712. /* Attempt to 'get' the MCH register we want */
  713. pdev = NULL;
  714. while (1) {
  715. pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
  716. PCI_DEVICE_ID_INTEL_I5000_DEV16, pdev);
  717. if (pdev == NULL) {
  718. i5000_printk(KERN_ERR,
  719. "MC: 'branchmap,control,errors' "
  720. "device not found:"
  721. "vendor 0x%x device 0x%x Func 2 "
  722. "(broken BIOS?)\n",
  723. PCI_VENDOR_ID_INTEL,
  724. PCI_DEVICE_ID_INTEL_I5000_DEV16);
  725. pci_dev_put(pvt->branchmap_werrors);
  726. return 1;
  727. }
  728. /* Scan for device 16 func 1 */
  729. if (PCI_FUNC(pdev->devfn) == 2)
  730. break;
  731. }
  732. pvt->fsb_error_regs = pdev;
  733. debugf1("System Address, processor bus- PCI Bus ID: %s %x:%x\n",
  734. pci_name(pvt->system_address),
  735. pvt->system_address->vendor, pvt->system_address->device);
  736. debugf1("Branchmap, control and errors - PCI Bus ID: %s %x:%x\n",
  737. pci_name(pvt->branchmap_werrors),
  738. pvt->branchmap_werrors->vendor, pvt->branchmap_werrors->device);
  739. debugf1("FSB Error Regs - PCI Bus ID: %s %x:%x\n",
  740. pci_name(pvt->fsb_error_regs),
  741. pvt->fsb_error_regs->vendor, pvt->fsb_error_regs->device);
  742. pdev = NULL;
  743. pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
  744. PCI_DEVICE_ID_I5000_BRANCH_0, pdev);
  745. if (pdev == NULL) {
  746. i5000_printk(KERN_ERR,
  747. "MC: 'BRANCH 0' device not found:"
  748. "vendor 0x%x device 0x%x Func 0 (broken BIOS?)\n",
  749. PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_I5000_BRANCH_0);
  750. pci_dev_put(pvt->branchmap_werrors);
  751. pci_dev_put(pvt->fsb_error_regs);
  752. return 1;
  753. }
  754. pvt->branch_0 = pdev;
  755. /* If this device claims to have more than 2 channels then
  756. * fetch Branch 1's information
  757. */
  758. if (pvt->maxch >= CHANNELS_PER_BRANCH) {
  759. pdev = NULL;
  760. pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
  761. PCI_DEVICE_ID_I5000_BRANCH_1, pdev);
  762. if (pdev == NULL) {
  763. i5000_printk(KERN_ERR,
  764. "MC: 'BRANCH 1' device not found:"
  765. "vendor 0x%x device 0x%x Func 0 "
  766. "(broken BIOS?)\n",
  767. PCI_VENDOR_ID_INTEL,
  768. PCI_DEVICE_ID_I5000_BRANCH_1);
  769. pci_dev_put(pvt->branchmap_werrors);
  770. pci_dev_put(pvt->fsb_error_regs);
  771. pci_dev_put(pvt->branch_0);
  772. return 1;
  773. }
  774. pvt->branch_1 = pdev;
  775. }
  776. return 0;
  777. }
  778. /*
  779. * i5000_put_devices 'put' all the devices that we have
  780. * reserved via 'get'
  781. */
  782. static void i5000_put_devices(struct mem_ctl_info *mci)
  783. {
  784. struct i5000_pvt *pvt;
  785. pvt = mci->pvt_info;
  786. pci_dev_put(pvt->branchmap_werrors); /* FUNC 1 */
  787. pci_dev_put(pvt->fsb_error_regs); /* FUNC 2 */
  788. pci_dev_put(pvt->branch_0); /* DEV 21 */
  789. /* Only if more than 2 channels do we release the second branch */
  790. if (pvt->maxch >= CHANNELS_PER_BRANCH)
  791. pci_dev_put(pvt->branch_1); /* DEV 22 */
  792. }
  793. /*
  794. * determine_amb_resent
  795. *
  796. * the information is contained in NUM_MTRS different registers
  797. * determineing which of the NUM_MTRS requires knowing
  798. * which channel is in question
