i5000_edac.c 42 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586
  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"
  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. #define MAX_BRANCHES 2
  248. /* Defines to extract the various fields from the
  249. * MTRx - Memory Technology Registers
  250. */
  251. #define MTR_DIMMS_PRESENT(mtr) ((mtr) & (0x1 << 8))
  252. #define MTR_DRAM_WIDTH(mtr) ((((mtr) >> 6) & 0x1) ? 8 : 4)
  253. #define MTR_DRAM_BANKS(mtr) ((((mtr) >> 5) & 0x1) ? 8 : 4)
  254. #define MTR_DRAM_BANKS_ADDR_BITS(mtr) ((MTR_DRAM_BANKS(mtr) == 8) ? 3 : 2)
  255. #define MTR_DIMM_RANK(mtr) (((mtr) >> 4) & 0x1)
  256. #define MTR_DIMM_RANK_ADDR_BITS(mtr) (MTR_DIMM_RANK(mtr) ? 2 : 1)
  257. #define MTR_DIMM_ROWS(mtr) (((mtr) >> 2) & 0x3)
  258. #define MTR_DIMM_ROWS_ADDR_BITS(mtr) (MTR_DIMM_ROWS(mtr) + 13)
  259. #define MTR_DIMM_COLS(mtr) ((mtr) & 0x3)
  260. #define MTR_DIMM_COLS_ADDR_BITS(mtr) (MTR_DIMM_COLS(mtr) + 10)
  261. /* enables the report of miscellaneous messages as CE errors - default off */
  262. static int misc_messages;
  263. /* Enumeration of supported devices */
  264. enum i5000_chips {
  265. I5000P = 0,
  266. I5000V = 1, /* future */
  267. I5000X = 2 /* future */
  268. };
  269. /* Device name and register DID (Device ID) */
  270. struct i5000_dev_info {
  271. const char *ctl_name; /* name for this device */
  272. u16 fsb_mapping_errors; /* DID for the branchmap,control */
  273. };
  274. /* Table of devices attributes supported by this driver */
  275. static const struct i5000_dev_info i5000_devs[] = {
  276. [I5000P] = {
  277. .ctl_name = "I5000",
  278. .fsb_mapping_errors = PCI_DEVICE_ID_INTEL_I5000_DEV16,
  279. },
  280. };
  281. struct i5000_dimm_info {
  282. int megabytes; /* size, 0 means not present */
  283. int dual_rank;
  284. };
  285. #define MAX_CHANNELS 6 /* max possible channels */
  286. #define MAX_CSROWS (8*2) /* max possible csrows per channel */
  287. /* driver private data structure */
  288. struct i5000_pvt {
  289. struct pci_dev *system_address; /* 16.0 */
  290. struct pci_dev *branchmap_werrors; /* 16.1 */
  291. struct pci_dev *fsb_error_regs; /* 16.2 */
  292. struct pci_dev *branch_0; /* 21.0 */
  293. struct pci_dev *branch_1; /* 22.0 */
  294. u16 tolm; /* top of low memory */
  295. u64 ambase; /* AMB BAR */
  296. u16 mir0, mir1, mir2;
  297. u16 b0_mtr[NUM_MTRS]; /* Memory Technlogy Reg */
  298. u16 b0_ambpresent0; /* Branch 0, Channel 0 */
  299. u16 b0_ambpresent1; /* Brnach 0, Channel 1 */
  300. u16 b1_mtr[NUM_MTRS]; /* Memory Technlogy Reg */
  301. u16 b1_ambpresent0; /* Branch 1, Channel 8 */
  302. u16 b1_ambpresent1; /* Branch 1, Channel 1 */
  303. /* DIMM information matrix, allocating architecture maximums */
  304. struct i5000_dimm_info dimm_info[MAX_CSROWS][MAX_CHANNELS];
  305. /* Actual values for this controller */
  306. int maxch; /* Max channels */
  307. int maxdimmperch; /* Max DIMMs per channel */
  308. };
  309. /* I5000 MCH error information retrieved from Hardware */
  310. struct i5000_error_info {
  311. /* These registers are always read from the MC */
  312. u32 ferr_fat_fbd; /* First Errors Fatal */
  313. u32 nerr_fat_fbd; /* Next Errors Fatal */
  314. u32 ferr_nf_fbd; /* First Errors Non-Fatal */
  315. u32 nerr_nf_fbd; /* Next Errors Non-Fatal */
  316. /* These registers are input ONLY if there was a Recoverable Error */
  317. u32 redmemb; /* Recoverable Mem Data Error log B */
  318. u16 recmema; /* Recoverable Mem Error log A */
  319. u32 recmemb; /* Recoverable Mem Error log B */
  320. /* These registers are input ONLY if there was a
  321. * Non-Recoverable Error */
  322. u16 nrecmema; /* Non-Recoverable Mem log A */
  323. u16 nrecmemb; /* Non-Recoverable Mem log B */
  324. };
  325. static struct edac_pci_ctl_info *i5000_pci;
  326. /*
  327. * i5000_get_error_info Retrieve the hardware error information from
  328. * the hardware and cache it in the 'info'
  329. * structure
  330. */
  331. static void i5000_get_error_info(struct mem_ctl_info *mci,
  332. struct i5000_error_info *info)
  333. {
  334. struct i5000_pvt *pvt;
  335. u32 value;
  336. pvt = mci->pvt_info;
  337. /* read in the 1st FATAL error register */
  338. pci_read_config_dword(pvt->branchmap_werrors, FERR_FAT_FBD, &value);
  339. /* Mask only the bits that the doc says are valid
  340. */
  341. value &= (FERR_FAT_FBDCHAN | FERR_FAT_MASK);
  342. /* If there is an error, then read in the */
  343. /* NEXT FATAL error register and the Memory Error Log Register A */
  344. if (value & FERR_FAT_MASK) {
  345. info->ferr_fat_fbd = value;
  346. /* harvest the various error data we need */
  347. pci_read_config_dword(pvt->branchmap_werrors,
  348. NERR_FAT_FBD, &info->nerr_fat_fbd);
  349. pci_read_config_word(pvt->branchmap_werrors,
  350. NRECMEMA, &info->nrecmema);
  351. pci_read_config_word(pvt->branchmap_werrors,
  352. NRECMEMB, &info->nrecmemb);
  353. /* Clear the error bits, by writing them back */
  354. pci_write_config_dword(pvt->branchmap_werrors,
  355. FERR_FAT_FBD, value);
  356. } else {
  357. info->ferr_fat_fbd = 0;
  358. info->nerr_fat_fbd = 0;
  359. info->nrecmema = 0;
  360. info->nrecmemb = 0;
  361. }
