i5000_edac.c 42 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592
  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. /* 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. "Bank=%d RAS=%d CAS=%d FATAL Err=0x%x (%s)",
  468. bank, ras, cas, allErrors, specific);
  469. /* Call the helper to output message */
  470. edac_mc_handle_error(HW_EVENT_ERR_FATAL, mci, 0, 0, 0,
  471. branch >> 1, -1, rank,
  472. rdwr ? "Write error" : "Read error",
  473. msg, NULL);
  474. }
  475. /*
  476. * i5000_process_fatal_error_info(struct mem_ctl_info *mci,
  477. * struct i5000_error_info *info,
  478. * int handle_errors);
  479. *
  480. * handle the Intel NON-FATAL errors, if any
  481. */
  482. static void i5000_process_nonfatal_error_info(struct mem_ctl_info *mci,
  483. struct i5000_error_info *info,
  484. int handle_errors)
  485. {
  486. char msg[EDAC_MC_LABEL_LEN + 1 + 170];
  487. char *specific = NULL;
  488. u32 allErrors;
  489. u32 ue_errors;
  490. u32 ce_errors;
  491. u32 misc_errors;
  492. int branch;
  493. int channel;
  494. int bank;
  495. int rank;
  496. int rdwr;
  497. int ras, cas;
  498. /* mask off the Error bits that are possible */
  499. allErrors = (info->ferr_nf_fbd & FERR_NF_MASK);
  500. if (!allErrors)
  501. return; /* if no error, return now */
  502. /* ONLY ONE of the possible error bits will be set, as per the docs */
  503. ue_errors = allErrors & FERR_NF_UNCORRECTABLE;
  504. if (ue_errors) {
  505. debugf0("\tUncorrected bits= 0x%x\n", ue_errors);
  506. branch = EXTRACT_FBDCHAN_INDX(info->ferr_nf_fbd);
  507. /*
  508. * According with i5000 datasheet, bit 28 has no significance
  509. * for errors M4Err-M12Err and M17Err-M21Err, on FERR_NF_FBD
  510. */
  511. channel = branch & 2;
  512. bank = NREC_BANK(info->nrecmema);
  513. rank = NREC_RANK(info->nrecmema);
  514. rdwr = NREC_RDWR(info->nrecmema);
  515. ras = NREC_RAS(info->nrecmemb);
  516. cas = NREC_CAS(info->nrecmemb);
  517. debugf0
  518. ("\t\tCSROW= %d Channels= %d,%d (Branch= %d "
  519. "DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
  520. rank, channel, channel + 1, branch >> 1, bank,
  521. rdwr ? "Write" : "Read", ras, cas);
  522. switch (ue_errors) {
  523. case FERR_NF_M12ERR:
  524. specific = "Non-Aliased Uncorrectable Patrol Data ECC";
  525. break;
  526. case FERR_NF_M11ERR:
  527. specific = "Non-Aliased Uncorrectable Spare-Copy "
  528. "Data ECC";
  529. break;
  530. case FERR_NF_M10ERR:
  531. specific = "Non-Aliased Uncorrectable Mirrored Demand "
  532. "Data ECC";
  533. break;
  534. case FERR_NF_M9ERR:
  535. specific = "Non-Aliased Uncorrectable Non-Mirrored "
  536. "Demand Data ECC";
  537. break;
  538. case FERR_NF_M8ERR:
  539. specific = "Aliased Uncorrectable Patrol Data ECC";
  540. break;
  541. case FERR_NF_M7ERR:
  542. specific = "Aliased Uncorrectable Spare-Copy Data ECC";
  543. break;
  544. case FERR_NF_M6ERR:
  545. specific = "Aliased Uncorrectable Mirrored Demand "
  546. "Data ECC";
  547. break;
  548. case FERR_NF_M5ERR:
  549. specific = "Aliased Uncorrectable Non-Mirrored Demand "
  550. "Data ECC";
  551. break;
  552. case FERR_NF_M4ERR:
  553. specific = "Uncorrectable Data ECC on Replay";
  554. break;
  555. }
  556. /* Form out message */
  557. snprintf(msg, sizeof(msg),
  558. "Rank=%d Bank=%d RAS=%d CAS=%d, UE Err=0x%x (%s)",
  559. rank, bank, ras, cas, ue_errors, specific);
  560. /* Call the helper to output message */
  561. edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 0, 0, 0,
  562. channel >> 1, -1, rank,
  563. rdwr ? "Write error" : "Read error",
  564. msg, NULL);
  565. }
  566. /* Check correctable errors */
  567. ce_errors = allErrors & FERR_NF_CORRECTABLE;