  799. *
  800. * 2 branches, each with 2 channels
  801. * b0_ambpresent0 for channel '0'
  802. * b0_ambpresent1 for channel '1'
  803. * b1_ambpresent0 for channel '2'
  804. * b1_ambpresent1 for channel '3'
  805. */
  806. static int determine_amb_present_reg(struct i5000_pvt *pvt, int channel)
  807. {
  808. int amb_present;
  809. if (channel < CHANNELS_PER_BRANCH) {
  810. if (channel & 0x1)
  811. amb_present = pvt->b0_ambpresent1;
  812. else
  813. amb_present = pvt->b0_ambpresent0;
  814. } else {
  815. if (channel & 0x1)
  816. amb_present = pvt->b1_ambpresent1;
  817. else
  818. amb_present = pvt->b1_ambpresent0;
  819. }
  820. return amb_present;
  821. }
  822. /*
  823. * determine_mtr(pvt, csrow, channel)
  824. *
  825. * return the proper MTR register as determine by the csrow and channel desired
  826. */
  827. static int determine_mtr(struct i5000_pvt *pvt, int csrow, int channel)
  828. {
  829. int mtr;
  830. if (channel < CHANNELS_PER_BRANCH)
  831. mtr = pvt->b0_mtr[csrow >> 1];
  832. else
  833. mtr = pvt->b1_mtr[csrow >> 1];
  834. return mtr;
  835. }
  836. /*
  837. */
  838. static void decode_mtr(int slot_row, u16 mtr)
  839. {
  840. int ans;
  841. ans = MTR_DIMMS_PRESENT(mtr);
  842. debugf2("\tMTR%d=0x%x: DIMMs are %s\n", slot_row, mtr,
  843. ans ? "Present" : "NOT Present");
  844. if (!ans)
  845. return;
  846. debugf2("\t\tWIDTH: x%d\n", MTR_DRAM_WIDTH(mtr));
  847. debugf2("\t\tNUMBANK: %d bank(s)\n", MTR_DRAM_BANKS(mtr));
  848. debugf2("\t\tNUMRANK: %s\n", MTR_DIMM_RANK(mtr) ? "double" : "single");
  849. debugf2("\t\tNUMROW: %s\n", numrow_toString[MTR_DIMM_ROWS(mtr)]);
  850. debugf2("\t\tNUMCOL: %s\n", numcol_toString[MTR_DIMM_COLS(mtr)]);
  851. }
  852. static void handle_channel(struct i5000_pvt *pvt, int csrow, int channel,
  853. struct i5000_dimm_info *dinfo)
  854. {
  855. int mtr;
  856. int amb_present_reg;
  857. int addrBits;
  858. mtr = determine_mtr(pvt, csrow, channel);
  859. if (MTR_DIMMS_PRESENT(mtr)) {
  860. amb_present_reg = determine_amb_present_reg(pvt, channel);
  861. /* Determine if there is a DIMM present in this DIMM slot */
  862. if (amb_present_reg & (1 << (csrow >> 1))) {
  863. dinfo->dual_rank = MTR_DIMM_RANK(mtr);
  864. if (!((dinfo->dual_rank == 0) &&
  865. ((csrow & 0x1) == 0x1))) {
  866. /* Start with the number of bits for a Bank
  867. * on the DRAM */
  868. addrBits = MTR_DRAM_BANKS_ADDR_BITS(mtr);
  869. /* Add thenumber of ROW bits */
  870. addrBits += MTR_DIMM_ROWS_ADDR_BITS(mtr);
  871. /* add the number of COLUMN bits */
  872. addrBits += MTR_DIMM_COLS_ADDR_BITS(mtr);
  873. addrBits += 6; /* add 64 bits per DIMM */
  874. addrBits -= 20; /* divide by 2^^20 */
  875. addrBits -= 3; /* 8 bits per bytes */
  876. dinfo->megabytes = 1 << addrBits;
  877. }
  878. }
  879. }
  880. }
  881. /*
  882. * calculate_dimm_size
  883. *
  884. * also will output a DIMM matrix map, if debug is enabled, for viewing
  885. * how the DIMMs are populated
  886. */
  887. static void calculate_dimm_size(struct i5000_pvt *pvt)
  888. {
  889. struct i5000_dimm_info *dinfo;
  890. int csrow, max_csrows;
  891. char *p, *mem_buffer;