  362. /* read in the 1st NON-FATAL error register */
  363. pci_read_config_dword(pvt->branchmap_werrors, FERR_NF_FBD, &value);
  364. /* If there is an error, then read in the 1st NON-FATAL error
  365. * register as well */
  366. if (value & FERR_NF_MASK) {
  367. info->ferr_nf_fbd = value;
  368. /* harvest the various error data we need */
  369. pci_read_config_dword(pvt->branchmap_werrors,
  370. NERR_NF_FBD, &info->nerr_nf_fbd);
  371. pci_read_config_word(pvt->branchmap_werrors,
  372. RECMEMA, &info->recmema);
  373. pci_read_config_dword(pvt->branchmap_werrors,
  374. RECMEMB, &info->recmemb);
  375. pci_read_config_dword(pvt->branchmap_werrors,
  376. REDMEMB, &info->redmemb);
  377. /* Clear the error bits, by writing them back */
  378. pci_write_config_dword(pvt->branchmap_werrors,
  379. FERR_NF_FBD, value);
  380. } else {
  381. info->ferr_nf_fbd = 0;
  382. info->nerr_nf_fbd = 0;
  383. info->recmema = 0;
  384. info->recmemb = 0;
  385. info->redmemb = 0;
  386. }
  387. }
  388. /*
  389. * i5000_process_fatal_error_info(struct mem_ctl_info *mci,
  390. * struct i5000_error_info *info,
  391. * int handle_errors);
  392. *
  393. * handle the Intel FATAL errors, if any
  394. */
  395. static void i5000_process_fatal_error_info(struct mem_ctl_info *mci,
  396. struct i5000_error_info *info,
  397. int handle_errors)
  398. {
  399. char msg[EDAC_MC_LABEL_LEN + 1 + 160];
  400. char *specific = NULL;
  401. u32 allErrors;
  402. int channel;
  403. int bank;
  404. int rank;
  405. int rdwr;
  406. int ras, cas;
  407. /* mask off the Error bits that are possible */
  408. allErrors = (info->ferr_fat_fbd & FERR_FAT_MASK);
  409. if (!allErrors)
  410. return; /* if no error, return now */
  411. channel = EXTRACT_FBDCHAN_INDX(info->ferr_fat_fbd);
  412. /* Use the NON-Recoverable macros to extract data */
  413. bank = NREC_BANK(info->nrecmema);
  414. rank = NREC_RANK(info->nrecmema);
  415. rdwr = NREC_RDWR(info->nrecmema);
  416. ras = NREC_RAS(info->nrecmemb);
  417. cas = NREC_CAS(info->nrecmemb);
  418. debugf0("\t\tCSROW= %d Channel= %d "
  419. "(DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
  420. rank, channel, bank,
  421. rdwr ? "Write" : "Read", ras, cas);
  422. /* Only 1 bit will be on */
  423. switch (allErrors) {
  424. case FERR_FAT_M1ERR:
  425. specific = "Alert on non-redundant retry or fast "
  426. "reset timeout";
  427. break;
  428. case FERR_FAT_M2ERR:
  429. specific = "Northbound CRC error on non-redundant "
  430. "retry";
  431. break;
  432. case FERR_FAT_M3ERR:
  433. {
  434. static int done;
  435. /*
  436. * This error is generated to inform that the intelligent
  437. * throttling is disabled and the temperature passed the
  438. * specified middle point. Since this is something the BIOS
  439. * should take care of, we'll warn only once to avoid
  440. * worthlessly flooding the log.
  441. */
  442. if (done)
  443. return;
  444. done++;
  445. specific = ">Tmid Thermal event with intelligent "
  446. "throttling disabled";
  447. }
  448. break;
  449. }
  450. /* Form out message */
  451. snprintf(msg, sizeof(msg),
  452. "Bank=%d RAS=%d CAS=%d FATAL Err=0x%x (%s)",
  453. bank, ras, cas, allErrors, specific);
  454. /* Call the helper to output message */
  455. edac_mc_handle_error(HW_EVENT_ERR_FATAL, mci, 0, 0, 0,
  456. channel >> 1, channel & 1, rank,
  457. rdwr ? "Write error" : "Read error",
  458. msg, NULL);
  459. }
  460. /*
  461. * i5000_process_fatal_error_info(struct mem_ctl_info *mci,
  462. * struct i5000_error_info *info,
  463. * int handle_errors);
  464. *
  465. * handle the Intel NON-FATAL errors, if any
  466. */
  467. static void i5000_process_nonfatal_error_info(struct mem_ctl_info *mci,
  468. struct i5000_error_info *info,
  469. int handle_errors)
  470. {
  471. char msg[EDAC_MC_LABEL_LEN + 1 + 170];
  472. char *specific = NULL;
  473. u32 allErrors;
  474. u32 ue_errors;
  475. u32 ce_errors;
  476. u32 misc_errors;
  477. int branch;
  478. int channel;
  479. int bank;
  480. int rank;
  481. int rdwr;
  482. int ras, cas;
  483. /* mask off the Error bits that are possible */
  484. allErrors = (info->ferr_nf_fbd & FERR_NF_MASK);
  485. if (!allErrors)
  486. return; /* if no error, return now */
  487. /* ONLY ONE of the possible error bits will be set, as per the docs */
  488. ue_errors = allErrors & FERR_NF_UNCORRECTABLE;
  489. if (ue_errors) {
  490. debugf0("\tUncorrected bits= 0x%x\n", ue_errors);
  491. branch = EXTRACT_FBDCHAN_INDX(info->ferr_nf_fbd);
  492. /*
  493. * According with i5000 datasheet, bit 28 has no significance
  494. * for errors M4Err-M12Err and M17Err-M21Err, on FERR_NF_FBD
  495. */
  496. channel = branch & 2;
  497. bank = NREC_BANK(info->nrecmema);
  498. rank = NREC_RANK(info->nrecmema);
  499. rdwr = NREC_RDWR(info->nrecmema);
  500. ras = NREC_RAS(info->nrecmemb);
  501. cas = NREC_CAS(info->nrecmemb);
  502. debugf0
  503. ("\t\tCSROW= %d Channels= %d,%d (Branch= %d "
  504. "DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
  505. rank, channel, channel + 1, branch >> 1, bank,
  506. rdwr ? "Write" : "Read", ras, cas);
  507. switch (ue_errors) {
  508. case FERR_NF_M12ERR:
  509. specific = "Non-Aliased Uncorrectable Patrol Data ECC";
  510. break;
  511. case FERR_NF_M11ERR:
  512. specific = "Non-Aliased Uncorrectable Spare-Copy "
  513. "Data ECC";
  514. break;
  515. case FERR_NF_M10ERR:
  516. specific = "Non-Aliased Uncorrectable Mirrored Demand "
  517. "Data ECC";
  518. break;
  519. case FERR_NF_M9ERR:
  520. specific = "Non-Aliased Uncorrectable Non-Mirrored "
  521. "Demand Data ECC";
  522. break;
  523. case FERR_NF_M8ERR:
  524. specific = "Aliased Uncorrectable Patrol Data ECC";
  525. break;
  526. case FERR_NF_M7ERR:
  527. specific = "Aliased Uncorrectable Spare-Copy Data ECC";
  528. break;
  529. case FERR_NF_M6ERR:
  530. specific = "Aliased Uncorrectable Mirrored Demand "
  531. "Data ECC";
  532. break;
  533. case FERR_NF_M5ERR:
  534. specific = "Aliased Uncorrectable Non-Mirrored Demand "
  535. "Data ECC";
  536. break;
  537. case FERR_NF_M4ERR:
  538. specific = "Uncorrectable Data ECC on Replay";
  539. break;
  540. }
  541. /* Form out message */
  542. snprintf(msg, sizeof(msg),
  543. "Rank=%d Bank=%d RAS=%d CAS=%d, UE Err=0x%x (%s)",
  544. rank, bank, ras, cas, ue_errors, specific);
  545. /* Call the helper to output message */
  546. edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 0, 0, 0,
  547. channel >> 1, -1, rank,
  548. rdwr ? "Write error" : "Read error",
  549. msg, NULL);
  550. }
  551. /* Check correctable errors */
  552. ce_errors = allErrors & FERR_NF_CORRECTABLE;
  553. if (ce_errors) {
  554. debugf0("\tCorrected bits= 0x%x\n", ce_errors);
  555. branch = EXTRACT_FBDCHAN_INDX(info->ferr_nf_fbd);
  556. channel = 0;
  557. if (REC_ECC_LOCATOR_ODD(info->redmemb))
  558. channel = 1;
  559. /* Convert channel to be based from zero, instead of
  560. * from branch base of 0 */
  561. channel += branch;
  562. bank = REC_BANK(info->recmema);
  563. rank = REC_RANK(info->recmema);
  564. rdwr = REC_RDWR(info->recmema);
  565. ras = REC_RAS(info->recmemb);
  566. cas = REC_CAS(info->recmemb);
  567. debugf0("\t\tCSROW= %d Channel= %d (Branch %d "
  568. "DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
  569. rank, channel, branch >> 1, bank,
  570. rdwr ? "Write" : "Read", ras, cas);
  571. switch (ce_errors) {
  572. case FERR_NF_M17ERR:
  573. specific = "Correctable Non-Mirrored Demand Data ECC";
  574. break;
  575. case FERR_NF_M18ERR:
  576. specific = "Correctable Mirrored Demand Data ECC";
  577. break;
  578. case FERR_NF_M19ERR:
  579. specific = "Correctable Spare-Copy Data ECC";
  580. break;
  581. case FERR_NF_M20ERR:
  582. specific = "Correctable Patrol Data ECC";
  583. break;
  584. }
  585. /* Form out message */
  586. snprintf(msg, sizeof(msg),
  587. "Rank=%d Bank=%d RDWR=%s RAS=%d "
  588. "CAS=%d, CE Err=0x%x (%s))", branch >> 1, bank,
  589. rdwr ? "Write" : "Read", ras, cas, ce_errors,
  590. specific);
  591. /* Call the helper to output message */
  592. edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 0, 0, 0,
  593. channel >> 1, channel % 2, rank,
  594. rdwr ? "Write error" : "Read error",
  595. msg, NULL);
  596. }
  597. if (!misc_messages)
  598. return;
  599. misc_errors = allErrors & (FERR_NF_NON_RETRY | FERR_NF_NORTH_CRC |
  600. FERR_NF_SPD_PROTOCOL | FERR_NF_DIMM_SPARE);
  601. if (misc_errors) {
  602. switch (misc_errors) {
  603. case FERR_NF_M13ERR:
  604. specific = "Non-Retry or Redundant Retry FBD Memory "
  605. "Alert or Redundant Fast Reset Timeout";
  606. break;
  607. case FERR_NF_M14ERR:
  608. specific = "Non-Retry or Redundant Retry FBD "
  609. "Configuration Alert";
  610. break;
  611. case FERR_NF_M15ERR:
  612. specific = "Non-Retry or Redundant Retry FBD "
  613. "Northbound CRC error on read data";
  614. break;
  615. case FERR_NF_M21ERR:
  616. specific = "FBD Northbound CRC error on "
  617. "FBD Sync Status";
  618. break;
  619. case FERR_NF_M22ERR:
  620. specific = "SPD protocol error";
  621. break;
  622. case FERR_NF_M27ERR:
  623. specific = "DIMM-spare copy started";
  624. break;
  625. case FERR_NF_M28ERR:
  626. specific = "DIMM-spare copy completed";
  627. break;
  628. }
  629. branch = EXTRACT_FBDCHAN_INDX(info->ferr_nf_fbd);
  630. /* Form out message */
  631. snprintf(msg, sizeof(msg),
  632. "Err=%#x (%s)", misc_errors, specific);
  633. /* Call the helper to output message */
  634. edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 0, 0, 0,
  635. branch >> 1, -1, -1,
  636. "Misc error", msg, NULL);
  637. }
  638. }
  639. /*
  640. * i5000_process_error_info Process the error info that is
  641. * in the 'info' structure, previously retrieved from hardware
  642. */
  643. static void i5000_process_error_info(struct mem_ctl_info *mci,
  644. struct i5000_error_info *info,
  645. int handle_errors)
  646. {
  647. /* First handle any fatal errors that occurred */
  648. i5000_process_fatal_error_info(mci, info, handle_errors);
  649. /* now handle any non-fatal errors that occurred */
  650. i5000_process_nonfatal_error_info(mci, info, handle_errors);
  651. }
  652. /*
  653. * i5000_clear_error Retrieve any error from the hardware
  654. * but do NOT process that error.
  655. * Used for 'clearing' out of previous errors
  656. * Called by the Core module.