  568. if (ce_errors) {
  569. debugf0("\tCorrected bits= 0x%x\n", ce_errors);
  570. branch = EXTRACT_FBDCHAN_INDX(info->ferr_nf_fbd);
  571. channel = 0;
  572. if (REC_ECC_LOCATOR_ODD(info->redmemb))
  573. channel = 1;
  574. /* Convert channel to be based from zero, instead of
  575. * from branch base of 0 */
  576. channel += branch;
  577. bank = REC_BANK(info->recmema);
  578. rank = REC_RANK(info->recmema);
  579. rdwr = REC_RDWR(info->recmema);
  580. ras = REC_RAS(info->recmemb);
  581. cas = REC_CAS(info->recmemb);
  582. debugf0("\t\tCSROW= %d Channel= %d (Branch %d "
  583. "DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
  584. rank, channel, branch >> 1, bank,
  585. rdwr ? "Write" : "Read", ras, cas);
  586. switch (ce_errors) {
  587. case FERR_NF_M17ERR:
  588. specific = "Correctable Non-Mirrored Demand Data ECC";
  589. break;
  590. case FERR_NF_M18ERR:
  591. specific = "Correctable Mirrored Demand Data ECC";
  592. break;
  593. case FERR_NF_M19ERR:
  594. specific = "Correctable Spare-Copy Data ECC";
  595. break;
  596. case FERR_NF_M20ERR:
  597. specific = "Correctable Patrol Data ECC";
  598. break;
  599. }
  600. /* Form out message */
  601. snprintf(msg, sizeof(msg),
  602. "Rank=%d Bank=%d RDWR=%s RAS=%d "
  603. "CAS=%d, CE Err=0x%x (%s))", branch >> 1, bank,
  604. rdwr ? "Write" : "Read", ras, cas, ce_errors,
  605. specific);
  606. /* Call the helper to output message */
  607. edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 0, 0, 0,
  608. channel >> 1, channel % 2, rank,
  609. rdwr ? "Write error" : "Read error",
  610. msg, NULL);
  611. }
  612. if (!misc_messages)
  613. return;
  614. misc_errors = allErrors & (FERR_NF_NON_RETRY | FERR_NF_NORTH_CRC |
  615. FERR_NF_SPD_PROTOCOL | FERR_NF_DIMM_SPARE);
  616. if (misc_errors) {
  617. switch (misc_errors) {
  618. case FERR_NF_M13ERR:
  619. specific = "Non-Retry or Redundant Retry FBD Memory "
  620. "Alert or Redundant Fast Reset Timeout";
  621. break;
  622. case FERR_NF_M14ERR:
  623. specific = "Non-Retry or Redundant Retry FBD "
  624. "Configuration Alert";
  625. break;
  626. case FERR_NF_M15ERR:
  627. specific = "Non-Retry or Redundant Retry FBD "
  628. "Northbound CRC error on read data";
  629. break;
  630. case FERR_NF_M21ERR:
  631. specific = "FBD Northbound CRC error on "
  632. "FBD Sync Status";
  633. break;
  634. case FERR_NF_M22ERR:
  635. specific = "SPD protocol error";
  636. break;
  637. case FERR_NF_M27ERR:
  638. specific = "DIMM-spare copy started";
  639. break;
  640. case FERR_NF_M28ERR:
  641. specific = "DIMM-spare copy completed";
  642. break;
  643. }
  644. branch = EXTRACT_FBDCHAN_INDX(info->ferr_nf_fbd);
  645. /* Form out message */
  646. snprintf(msg, sizeof(msg),
  647. "Err=%#x (%s)", misc_errors, specific);
  648. /* Call the helper to output message */
  649. edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 0, 0, 0,
  650. branch >> 1, -1, -1,
  651. "Misc error", msg, NULL);
  652. }
  653. }
  654. /*
  655. * i5000_process_error_info Process the error info that is
  656. * in the 'info' structure, previously retrieved from hardware
  657. */
  658. static void i5000_process_error_info(struct mem_ctl_info *mci,
  659. struct i5000_error_info *info,
  660. int handle_errors)
  661. {
  662. /* First handle any fatal errors that occurred */
  663. i5000_process_fatal_error_info(mci, info, handle_errors);
  664. /* now handle any non-fatal errors that occurred */
  665. i5000_process_nonfatal_error_info(mci, info, handle_errors);
  666. }
  667. /*
  668. * i5000_clear_error Retrieve any error from the hardware
  669. * but do NOT process that error.