  892. int space, n;
  893. int channel;
  894. /* ================= Generate some debug output ================= */
  895. space = PAGE_SIZE;
  896. mem_buffer = p = kmalloc(space, GFP_KERNEL);
  897. if (p == NULL) {
  898. i5000_printk(KERN_ERR, "MC: %s:%s() kmalloc() failed\n",
  899. __FILE__, __func__);
  900. return;
  901. }
  902. n = snprintf(p, space, "\n");
  903. p += n;
  904. space -= n;
  905. /* Scan all the actual CSROWS (which is # of DIMMS * 2)
  906. * and calculate the information for each DIMM
  907. * Start with the highest csrow first, to display it first
  908. * and work toward the 0th csrow
  909. */
  910. max_csrows = pvt->maxdimmperch * 2;
  911. for (csrow = max_csrows - 1; csrow >= 0; csrow--) {
  912. /* on an odd csrow, first output a 'boundary' marker,
  913. * then reset the message buffer */
  914. if (csrow & 0x1) {
  915. n = snprintf(p, space, "---------------------------"
  916. "--------------------------------");
  917. p += n;
  918. space -= n;
  919. debugf2("%s\n", mem_buffer);
  920. p = mem_buffer;
  921. space = PAGE_SIZE;
  922. }
  923. n = snprintf(p, space, "csrow %2d ", csrow);
  924. p += n;
  925. space -= n;
  926. for (channel = 0; channel < pvt->maxch; channel++) {
  927. dinfo = &pvt->dimm_info[csrow][channel];
  928. handle_channel(pvt, csrow, channel, dinfo);
  929. n = snprintf(p, space, "%4d MB | ", dinfo->megabytes);
  930. p += n;
  931. space -= n;
  932. }
  933. n = snprintf(p, space, "\n");
  934. p += n;
  935. space -= n;
  936. }
  937. /* Output the last bottom 'boundary' marker */
  938. n = snprintf(p, space, "---------------------------"
  939. "--------------------------------\n");
  940. p += n;
  941. space -= n;
  942. /* now output the 'channel' labels */
  943. n = snprintf(p, space, " ");
  944. p += n;
  945. space -= n;
  946. for (channel = 0; channel < pvt->maxch; channel++) {
  947. n = snprintf(p, space, "channel %d | ", channel);
  948. p += n;
  949. space -= n;
  950. }
  951. n = snprintf(p, space, "\n");
  952. p += n;
  953. space -= n;
  954. /* output the last message and free buffer */
  955. debugf2("%s\n", mem_buffer);
  956. kfree(mem_buffer);
  957. }
  958. /*
  959. * i5000_get_mc_regs read in the necessary registers and
  960. * cache locally
  961. *
  962. * Fills in the private data members
  963. */
  964. static void i5000_get_mc_regs(struct mem_ctl_info *mci)
  965. {
  966. struct i5000_pvt *pvt;
  967. u32 actual_tolm;
  968. u16 limit;
  969. int slot_row;
  970. int maxch;
  971. int maxdimmperch;
  972. int way0, way1;
  973. pvt = mci->pvt_info;
  974. pci_read_config_dword(pvt->system_address, AMBASE,
  975. (u32 *) & pvt->ambase);
  976. pci_read_config_dword(pvt->system_address, AMBASE + sizeof(u32),
  977. ((u32 *) & pvt->ambase) + sizeof(u32));
  978. maxdimmperch = pvt->maxdimmperch;
  979. maxch = pvt->maxch;
  980. debugf2("AMBASE= 0x%lx MAXCH= %d MAX-DIMM-Per-CH= %d\n",
  981. (long unsigned int)pvt->ambase, pvt->maxch, pvt->maxdimmperch);
  982. /* Get the Branch Map regs */
  983. pci_read_config_word(pvt->branchmap_werrors, TOLM, &pvt->tolm);
  984. pvt->tolm >>= 12;
  985. debugf2("\nTOLM (number of 256M regions) =%u (0x%x)\n", pvt->tolm,