  657. */
  658. static void i5000_clear_error(struct mem_ctl_info *mci)
  659. {
  660. struct i5000_error_info info;
  661. i5000_get_error_info(mci, &info);
  662. }
  663. /*
  664. * i5000_check_error Retrieve and process errors reported by the
  665. * hardware. Called by the Core module.
  666. */
  667. static void i5000_check_error(struct mem_ctl_info *mci)
  668. {
  669. struct i5000_error_info info;
  670. debugf4("MC%d\n", mci->mc_idx);
  671. i5000_get_error_info(mci, &info);
  672. i5000_process_error_info(mci, &info, 1);
  673. }
  674. /*
  675. * i5000_get_devices Find and perform 'get' operation on the MCH's
  676. * device/functions we want to reference for this driver
  677. *
  678. * Need to 'get' device 16 func 1 and func 2
  679. */
  680. static int i5000_get_devices(struct mem_ctl_info *mci, int dev_idx)
  681. {
  682. //const struct i5000_dev_info *i5000_dev = &i5000_devs[dev_idx];
  683. struct i5000_pvt *pvt;
  684. struct pci_dev *pdev;
  685. pvt = mci->pvt_info;
  686. /* Attempt to 'get' the MCH register we want */
  687. pdev = NULL;
  688. while (1) {
  689. pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
  690. PCI_DEVICE_ID_INTEL_I5000_DEV16, pdev);
  691. /* End of list, leave */
  692. if (pdev == NULL) {
  693. i5000_printk(KERN_ERR,
  694. "'system address,Process Bus' "
  695. "device not found:"
  696. "vendor 0x%x device 0x%x FUNC 1 "
  697. "(broken BIOS?)\n",
  698. PCI_VENDOR_ID_INTEL,
  699. PCI_DEVICE_ID_INTEL_I5000_DEV16);
  700. return 1;
  701. }
  702. /* Scan for device 16 func 1 */
  703. if (PCI_FUNC(pdev->devfn) == 1)
  704. break;
  705. }
  706. pvt->branchmap_werrors = pdev;
  707. /* Attempt to 'get' the MCH register we want */
  708. pdev = NULL;
  709. while (1) {
  710. pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
  711. PCI_DEVICE_ID_INTEL_I5000_DEV16, pdev);
  712. if (pdev == NULL) {
  713. i5000_printk(KERN_ERR,
  714. "MC: 'branchmap,control,errors' "
  715. "device not found:"
  716. "vendor 0x%x device 0x%x Func 2 "
  717. "(broken BIOS?)\n",
  718. PCI_VENDOR_ID_INTEL,
  719. PCI_DEVICE_ID_INTEL_I5000_DEV16);
  720. pci_dev_put(pvt->branchmap_werrors);
  721. return 1;
  722. }
  723. /* Scan for device 16 func 1 */
  724. if (PCI_FUNC(pdev->devfn) == 2)
  725. break;
  726. }
  727. pvt->fsb_error_regs = pdev;
  728. debugf1("System Address, processor bus- PCI Bus ID: %s %x:%x\n",
  729. pci_name(pvt->system_address),
  730. pvt->system_address->vendor, pvt->system_address->device);
  731. debugf1("Branchmap, control and errors - PCI Bus ID: %s %x:%x\n",
  732. pci_name(pvt->branchmap_werrors),
  733. pvt->branchmap_werrors->vendor, pvt->branchmap_werrors->device);
  734. debugf1("FSB Error Regs - PCI Bus ID: %s %x:%x\n",
  735. pci_name(pvt->fsb_error_regs),
  736. pvt->fsb_error_regs->vendor, pvt->fsb_error_regs->device);
  737. pdev = NULL;
  738. pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
  739. PCI_DEVICE_ID_I5000_BRANCH_0, pdev);
  740. if (pdev == NULL) {
  741. i5000_printk(KERN_ERR,
  742. "MC: 'BRANCH 0' device not found:"
  743. "vendor 0x%x device 0x%x Func 0 (broken BIOS?)\n",
  744. PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_I5000_BRANCH_0);
  745. pci_dev_put(pvt->branchmap_werrors);
  746. pci_dev_put(pvt->fsb_error_regs);
  747. return 1;
  748. }
  749. pvt->branch_0 = pdev;
  750. /* If this device claims to have more than 2 channels then
  751. * fetch Branch 1's information
  752. */
  753. if (pvt->maxch >= CHANNELS_PER_BRANCH) {
  754. pdev = NULL;
  755. pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
  756. PCI_DEVICE_ID_I5000_BRANCH_1, pdev);
  757. if (pdev == NULL) {
  758. i5000_printk(KERN_ERR,
  759. "MC: 'BRANCH 1' device not found:"
  760. "vendor 0x%x device 0x%x Func 0 "
  761. "(broken BIOS?)\n",
  762. PCI_VENDOR_ID_INTEL,
  763. PCI_DEVICE_ID_I5000_BRANCH_1);
  764. pci_dev_put(pvt->branchmap_werrors);
  765. pci_dev_put(pvt->fsb_error_regs);
  766. pci_dev_put(pvt->branch_0);
  767. return 1;
  768. }
  769. pvt->branch_1 = pdev;
  770. }
  771. return 0;
  772. }
  773. /*
  774. * i5000_put_devices 'put' all the devices that we have
  775. * reserved via 'get'
  776. */
  777. static void i5000_put_devices(struct mem_ctl_info *mci)
  778. {
  779. struct i5000_pvt *pvt;
  780. pvt = mci->pvt_info;
  781. pci_dev_put(pvt->branchmap_werrors); /* FUNC 1 */
  782. pci_dev_put(pvt->fsb_error_regs); /* FUNC 2 */
  783. pci_dev_put(pvt->branch_0); /* DEV 21 */
  784. /* Only if more than 2 channels do we release the second branch */
  785. if (pvt->maxch >= CHANNELS_PER_BRANCH)
  786. pci_dev_put(pvt->branch_1); /* DEV 22 */
  787. }
  788. /*
  789. * determine_amb_resent
  790. *
  791. * the information is contained in NUM_MTRS different registers
  792. * determineing which of the NUM_MTRS requires knowing
  793. * which channel is in question
  794. *