  670. * Used for 'clearing' out of previous errors
  671. * Called by the Core module.
  672. */
  673. static void i5000_clear_error(struct mem_ctl_info *mci)
  674. {
  675. struct i5000_error_info info;
  676. i5000_get_error_info(mci, &info);
  677. }
  678. /*
  679. * i5000_check_error Retrieve and process errors reported by the
  680. * hardware. Called by the Core module.
  681. */
  682. static void i5000_check_error(struct mem_ctl_info *mci)
  683. {
  684. struct i5000_error_info info;
  685. debugf4("MC%d: %s: %s()\n", mci->mc_idx, __FILE__, __func__);
  686. i5000_get_error_info(mci, &info);
  687. i5000_process_error_info(mci, &info, 1);
  688. }
  689. /*
  690. * i5000_get_devices Find and perform 'get' operation on the MCH's
  691. * device/functions we want to reference for this driver
  692. *
  693. * Need to 'get' device 16 func 1 and func 2
  694. */
  695. static int i5000_get_devices(struct mem_ctl_info *mci, int dev_idx)
  696. {
  697. //const struct i5000_dev_info *i5000_dev = &i5000_devs[dev_idx];
  698. struct i5000_pvt *pvt;
  699. struct pci_dev *pdev;
  700. pvt = mci->pvt_info;
  701. /* Attempt to 'get' the MCH register we want */
  702. pdev = NULL;
  703. while (1) {
  704. pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
  705. PCI_DEVICE_ID_INTEL_I5000_DEV16, pdev);
  706. /* End of list, leave */
  707. if (pdev == NULL) {
  708. i5000_printk(KERN_ERR,
  709. "'system address,Process Bus' "
  710. "device not found:"
  711. "vendor 0x%x device 0x%x FUNC 1 "
  712. "(broken BIOS?)\n",
  713. PCI_VENDOR_ID_INTEL,
  714. PCI_DEVICE_ID_INTEL_I5000_DEV16);
  715. return 1;
  716. }
  717. /* Scan for device 16 func 1 */
  718. if (PCI_FUNC(pdev->devfn) == 1)
  719. break;
  720. }
  721. pvt->branchmap_werrors = pdev;
  722. /* Attempt to 'get' the MCH register we want */
  723. pdev = NULL;
  724. while (1) {
  725. pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
  726. PCI_DEVICE_ID_INTEL_I5000_DEV16, pdev);
  727. if (pdev == NULL) {
  728. i5000_printk(KERN_ERR,
  729. "MC: 'branchmap,control,errors' "
  730. "device not found:"
  731. "vendor 0x%x device 0x%x Func 2 "
  732. "(broken BIOS?)\n",
  733. PCI_VENDOR_ID_INTEL,
  734. PCI_DEVICE_ID_INTEL_I5000_DEV16);
  735. pci_dev_put(pvt->branchmap_werrors);
  736. return 1;
  737. }
  738. /* Scan for device 16 func 1 */
  739. if (PCI_FUNC(pdev->devfn) == 2)
  740. break;
  741. }
  742. pvt->fsb_error_regs = pdev;
  743. debugf1("System Address, processor bus- PCI Bus ID: %s %x:%x\n",
  744. pci_name(pvt->system_address),
  745. pvt->system_address->vendor, pvt->system_address->device);
  746. debugf1("Branchmap, control and errors - PCI Bus ID: %s %x:%x\n",
  747. pci_name(pvt->branchmap_werrors),
  748. pvt->branchmap_werrors->vendor, pvt->branchmap_werrors->device);
  749. debugf1("FSB Error Regs - PCI Bus ID: %s %x:%x\n",
  750. pci_name(pvt->fsb_error_regs),
  751. pvt->fsb_error_regs->vendor, pvt->fsb_error_regs->device);
  752. pdev = NULL;
  753. pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
  754. PCI_DEVICE_ID_I5000_BRANCH_0, pdev);
  755. if (pdev == NULL) {
  756. i5000_printk(KERN_ERR,
  757. "MC: 'BRANCH 0' device not found:"
  758. "vendor 0x%x device 0x%x Func 0 (broken BIOS?)\n",
  759. PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_I5000_BRANCH_0);