  986. pvt->tolm);
  987. actual_tolm = pvt->tolm << 28;
  988. debugf2("Actual TOLM byte addr=%u (0x%x)\n", actual_tolm, actual_tolm);
  989. pci_read_config_word(pvt->branchmap_werrors, MIR0, &pvt->mir0);
  990. pci_read_config_word(pvt->branchmap_werrors, MIR1, &pvt->mir1);
  991. pci_read_config_word(pvt->branchmap_werrors, MIR2, &pvt->mir2);
  992. /* Get the MIR[0-2] regs */
  993. limit = (pvt->mir0 >> 4) & 0x0FFF;
  994. way0 = pvt->mir0 & 0x1;
  995. way1 = pvt->mir0 & 0x2;
  996. debugf2("MIR0: limit= 0x%x WAY1= %u WAY0= %x\n", limit, way1, way0);
  997. limit = (pvt->mir1 >> 4) & 0x0FFF;
  998. way0 = pvt->mir1 & 0x1;
  999. way1 = pvt->mir1 & 0x2;
  1000. debugf2("MIR1: limit= 0x%x WAY1= %u WAY0= %x\n", limit, way1, way0);
  1001. limit = (pvt->mir2 >> 4) & 0x0FFF;
  1002. way0 = pvt->mir2 & 0x1;
  1003. way1 = pvt->mir2 & 0x2;
  1004. debugf2("MIR2: limit= 0x%x WAY1= %u WAY0= %x\n", limit, way1, way0);
  1005. /* Get the MTR[0-3] regs */
  1006. for (slot_row = 0; slot_row < NUM_MTRS; slot_row++) {
  1007. int where = MTR0 + (slot_row * sizeof(u32));
  1008. pci_read_config_word(pvt->branch_0, where,
  1009. &pvt->b0_mtr[slot_row]);
  1010. debugf2("MTR%d where=0x%x B0 value=0x%x\n", slot_row, where,
  1011. pvt->b0_mtr[slot_row]);
  1012. if (pvt->maxch >= CHANNELS_PER_BRANCH) {
  1013. pci_read_config_word(pvt->branch_1, where,
  1014. &pvt->b1_mtr[slot_row]);
  1015. debugf2("MTR%d where=0x%x B1 value=0x%x\n", slot_row,
  1016. where, pvt->b0_mtr[slot_row]);
  1017. } else {
  1018. pvt->b1_mtr[slot_row] = 0;
  1019. }
  1020. }
  1021. /* Read and dump branch 0's MTRs */
  1022. debugf2("\nMemory Technology Registers:\n");
  1023. debugf2(" Branch 0:\n");
  1024. for (slot_row = 0; slot_row < NUM_MTRS; slot_row++) {
  1025. decode_mtr(slot_row, pvt->b0_mtr[slot_row]);
  1026. }
  1027. pci_read_config_word(pvt->branch_0, AMB_PRESENT_0,
  1028. &pvt->b0_ambpresent0);
  1029. debugf2("\t\tAMB-Branch 0-present0 0x%x:\n", pvt->b0_ambpresent0);
  1030. pci_read_config_word(pvt->branch_0, AMB_PRESENT_1,
  1031. &pvt->b0_ambpresent1);
  1032. debugf2("\t\tAMB-Branch 0-present1 0x%x:\n", pvt->b0_ambpresent1);
  1033. /* Only if we have 2 branchs (4 channels) */
  1034. if (pvt->maxch < CHANNELS_PER_BRANCH) {
  1035. pvt->b1_ambpresent0 = 0;
  1036. pvt->b1_ambpresent1 = 0;
  1037. } else {
  1038. /* Read and dump branch 1's MTRs */
  1039. debugf2(" Branch 1:\n");
  1040. for (slot_row = 0; slot_row < NUM_MTRS; slot_row++) {
  1041. decode_mtr(slot_row, pvt->b1_mtr[slot_row]);
  1042. }
  1043. pci_read_config_word(pvt->branch_1, AMB_PRESENT_0,
  1044. &pvt->b1_ambpresent0);
  1045. debugf2("\t\tAMB-Branch 1-present0 0x%x:\n",
  1046. pvt->b1_ambpresent0);
  1047. pci_read_config_word(pvt->branch_1, AMB_PRESENT_1,
  1048. &pvt->b1_ambpresent1);
  1049. debugf2("\t\tAMB-Branch 1-present1 0x%x:\n",
  1050. pvt->b1_ambpresent1);
  1051. }
  1052. /* Go and determine the size of each DIMM and place in an
  1053. * orderly matrix */
  1054. calculate_dimm_size(pvt);
  1055. }
  1056. /*
  1057. * i5000_init_csrows Initialize the 'csrows' table within
  1058. * the mci control structure with the
  1059. * addressing of memory.