  795. * 2 branches, each with 2 channels
  796. * b0_ambpresent0 for channel '0'
  797. * b0_ambpresent1 for channel '1'
  798. * b1_ambpresent0 for channel '2'
  799. * b1_ambpresent1 for channel '3'
  800. */
  801. static int determine_amb_present_reg(struct i5000_pvt *pvt, int channel)
  802. {
  803. int amb_present;
  804. if (channel < CHANNELS_PER_BRANCH) {
  805. if (channel & 0x1)
  806. amb_present = pvt->b0_ambpresent1;
  807. else
  808. amb_present = pvt->b0_ambpresent0;
  809. } else {
  810. if (channel & 0x1)
  811. amb_present = pvt->b1_ambpresent1;
  812. else
  813. amb_present = pvt->b1_ambpresent0;
  814. }
  815. return amb_present;
  816. }
  817. /*
  818. * determine_mtr(pvt, csrow, channel)
  819. *
  820. * return the proper MTR register as determine by the csrow and channel desired
  821. */
  822. static int determine_mtr(struct i5000_pvt *pvt, int slot, int channel)
  823. {
  824. int mtr;
  825. if (channel < CHANNELS_PER_BRANCH)
  826. mtr = pvt->b0_mtr[slot];
  827. else
  828. mtr = pvt->b1_mtr[slot];
  829. return mtr;
  830. }
  831. /*
  832. */
  833. static void decode_mtr(int slot_row, u16 mtr)
  834. {
  835. int ans;
  836. ans = MTR_DIMMS_PRESENT(mtr);
  837. debugf2("\tMTR%d=0x%x: DIMMs are %s\n", slot_row, mtr,
  838. ans ? "Present" : "NOT Present");
  839. if (!ans)
  840. return;
  841. debugf2("\t\tWIDTH: x%d\n", MTR_DRAM_WIDTH(mtr));
  842. debugf2("\t\tNUMBANK: %d bank(s)\n", MTR_DRAM_BANKS(mtr));
  843. debugf2("\t\tNUMRANK: %s\n", MTR_DIMM_RANK(mtr) ? "double" : "single");
  844. debugf2("\t\tNUMROW: %s\n",
  845. MTR_DIMM_ROWS(mtr) == 0 ? "8,192 - 13 rows" :
  846. MTR_DIMM_ROWS(mtr) == 1 ? "16,384 - 14 rows" :
  847. MTR_DIMM_ROWS(mtr) == 2 ? "32,768 - 15 rows" :
  848. "reserved");
  849. debugf2("\t\tNUMCOL: %s\n",
  850. MTR_DIMM_COLS(mtr) == 0 ? "1,024 - 10 columns" :
  851. MTR_DIMM_COLS(mtr) == 1 ? "2,048 - 11 columns" :
  852. MTR_DIMM_COLS(mtr) == 2 ? "4,096 - 12 columns" :
  853. "reserved");
  854. }
  855. static void handle_channel(struct i5000_pvt *pvt, int slot, int channel,
  856. struct i5000_dimm_info *dinfo)
  857. {
  858. int mtr;
  859. int amb_present_reg;
  860. int addrBits;
  861. mtr = determine_mtr(pvt, slot, channel);
  862. if (MTR_DIMMS_PRESENT(mtr)) {
  863. amb_present_reg = determine_amb_present_reg(pvt, channel);
  864. /* Determine if there is a DIMM present in this DIMM slot */
  865. if (amb_present_reg) {
  866. dinfo->dual_rank = MTR_DIMM_RANK(mtr);
  867. /* Start with the number of bits for a Bank
  868. * on the DRAM */
  869. addrBits = MTR_DRAM_BANKS_ADDR_BITS(mtr);
  870. /* Add the number of ROW bits */
  871. addrBits += MTR_DIMM_ROWS_ADDR_BITS(mtr);
  872. /* add the number of COLUMN bits */
  873. addrBits += MTR_DIMM_COLS_ADDR_BITS(mtr);
  874. addrBits += 6; /* add 64 bits per DIMM */
  875. addrBits -= 20; /* divide by 2^^20 */
  876. addrBits -= 3; /* 8 bits per bytes */
  877. dinfo->megabytes = 1 << addrBits;
  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 slot, channel, branch;
  891. char *p, *mem_buffer;
  892. int space, n;
  893. /* ================= Generate some debug output ================= */
  894. space = PAGE_SIZE;
  895. mem_buffer = p = kmalloc(space, GFP_KERNEL);
  896. if (p == NULL) {
  897. i5000_printk(KERN_ERR, "MC: %s:%s() kmalloc() failed\n",
  898. __FILE__, __func__);
  899. return;
  900. }
  901. /* Scan all the actual slots
  902. * and calculate the information for each DIMM
  903. * Start with the highest slot first, to display it first
  904. * and work toward the 0th slot
  905. */
  906. for (slot = pvt->maxdimmperch - 1; slot >= 0; slot--) {
  907. /* on an odd slot, first output a 'boundary' marker,
  908. * then reset the message buffer */
  909. if (slot & 0x1) {
  910. n = snprintf(p, space, "--------------------------"
  911. "--------------------------------");
  912. p += n;
  913. space -= n;
  914. debugf2("%s\n", mem_buffer);
  915. p = mem_buffer;
  916. space = PAGE_SIZE;
  917. }
  918. n = snprintf(p, space, "slot %2d ", slot);
  919. p += n;
  920. space -= n;
  921. for (channel = 0; channel < pvt->maxch; channel++) {
  922. dinfo = &pvt->dimm_info[slot][channel];
  923. handle_channel(pvt, slot, channel, dinfo);
  924. if (dinfo->megabytes)
  925. n = snprintf(p, space, "%4d MB %dR| ",
  926. dinfo->megabytes, dinfo->dual_rank + 1);
  927. else
  928. n = snprintf(p, space, "%4d MB | ", 0);
  929. p += n;
  930. space -= n;
  931. }
  932. p += n;
  933. space -= n;
  934. debugf2("%s\n", mem_buffer);
  935. p = mem_buffer;
  936. space = PAGE_SIZE;
  937. }
  938. /* Output the last bottom 'boundary' marker */
  939. n = snprintf(p, space, "--------------------------"
  940. "--------------------------------");
  941. p += n;
  942. space -= n;
  943. debugf2("%s\n", mem_buffer);
  944. p = mem_buffer;
  945. space = PAGE_SIZE;