  760. pci_dev_put(pvt->branchmap_werrors);
  761. pci_dev_put(pvt->fsb_error_regs);
  762. return 1;
  763. }
  764. pvt->branch_0 = pdev;
  765. /* If this device claims to have more than 2 channels then
  766. * fetch Branch 1's information
  767. */
  768. if (pvt->maxch >= CHANNELS_PER_BRANCH) {
  769. pdev = NULL;
  770. pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
  771. PCI_DEVICE_ID_I5000_BRANCH_1, pdev);
  772. if (pdev == NULL) {
  773. i5000_printk(KERN_ERR,
  774. "MC: 'BRANCH 1' device not found:"
  775. "vendor 0x%x device 0x%x Func 0 "
  776. "(broken BIOS?)\n",
  777. PCI_VENDOR_ID_INTEL,
  778. PCI_DEVICE_ID_I5000_BRANCH_1);
  779. pci_dev_put(pvt->branchmap_werrors);
  780. pci_dev_put(pvt->fsb_error_regs);
  781. pci_dev_put(pvt->branch_0);
  782. return 1;
  783. }
  784. pvt->branch_1 = pdev;
  785. }
  786. return 0;
  787. }
  788. /*
  789. * i5000_put_devices 'put' all the devices that we have
  790. * reserved via 'get'
  791. */
  792. static void i5000_put_devices(struct mem_ctl_info *mci)
  793. {
  794. struct i5000_pvt *pvt;
  795. pvt = mci->pvt_info;
  796. pci_dev_put(pvt->branchmap_werrors); /* FUNC 1 */
  797. pci_dev_put(pvt->fsb_error_regs); /* FUNC 2 */
  798. pci_dev_put(pvt->branch_0); /* DEV 21 */
  799. /* Only if more than 2 channels do we release the second branch */
  800. if (pvt->maxch >= CHANNELS_PER_BRANCH)
  801. pci_dev_put(pvt->branch_1); /* DEV 22 */
  802. }
  803. /*
  804. * determine_amb_resent
  805. *
  806. * the information is contained in NUM_MTRS different registers
  807. * determineing which of the NUM_MTRS requires knowing
  808. * which channel is in question
  809. *
  810. * 2 branches, each with 2 channels
  811. * b0_ambpresent0 for channel '0'
  812. * b0_ambpresent1 for channel '1'
  813. * b1_ambpresent0 for channel '2'
  814. * b1_ambpresent1 for channel '3'
  815. */
  816. static int determine_amb_present_reg(struct i5000_pvt *pvt, int channel)
  817. {
  818. int amb_present;
  819. if (channel < CHANNELS_PER_BRANCH) {
  820. if (channel & 0x1)
  821. amb_present = pvt->b0_ambpresent1;
  822. else
  823. amb_present = pvt->b0_ambpresent0;
  824. } else {
  825. if (channel & 0x1)
  826. amb_present = pvt->b1_ambpresent1;
  827. else
  828. amb_present = pvt->b1_ambpresent0;
  829. }
  830. return amb_present;
  831. }
  832. /*
  833. * determine_mtr(pvt, csrow, channel)
  834. *
  835. * return the proper MTR register as determine by the csrow and channel desired
  836. */
  837. static int determine_mtr(struct i5000_pvt *pvt, int csrow, int channel)
  838. {
  839. int mtr;
  840. if (channel < CHANNELS_PER_BRANCH)
  841. mtr = pvt->b0_mtr[csrow >> 1];
  842. else
  843. mtr = pvt->b1_mtr[csrow >> 1];
  844. return mtr;
  845. }
  846. /*
  847. */
  848. static void decode_mtr(int slot_row, u16 mtr)
  849. {
  850. int ans;
  851. ans = MTR_DIMMS_PRESENT(mtr);
  852. debugf2("\tMTR%d=0x%x: DIMMs are %s\n", slot_row, mtr,
  853. ans ? "Present" : "NOT Present");
  854. if (!ans)
  855. return;
  856. debugf2("\t\tWIDTH: x%d\n", MTR_DRAM_WIDTH(mtr));
  857. debugf2("\t\tNUMBANK: %d bank(s)\n", MTR_DRAM_BANKS(mtr));
  858. debugf2("\t\tNUMRANK: %s\n", MTR_DIMM_RANK(mtr) ? "double" : "single");
  859. debugf2("\t\tNUMROW: %s\n", numrow_toString[MTR_DIMM_ROWS(mtr)]);
  860. debugf2("\t\tNUMCOL: %s\n", numcol_toString[MTR_DIMM_COLS(mtr)]);