  1060. *
  1061. * return:
  1062. * 0 success
  1063. * 1 no actual memory found on this MC
  1064. */
  1065. static int i5000_init_csrows(struct mem_ctl_info *mci)
  1066. {
  1067. struct i5000_pvt *pvt;
  1068. struct csrow_info *p_csrow;
  1069. int empty, channel_count;
  1070. int max_csrows;
  1071. int mtr;
  1072. int csrow_megs;
  1073. int channel;
  1074. int csrow;
  1075. pvt = mci->pvt_info;
  1076. channel_count = pvt->maxch;
  1077. max_csrows = pvt->maxdimmperch * 2;
  1078. empty = 1; /* Assume NO memory */
  1079. for (csrow = 0; csrow < max_csrows; csrow++) {
  1080. p_csrow = &mci->csrows[csrow];
  1081. p_csrow->csrow_idx = csrow;
  1082. /* use branch 0 for the basis */
  1083. mtr = pvt->b0_mtr[csrow >> 1];
  1084. /* if no DIMMS on this row, continue */
  1085. if (!MTR_DIMMS_PRESENT(mtr))
  1086. continue;
  1087. /* FAKE OUT VALUES, FIXME */
  1088. p_csrow->first_page = 0 + csrow * 20;
  1089. p_csrow->last_page = 9 + csrow * 20;
  1090. p_csrow->page_mask = 0xFFF;
  1091. p_csrow->grain = 8;
  1092. csrow_megs = 0;
  1093. for (channel = 0; channel < pvt->maxch; channel++) {
  1094. csrow_megs += pvt->dimm_info[csrow][channel].megabytes;
  1095. }
  1096. p_csrow->nr_pages = csrow_megs << 8;
  1097. /* Assume DDR2 for now */
  1098. p_csrow->mtype = MEM_FB_DDR2;
  1099. /* ask what device type on this row */
  1100. if (MTR_DRAM_WIDTH(mtr))
  1101. p_csrow->dtype = DEV_X8;
  1102. else
  1103. p_csrow->dtype = DEV_X4;
  1104. p_csrow->edac_mode = EDAC_S8ECD8ED;
  1105. empty = 0;
  1106. }
  1107. return empty;
  1108. }
  1109. /*
  1110. * i5000_enable_error_reporting
  1111. * Turn on the memory reporting features of the hardware
  1112. */
  1113. static void i5000_enable_error_reporting(struct mem_ctl_info *mci)
  1114. {
  1115. struct i5000_pvt *pvt;
  1116. u32 fbd_error_mask;
  1117. pvt = mci->pvt_info;
  1118. /* Read the FBD Error Mask Register */
  1119. pci_read_config_dword(pvt->branchmap_werrors, EMASK_FBD,
  1120. &fbd_error_mask);
  1121. /* Enable with a '0' */
  1122. fbd_error_mask &= ~(ENABLE_EMASK_ALL);
  1123. pci_write_config_dword(pvt->branchmap_werrors, EMASK_FBD,
  1124. fbd_error_mask);
  1125. }
  1126. /*
  1127. * i5000_get_dimm_and_channel_counts(pdev, &num_csrows, &num_channels)
  1128. *
  1129. * ask the device how many channels are present and how many CSROWS
  1130. * as well
  1131. */
  1132. static void i5000_get_dimm_and_channel_counts(struct pci_dev *pdev,
  1133. int *num_dimms_per_channel,
  1134. int *num_channels)
  1135. {
  1136. u8 value;
  1137. /* Need to retrieve just how many channels and dimms per channel are
  1138. * supported on this memory controller
  1139. */
  1140. pci_read_config_byte(pdev, MAXDIMMPERCH, &value);
  1141. *num_dimms_per_channel = (int)value *2;
  1142. pci_read_config_byte(pdev, MAXCH, &value);
  1143. *num_channels = (int)value;
  1144. }
  1145. /*
  1146. * i5000_probe1 Probe for ONE instance of device to see if it is
  1147. * present.