  946. /* now output the 'channel' labels */
  947. n = snprintf(p, space, " ");
  948. p += n;
  949. space -= n;
  950. for (channel = 0; channel < pvt->maxch; channel++) {
  951. n = snprintf(p, space, "channel %d | ", channel);
  952. p += n;
  953. space -= n;
  954. }
  955. debugf2("%s\n", mem_buffer);
  956. p = mem_buffer;
  957. space = PAGE_SIZE;
  958. n = snprintf(p, space, " ");
  959. p += n;
  960. for (branch = 0; branch < MAX_BRANCHES; branch++) {
  961. n = snprintf(p, space, " branch %d | ", branch);
  962. p += n;
  963. space -= n;
  964. }
  965. /* output the last message and free buffer */
  966. debugf2("%s\n", mem_buffer);
  967. kfree(mem_buffer);
  968. }
  969. /*
  970. * i5000_get_mc_regs read in the necessary registers and
  971. * cache locally
  972. *
  973. * Fills in the private data members
  974. */
  975. static void i5000_get_mc_regs(struct mem_ctl_info *mci)
  976. {
  977. struct i5000_pvt *pvt;
  978. u32 actual_tolm;
  979. u16 limit;
  980. int slot_row;
  981. int maxch;
  982. int maxdimmperch;
  983. int way0, way1;
  984. pvt = mci->pvt_info;
  985. pci_read_config_dword(pvt->system_address, AMBASE,
  986. (u32 *) & pvt->ambase);
  987. pci_read_config_dword(pvt->system_address, AMBASE + sizeof(u32),
  988. ((u32 *) & pvt->ambase) + sizeof(u32));
  989. maxdimmperch = pvt->maxdimmperch;
  990. maxch = pvt->maxch;
  991. debugf2("AMBASE= 0x%lx MAXCH= %d MAX-DIMM-Per-CH= %d\n",
  992. (long unsigned int)pvt->ambase, pvt->maxch, pvt->maxdimmperch);
  993. /* Get the Branch Map regs */
  994. pci_read_config_word(pvt->branchmap_werrors, TOLM, &pvt->tolm);
  995. pvt->tolm >>= 12;
  996. debugf2("\nTOLM (number of 256M regions) =%u (0x%x)\n", pvt->tolm,
  997. pvt->tolm);
  998. actual_tolm = pvt->tolm << 28;
  999. debugf2("Actual TOLM byte addr=%u (0x%x)\n", actual_tolm, actual_tolm);
  1000. pci_read_config_word(pvt->branchmap_werrors, MIR0, &pvt->mir0);
  1001. pci_read_config_word(pvt->branchmap_werrors, MIR1, &pvt->mir1);
  1002. pci_read_config_word(pvt->branchmap_werrors, MIR2, &pvt->mir2);
  1003. /* Get the MIR[0-2] regs */
  1004. limit = (pvt->mir0 >> 4) & 0x0FFF;
  1005. way0 = pvt->mir0 & 0x1;
  1006. way1 = pvt->mir0 & 0x2;
  1007. debugf2("MIR0: limit= 0x%x WAY1= %u WAY0= %x\n", limit, way1, way0);
  1008. limit = (pvt->mir1 >> 4) & 0x0FFF;
  1009. way0 = pvt->mir1 & 0x1;
  1010. way1 = pvt->mir1 & 0x2;
  1011. debugf2("MIR1: limit= 0x%x WAY1= %u WAY0= %x\n", limit, way1, way0);
  1012. limit = (pvt->mir2 >> 4) & 0x0FFF;
  1013. way0 = pvt->mir2 & 0x1;
  1014. way1 = pvt->mir2 & 0x2;
  1015. debugf2("MIR2: limit= 0x%x WAY1= %u WAY0= %x\n", limit, way1, way0);
  1016. /* Get the MTR[0-3] regs */
  1017. for (slot_row = 0; slot_row < NUM_MTRS; slot_row++) {
  1018. int where = MTR0 + (slot_row * sizeof(u32));
  1019. pci_read_config_word(pvt->branch_0, where,
  1020. &pvt->b0_mtr[slot_row]);
  1021. debugf2("MTR%d where=0x%x B0 value=0x%x\n", slot_row, where,
  1022. pvt->b0_mtr[slot_row]);
  1023. if (pvt->maxch >= CHANNELS_PER_BRANCH) {
  1024. pci_read_config_word(pvt->branch_1, where,
  1025. &pvt->b1_mtr[slot_row]);
  1026. debugf2("MTR%d where=0x%x B1 value=0x%x\n", slot_row,
  1027. where, pvt->b1_mtr[slot_row]);
  1028. } else {
  1029. pvt->b1_mtr[slot_row] = 0;
  1030. }
  1031. }
  1032. /* Read and dump branch 0's MTRs */
  1033. debugf2("\nMemory Technology Registers:\n");
  1034. debugf2(" Branch 0:\n");
  1035. for (slot_row = 0; slot_row < NUM_MTRS; slot_row++) {
  1036. decode_mtr(slot_row, pvt->b0_mtr[slot_row]);
  1037. }
  1038. pci_read_config_word(pvt->branch_0, AMB_PRESENT_0,
  1039. &pvt->b0_ambpresent0);
  1040. debugf2("\t\tAMB-Branch 0-present0 0x%x:\n", pvt->b0_ambpresent0);
  1041. pci_read_config_word(pvt->branch_0, AMB_PRESENT_1,
  1042. &pvt->b0_ambpresent1);
  1043. debugf2("\t\tAMB-Branch 0-present1 0x%x:\n", pvt->b0_ambpresent1);
  1044. /* Only if we have 2 branchs (4 channels) */
  1045. if (pvt->maxch < CHANNELS_PER_BRANCH) {
  1046. pvt->b1_ambpresent0 = 0;
  1047. pvt->b1_ambpresent1 = 0;
  1048. } else {
  1049. /* Read and dump branch 1's MTRs */
  1050. debugf2(" Branch 1:\n");
  1051. for (slot_row = 0; slot_row < NUM_MTRS; slot_row++) {
  1052. decode_mtr(slot_row, pvt->b1_mtr[slot_row]);
  1053. }
  1054. pci_read_config_word(pvt->branch_1, AMB_PRESENT_0,
  1055. &pvt->b1_ambpresent0);
  1056. debugf2("\t\tAMB-Branch 1-present0 0x%x:\n",
  1057. pvt->b1_ambpresent0);
  1058. pci_read_config_word(pvt->branch_1, AMB_PRESENT_1,
  1059. &pvt->b1_ambpresent1);
  1060. debugf2("\t\tAMB-Branch 1-present1 0x%x:\n",
  1061. pvt->b1_ambpresent1);
  1062. }
  1063. /* Go and determine the size of each DIMM and place in an
  1064. * orderly matrix */
  1065. calculate_dimm_size(pvt);
  1066. }
  1067. /*
  1068. * i5000_init_csrows Initialize the 'csrows' table within
  1069. * the mci control structure with the
  1070. * addressing of memory.