  861. }
  862. static void handle_channel(struct i5000_pvt *pvt, int csrow, int channel,
  863. struct i5000_dimm_info *dinfo)
  864. {
  865. int mtr;
  866. int amb_present_reg;
  867. int addrBits;
  868. mtr = determine_mtr(pvt, csrow, channel);
  869. if (MTR_DIMMS_PRESENT(mtr)) {
  870. amb_present_reg = determine_amb_present_reg(pvt, channel);
  871. /* Determine if there is a DIMM present in this DIMM slot */
  872. if (amb_present_reg & (1 << (csrow >> 1))) {
  873. dinfo->dual_rank = MTR_DIMM_RANK(mtr);
  874. if (!((dinfo->dual_rank == 0) &&
  875. ((csrow & 0x1) == 0x1))) {
  876. /* Start with the number of bits for a Bank
  877. * on the DRAM */
  878. addrBits = MTR_DRAM_BANKS_ADDR_BITS(mtr);
  879. /* Add thenumber of ROW bits */
  880. addrBits += MTR_DIMM_ROWS_ADDR_BITS(mtr);
  881. /* add the number of COLUMN bits */
  882. addrBits += MTR_DIMM_COLS_ADDR_BITS(mtr);
  883. addrBits += 6; /* add 64 bits per DIMM */
  884. addrBits -= 20; /* divide by 2^^20 */
  885. addrBits -= 3; /* 8 bits per bytes */
  886. dinfo->megabytes = 1 << addrBits;
  887. }
  888. }
  889. }
  890. }
  891. /*
  892. * calculate_dimm_size
  893. *
  894. * also will output a DIMM matrix map, if debug is enabled, for viewing
  895. * how the DIMMs are populated
  896. */
  897. static void calculate_dimm_size(struct i5000_pvt *pvt)
  898. {
  899. struct i5000_dimm_info *dinfo;
  900. int csrow, max_csrows;
  901. char *p, *mem_buffer;
  902. int space, n;
  903. int channel;
  904. /* ================= Generate some debug output ================= */
  905. space = PAGE_SIZE;
  906. mem_buffer = p = kmalloc(space, GFP_KERNEL);
  907. if (p == NULL) {
  908. i5000_printk(KERN_ERR, "MC: %s:%s() kmalloc() failed\n",
  909. __FILE__, __func__);
  910. return;
  911. }
  912. n = snprintf(p, space, "\n");
  913. p += n;
  914. space -= n;
  915. /* Scan all the actual CSROWS (which is # of DIMMS * 2)
  916. * and calculate the information for each DIMM
  917. * Start with the highest csrow first, to display it first
  918. * and work toward the 0th csrow
  919. */
  920. max_csrows = pvt->maxdimmperch * 2;
  921. for (csrow = max_csrows - 1; csrow >= 0; csrow--) {
  922. /* on an odd csrow, first output a 'boundary' marker,
  923. * then reset the message buffer */
  924. if (csrow & 0x1) {
  925. n = snprintf(p, space, "---------------------------"
  926. "--------------------------------");
  927. p += n;
  928. space -= n;
  929. debugf2("%s\n", mem_buffer);
  930. p = mem_buffer;
  931. space = PAGE_SIZE;
  932. }
  933. n = snprintf(p, space, "csrow %2d ", csrow);
  934. p += n;
  935. space -= n;
  936. for (channel = 0; channel < pvt->maxch; channel++) {
  937. dinfo = &pvt->dimm_info[csrow][channel];
  938. handle_channel(pvt, csrow, channel, dinfo);
  939. n = snprintf(p, space, "%4d MB | ", dinfo->megabytes);
  940. p += n;
  941. space -= n;
  942. }
  943. n = snprintf(p, space, "\n");
  944. p += n;
  945. space -= n;
  946. }
  947. /* Output the last bottom 'boundary' marker */
  948. n = snprintf(p, space, "---------------------------"
  949. "--------------------------------\n");
  950. p += n;
  951. space -= n;
  952. /* now output the 'channel' labels */
  953. n = snprintf(p, space, " ");
  954. p += n;
  955. space -= n;
  956. for (channel = 0; channel < pvt->maxch; channel++) {