  1148. * return:
  1149. * 0 for FOUND a device
  1150. * < 0 for error code
  1151. */
  1152. static int i5000_probe1(struct pci_dev *pdev, int dev_idx)
  1153. {
  1154. struct mem_ctl_info *mci;
  1155. struct i5000_pvt *pvt;
  1156. int num_channels;
  1157. int num_dimms_per_channel;
  1158. int num_csrows;
  1159. debugf0("MC: " __FILE__ ": %s(), pdev bus %u dev=0x%x fn=0x%x\n",
  1160. __func__,
  1161. pdev->bus->number,
  1162. PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
  1163. /* We only are looking for func 0 of the set */
  1164. if (PCI_FUNC(pdev->devfn) != 0)
  1165. return -ENODEV;
  1166. /* Ask the devices for the number of CSROWS and CHANNELS so
  1167. * that we can calculate the memory resources, etc
  1168. *
  1169. * The Chipset will report what it can handle which will be greater
  1170. * or equal to what the motherboard manufacturer will implement.
  1171. *
  1172. * As we don't have a motherboard identification routine to determine
  1173. * actual number of slots/dimms per channel, we thus utilize the
  1174. * resource as specified by the chipset. Thus, we might have
  1175. * have more DIMMs per channel than actually on the mobo, but this
  1176. * allows the driver to support upto the chipset max, without
  1177. * some fancy mobo determination.
  1178. */
  1179. i5000_get_dimm_and_channel_counts(pdev, &num_dimms_per_channel,
  1180. &num_channels);
  1181. num_csrows = num_dimms_per_channel * 2;
  1182. debugf0("MC: %s(): Number of - Channels= %d DIMMS= %d CSROWS= %d\n",
  1183. __func__, num_channels, num_dimms_per_channel, num_csrows);
  1184. /* allocate a new MC control structure */
  1185. mci = edac_mc_alloc(sizeof(*pvt), num_csrows, num_channels, 0);
  1186. if (mci == NULL)
  1187. return -ENOMEM;
  1188. kobject_get(&mci->edac_mci_kobj);
  1189. debugf0("MC: " __FILE__ ": %s(): mci = %p\n", __func__, mci);
  1190. mci->dev = &pdev->dev; /* record ptr to the generic device */
  1191. pvt = mci->pvt_info;
  1192. pvt->system_address = pdev; /* Record this device in our private */
  1193. pvt->maxch = num_channels;
  1194. pvt->maxdimmperch = num_dimms_per_channel;
  1195. /* 'get' the pci devices we want to reserve for our use */
  1196. if (i5000_get_devices(mci, dev_idx))
  1197. goto fail0;
  1198. /* Time to get serious */
  1199. i5000_get_mc_regs(mci); /* retrieve the hardware registers */
  1200. mci->mc_idx = 0;
  1201. mci->mtype_cap = MEM_FLAG_FB_DDR2;
  1202. mci->edac_ctl_cap = EDAC_FLAG_NONE;
  1203. mci->edac_cap = EDAC_FLAG_NONE;
  1204. mci->mod_name = "i5000_edac.c";
  1205. mci->mod_ver = I5000_REVISION;
  1206. mci->ctl_name = i5000_devs[dev_idx].ctl_name;
  1207. mci->dev_name = pci_name(pdev);
  1208. mci->ctl_page_to_phys = NULL;
  1209. /* Set the function pointer to an actual operation function */
  1210. mci->edac_check = i5000_check_error;
  1211. /* initialize the MC control structure 'csrows' table
  1212. * with the mapping and control information */
  1213. if (i5000_init_csrows(mci)) {
  1214. debugf0("MC: Setting mci->edac_cap to EDAC_FLAG_NONE\n"
  1215. " because i5000_init_csrows() returned nonzero "
  1216. "value\n");
  1217. mci->edac_cap = EDAC_FLAG_NONE; /* no csrows found */
  1218. } else {
  1219. debugf1("MC: Enable error reporting now\n");
  1220. i5000_enable_error_reporting(mci);
  1221. }
  1222. /* add this new MC control structure to EDAC's list of MCs */
  1223. if (edac_mc_add_mc(mci)) {
  1224. debugf0("MC: " __FILE__
  1225. ": %s(): failed edac_mc_add_mc()\n", __func__);
  1226. /* FIXME: perhaps some code should go here that disables error
  1227. * reporting if we just enabled it
  1228. */
  1229. goto fail1;
  1230. }
  1231. i5000_clear_error(mci);