  1071. *
  1072. * return:
  1073. * 0 success
  1074. * 1 no actual memory found on this MC
  1075. */
  1076. static int i5000_init_csrows(struct mem_ctl_info *mci)
  1077. {
  1078. struct i5000_pvt *pvt;
  1079. struct dimm_info *dimm;
  1080. int empty, channel_count;
  1081. int max_csrows;
  1082. int mtr;
  1083. int csrow_megs;
  1084. int channel;
  1085. int slot;
  1086. pvt = mci->pvt_info;
  1087. channel_count = pvt->maxch;
  1088. max_csrows = pvt->maxdimmperch * 2;
  1089. empty = 1; /* Assume NO memory */
  1090. /*
  1091. * FIXME: The memory layout used to map slot/channel into the
  1092. * real memory architecture is weird: branch+slot are "csrows"
  1093. * and channel is channel. That required an extra array (dimm_info)
  1094. * to map the dimms. A good cleanup would be to remove this array,
  1095. * and do a loop here with branch, channel, slot
  1096. */
  1097. for (slot = 0; slot < max_csrows; slot++) {
  1098. for (channel = 0; channel < pvt->maxch; channel++) {
  1099. mtr = determine_mtr(pvt, slot, channel);
  1100. if (!MTR_DIMMS_PRESENT(mtr))
  1101. continue;
  1102. dimm = EDAC_DIMM_PTR(mci->layers, mci->dimms, mci->n_layers,
  1103. channel / MAX_BRANCHES,
  1104. channel % MAX_BRANCHES, slot);
  1105. csrow_megs = pvt->dimm_info[slot][channel].megabytes;
  1106. dimm->grain = 8;
  1107. /* Assume DDR2 for now */
  1108. dimm->mtype = MEM_FB_DDR2;
  1109. /* ask what device type on this row */
  1110. if (MTR_DRAM_WIDTH(mtr))
  1111. dimm->dtype = DEV_X8;
  1112. else
  1113. dimm->dtype = DEV_X4;
  1114. dimm->edac_mode = EDAC_S8ECD8ED;
  1115. dimm->nr_pages = csrow_megs << 8;
  1116. }
  1117. empty = 0;
  1118. }
  1119. return empty;
  1120. }
  1121. /*
  1122. * i5000_enable_error_reporting
  1123. * Turn on the memory reporting features of the hardware
  1124. */
  1125. static void i5000_enable_error_reporting(struct mem_ctl_info *mci)
  1126. {
  1127. struct i5000_pvt *pvt;
  1128. u32 fbd_error_mask;
  1129. pvt = mci->pvt_info;
  1130. /* Read the FBD Error Mask Register */
  1131. pci_read_config_dword(pvt->branchmap_werrors, EMASK_FBD,
  1132. &fbd_error_mask);
  1133. /* Enable with a '0' */
  1134. fbd_error_mask &= ~(ENABLE_EMASK_ALL);
  1135. pci_write_config_dword(pvt->branchmap_werrors, EMASK_FBD,
  1136. fbd_error_mask);
  1137. }
  1138. /*
  1139. * i5000_get_dimm_and_channel_counts(pdev, &nr_csrows, &num_channels)
  1140. *
  1141. * ask the device how many channels are present and how many CSROWS
  1142. * as well
  1143. */
  1144. static void i5000_get_dimm_and_channel_counts(struct pci_dev *pdev,
  1145. int *num_dimms_per_channel,
  1146. int *num_channels)
  1147. {
  1148. u8 value;
  1149. /* Need to retrieve just how many channels and dimms per channel are
  1150. * supported on this memory controller
  1151. */
  1152. pci_read_config_byte(pdev, MAXDIMMPERCH, &value);
  1153. *num_dimms_per_channel = (int)value;
  1154. pci_read_config_byte(pdev, MAXCH, &value);
  1155. *num_channels = (int)value;
  1156. }
  1157. /*
  1158. * i5000_probe1 Probe for ONE instance of device to see if it is
  1159. * present.
  1160. * return:
  1161. * 0 for FOUND a device
  1162. * < 0 for error code
  1163. */
  1164. static int i5000_probe1(struct pci_dev *pdev, int dev_idx)
  1165. {
  1166. struct mem_ctl_info *mci;
  1167. struct edac_mc_layer layers[3];
  1168. struct i5000_pvt *pvt;
  1169. int num_channels;
  1170. int num_dimms_per_channel;
  1171. debugf0("MC: %s(), pdev bus %u dev=0x%x fn=0x%x\n",
  1172. __FILE__, pdev->bus->number,
  1173. PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
  1174. /* We only are looking for func 0 of the set */
  1175. if (PCI_FUNC(pdev->devfn) != 0)
  1176. return -ENODEV;
  1177. /* Ask the devices for the number of CSROWS and CHANNELS so
  1178. * that we can calculate the memory resources, etc
  1179. *
  1180. * The Chipset will report what it can handle which will be greater
  1181. * or equal to what the motherboard manufacturer will implement.
  1182. *
  1183. * As we don't have a motherboard identification routine to determine
  1184. * actual number of slots/dimms per channel, we thus utilize the
  1185. * resource as specified by the chipset. Thus, we might have
  1186. * have more DIMMs per channel than actually on the mobo, but this
  1187. * allows the driver to support up to the chipset max, without
  1188. * some fancy mobo determination.
  1189. */
  1190. i5000_get_dimm_and_channel_counts(pdev, &num_dimms_per_channel,
  1191. &num_channels);
  1192. debugf0("MC: Number of Branches=2 Channels= %d DIMMS= %d\n",
  1193. num_channels, num_dimms_per_channel);
  1194. /* allocate a new MC control structure */
  1195. layers[0].type = EDAC_MC_LAYER_BRANCH;
  1196. layers[0].size = MAX_BRANCHES;
  1197. layers[0].is_virt_csrow = false;
  1198. layers[1].type = EDAC_MC_LAYER_CHANNEL;
  1199. layers[1].size = num_channels / MAX_BRANCHES;
  1200. layers[1].is_virt_csrow = false;
  1201. layers[2].type = EDAC_MC_LAYER_SLOT;
  1202. layers[2].size = num_dimms_per_channel;
  1203. layers[2].is_virt_csrow = true;
  1204. mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, sizeof(*pvt));
  1205. if (mci == NULL)
  1206. return -ENOMEM;
  1207. debugf0("MC: %s(): mci = %p\n", __FILE__, mci);
  1208. mci->pdev = &pdev->dev; /* record ptr to the generic device */
  1209. pvt = mci->pvt_info;
  1210. pvt->system_address = pdev; /* Record this device in our private */
  1211. pvt->maxch = num_channels;
  1212. pvt->maxdimmperch = num_dimms_per_channel;
  1213. /* 'get' the pci devices we want to reserve for our use */
  1214. if (i5000_get_devices(mci, dev_idx))
  1215. goto fail0;
  1216. /* Time to get serious */
  1217. i5000_get_mc_regs(mci); /* retrieve the hardware registers */
  1218. mci->mc_idx = 0;
  1219. mci->mtype_cap = MEM_FLAG_FB_DDR2;
  1220. mci->edac_ctl_cap = EDAC_FLAG_NONE;
  1221. mci->edac_cap = EDAC_FLAG_NONE;
  1222. mci->mod_name = "i5000_edac.c";
  1223. mci->mod_ver = I5000_REVISION;
  1224. mci->ctl_name = i5000_devs[dev_idx].ctl_name;
  1225. mci->dev_name = pci_name(pdev);
  1226. mci->ctl_page_to_phys = NULL;
  1227. /* Set the function pointer to an actual operation function */
  1228. mci->edac_check = i5000_check_error;
  1229. /* initialize the MC control structure 'csrows' table
  1230. * with the mapping and control information */
  1231. if (i5000_init_csrows(mci)) {
  1232. debugf0("MC: Setting mci->edac_cap to EDAC_FLAG_NONE\n"
  1233. " because i5000_init_csrows() returned nonzero "
  1234. "value\n");
  1235. mci->edac_cap = EDAC_FLAG_NONE; /* no csrows found */
  1236. } else {
  1237. debugf1("MC: Enable error reporting now\n");
  1238. i5000_enable_error_reporting(mci);
  1239. }
  1240. /* add this new MC control structure to EDAC's list of MCs */
  1241. if (edac_mc_add_mc(mci)) {
  1242. debugf0("MC: %s(): failed edac_mc_add_mc()\n",
  1243. __FILE__);
  1244. /* FIXME: perhaps some code should go here that disables error
  1245. * reporting if we just enabled it
  1246. */
  1247. goto fail1;
  1248. }
  1249. i5000_clear_error(mci);
  1250. /* allocating generic PCI control info */
  1251. i5000_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
  1252. if (!i5000_pci) {
  1253. printk(KERN_WARNING
  1254. "%s(): Unable to create PCI control\n",
  1255. __func__);
  1256. printk(KERN_WARNING
  1257. "%s(): PCI error report via EDAC not setup\n",
  1258. __func__);
  1259. }
  1260. return 0;
  1261. /* Error exit unwinding stack */
  1262. fail1:
  1263. i5000_put_devices(mci);
  1264. fail0:
  1265. edac_mc_free(mci);
  1266. return -ENODEV;
  1267. }
  1268. /*
  1269. * i5000_init_one constructor for one instance of device
  1270. *
  1271. * returns:
  1272. * negative on error
  1273. * count (>= 0)
  1274. */
  1275. static int __devinit i5000_init_one(struct pci_dev *pdev,
  1276. const struct pci_device_id *id)
  1277. {
  1278. int rc;
  1279. debugf0("MC: %s()\n", __FILE__);
  1280. /* wake up device */
  1281. rc = pci_enable_device(pdev);
  1282. if (rc)
  1283. return rc;
  1284. /* now probe and enable the device */
  1285. return i5000_probe1(pdev, id->driver_data);
  1286. }
  1287. /*
  1288. * i5000_remove_one destructor for one instance of device
  1289. *
  1290. */
  1291. static void __devexit i5000_remove_one(struct pci_dev *pdev)
  1292. {
  1293. struct mem_ctl_info *mci;
  1294. debugf0("%s()\n", __FILE__);
  1295. if (i5000_pci)
  1296. edac_pci_release_generic_ctl(i5000_pci);
  1297. if ((mci = edac_mc_del_mc(&pdev->dev)) == NULL)
  1298. return;
  1299. /* retrieve references to resources, and free those resources */
  1300. i5000_put_devices(mci);
  1301. edac_mc_free(mci);
  1302. }
  1303. /*
  1304. * pci_device_id table for which devices we are looking for
  1305. *
  1306. * The "E500P" device is the first device supported.
  1307. */
  1308. static DEFINE_PCI_DEVICE_TABLE(i5000_pci_tbl) = {
  1309. {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_I5000_DEV16),
  1310. .driver_data = I5000P},
  1311. {0,} /* 0 terminated list. */
  1312. };
  1313. MODULE_DEVICE_TABLE(pci, i5000_pci_tbl);
  1314. /*
  1315. * i5000_driver pci_driver structure for this module
  1316. *
  1317. */
  1318. static struct pci_driver i5000_driver = {
  1319. .name = KBUILD_BASENAME,
  1320. .probe = i5000_init_one,
  1321. .remove = __devexit_p(i5000_remove_one),
  1322. .id_table = i5000_pci_tbl,
  1323. };
  1324. /*
  1325. * i5000_init Module entry function
  1326. * Try to initialize this module for its devices
  1327. */
  1328. static int __init i5000_init(void)
  1329. {
  1330. int pci_rc;
  1331. debugf2("MC: %s()\n", __FILE__);
  1332. /* Ensure that the OPSTATE is set correctly for POLL or NMI */
  1333. opstate_init();
  1334. pci_rc = pci_register_driver(&i5000_driver);
  1335. return (pci_rc < 0) ? pci_rc : 0;
  1336. }
  1337. /*
  1338. * i5000_exit() Module exit function
  1339. * Unregister the driver
  1340. */
  1341. static void __exit i5000_exit(void)
  1342. {
  1343. debugf2("MC: %s()\n", __FILE__);
  1344. pci_unregister_driver(&i5000_driver);
  1345. }
  1346. module_init(i5000_init);
  1347. module_exit(i5000_exit);
  1348. MODULE_LICENSE("GPL");
  1349. MODULE_AUTHOR
  1350. ("Linux Networx (http://lnxi.com) Doug Thompson <norsk5@xmission.com>");
  1351. MODULE_DESCRIPTION("MC Driver for Intel I5000 memory controllers - "
  1352. I5000_REVISION);
  1353. module_param(edac_op_state, int, 0444);
  1354. MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
  1355. module_param(misc_messages, int, 0444);
  1356. MODULE_PARM_DESC(misc_messages, "Log miscellaneous non fatal messages");