  1232. /* allocating generic PCI control info */
  1233. i5000_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
  1234. if (!i5000_pci) {
  1235. printk(KERN_WARNING
  1236. "%s(): Unable to create PCI control\n",
  1237. __func__);
  1238. printk(KERN_WARNING
  1239. "%s(): PCI error report via EDAC not setup\n",
  1240. __func__);
  1241. }
  1242. return 0;
  1243. /* Error exit unwinding stack */
  1244. fail1:
  1245. i5000_put_devices(mci);
  1246. fail0:
  1247. kobject_put(&mci->edac_mci_kobj);
  1248. edac_mc_free(mci);
  1249. return -ENODEV;
  1250. }
  1251. /*
  1252. * i5000_init_one constructor for one instance of device
  1253. *
  1254. * returns:
  1255. * negative on error
  1256. * count (>= 0)
  1257. */
  1258. static int __devinit i5000_init_one(struct pci_dev *pdev,
  1259. const struct pci_device_id *id)
  1260. {
  1261. int rc;
  1262. debugf0("MC: " __FILE__ ": %s()\n", __func__);
  1263. /* wake up device */
  1264. rc = pci_enable_device(pdev);
  1265. if (rc == -EIO)
  1266. return rc;
  1267. /* now probe and enable the device */
  1268. return i5000_probe1(pdev, id->driver_data);
  1269. }
  1270. /*
  1271. * i5000_remove_one destructor for one instance of device
  1272. *
  1273. */
  1274. static void __devexit i5000_remove_one(struct pci_dev *pdev)
  1275. {
  1276. struct mem_ctl_info *mci;
  1277. debugf0(__FILE__ ": %s()\n", __func__);
  1278. if (i5000_pci)
  1279. edac_pci_release_generic_ctl(i5000_pci);
  1280. if ((mci = edac_mc_del_mc(&pdev->dev)) == NULL)
  1281. return;
  1282. /* retrieve references to resources, and free those resources */
  1283. i5000_put_devices(mci);
  1284. kobject_put(&mci->edac_mci_kobj);
  1285. edac_mc_free(mci);
  1286. }
  1287. /*
  1288. * pci_device_id table for which devices we are looking for
  1289. *
  1290. * The "E500P" device is the first device supported.
  1291. */
  1292. static const struct pci_device_id i5000_pci_tbl[] __devinitdata = {
  1293. {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_I5000_DEV16),
  1294. .driver_data = I5000P},
  1295. {0,} /* 0 terminated list. */
  1296. };
  1297. MODULE_DEVICE_TABLE(pci, i5000_pci_tbl);
  1298. /*
  1299. * i5000_driver pci_driver structure for this module
  1300. *
  1301. */
  1302. static struct pci_driver i5000_driver = {
  1303. .name = KBUILD_BASENAME,
  1304. .probe = i5000_init_one,
  1305. .remove = __devexit_p(i5000_remove_one),
  1306. .id_table = i5000_pci_tbl,
  1307. };
  1308. /*
  1309. * i5000_init Module entry function
  1310. * Try to initialize this module for its devices
  1311. */
  1312. static int __init i5000_init(void)
  1313. {
  1314. int pci_rc;
  1315. debugf2("MC: " __FILE__ ": %s()\n", __func__);
  1316. /* Ensure that the OPSTATE is set correctly for POLL or NMI */
  1317. opstate_init();
  1318. pci_rc = pci_register_driver(&i5000_driver);
  1319. return (pci_rc < 0) ? pci_rc : 0;
  1320. }
  1321. /*
  1322. * i5000_exit() Module exit function
  1323. * Unregister the driver
  1324. */
  1325. static void __exit i5000_exit(void)
  1326. {
  1327. debugf2("MC: " __FILE__ ": %s()\n", __func__);
  1328. pci_unregister_driver(&i5000_driver);
  1329. }
  1330. module_init(i5000_init);
  1331. module_exit(i5000_exit);
  1332. MODULE_LICENSE("GPL");
  1333. MODULE_AUTHOR
  1334. ("Linux Networx (http://lnxi.com) Doug Thompson <norsk5@xmission.com>");
  1335. MODULE_DESCRIPTION("MC Driver for Intel I5000 memory controllers - "
  1336. I5000_REVISION);
  1337. module_param(edac_op_state, int, 0444);
  1338. MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
  1339. module_param(misc_messages, int, 0444);
  1340. MODULE_PARM_DESC(misc_messages, "Log miscellaneous non fatal messages");