  957. n = snprintf(p, space, "channel %d | ", channel);
  958. p += n;
  959. space -= n;
  960. }
  961. n = snprintf(p, space, "\n");
  962. p += n;
  963. space -= n;
  964. /* output the last message and free buffer */
  965. debugf2("%s\n", mem_buffer);
  966. kfree(mem_buffer);
  967. }
  968. /*
  969. * i5000_get_mc_regs read in the necessary registers and
  970. * cache locally
  971. *
  972. * Fills in the private data members
  973. */
  974. static void i5000_get_mc_regs(struct mem_ctl_info *mci)
  975. {
  976. struct i5000_pvt *pvt;
  977. u32 actual_tolm;
  978. u16 limit;
  979. int slot_row;
  980. int maxch;
  981. int maxdimmperch;
  982. int way0, way1;
  983. pvt = mci->pvt_info;
  984. pci_read_config_dword(pvt->system_address, AMBASE,
  985. (u32 *) & pvt->ambase);
  986. pci_read_config_dword(pvt->system_address, AMBASE + sizeof(u32),
  987. ((u32 *) & pvt->ambase) + sizeof(u32));
  988. maxdimmperch = pvt->maxdimmperch;
  989. maxch = pvt->maxch;
  990. debugf2("AMBASE= 0x%lx MAXCH= %d MAX-DIMM-Per-CH= %d\n",
  991. (long unsigned int)pvt->ambase, pvt->maxch, pvt->maxdimmperch);
  992. /* Get the Branch Map regs */
  993. pci_read_config_word(pvt->branchmap_werrors, TOLM, &pvt->tolm);
  994. pvt->tolm >>= 12;
  995. debugf2("\nTOLM (number of 256M regions) =%u (0x%x)\n", pvt->tolm,
  996. pvt->tolm);
  997. actual_tolm = pvt->tolm << 28;
  998. debugf2("Actual TOLM byte addr=%u (0x%x)\n", actual_tolm, actual_tolm);
  999. pci_read_config_word(pvt->branchmap_werrors, MIR0, &pvt->mir0);
  1000. pci_read_config_word(pvt->branchmap_werrors, MIR1, &pvt->mir1);
  1001. pci_read_config_word(pvt->branchmap_werrors, MIR2, &pvt->mir2);
  1002. /* Get the MIR[0-2] regs */
  1003. limit = (pvt->mir0 >> 4) & 0x0FFF;
  1004. way0 = pvt->mir0 & 0x1;
  1005. way1 = pvt->mir0 & 0x2;
  1006. debugf2("MIR0: limit= 0x%x WAY1= %u WAY0= %x\n", limit, way1, way0);
  1007. limit = (pvt->mir1 >> 4) & 0x0FFF;
  1008. way0 = pvt->mir1 & 0x1;
  1009. way1 = pvt->mir1 & 0x2;
  1010. debugf2("MIR1: limit= 0x%x WAY1= %u WAY0= %x\n", limit, way1, way0);
  1011. limit = (pvt->mir2 >> 4) & 0x0FFF;
  1012. way0 = pvt->mir2 & 0x1;
  1013. way1 = pvt->mir2 & 0x2;
  1014. debugf2("MIR2: limit= 0x%x WAY1= %u WAY0= %x\n", limit, way1, way0);
  1015. /* Get the MTR[0-3] regs */
  1016. for (slot_row = 0; slot_row < NUM_MTRS; slot_row++) {
  1017. int where = MTR0 + (slot_row * sizeof(u32));
  1018. pci_read_config_word(pvt->branch_0, where,
  1019. &pvt->b0_mtr[slot_row]);
  1020. debugf2("MTR%d where=0x%x B0 value=0x%x\n", slot_row, where,
  1021. pvt->b0_mtr[slot_row]);
  1022. if (pvt->maxch >= CHANNELS_PER_BRANCH) {
  1023. pci_read_config_word(pvt->branch_1, where,
  1024. &pvt->b1_mtr[slot_row]);
  1025. debugf2("MTR%d where=0x%x B1 value=0x%x\n", slot_row,
  1026. where, pvt->b1_mtr[slot_row]);
  1027. } else {
  1028. pvt->b1_mtr[slot_row] = 0;
  1029. }
  1030. }
  1031. /* Read and dump branch 0's MTRs */
  1032. debugf2("\nMemory Technology Registers:\n");
  1033. debugf2(" Branch 0:\n");
  1034. for (slot_row = 0; slot_row < NUM_MTRS; slot_row++) {
  1035. decode_mtr(slot_row, pvt->b0_mtr[slot_row]);
  1036. }
  1037. pci_read_config_word(pvt->branch_0, AMB_PRESENT_0,
  1038. &pvt->b0_ambpresent0);
  1039. debugf2("\t\tAMB-Branch 0-present0 0x%x:\n", pvt->b0_ambpresent0);
  1040. pci_read_config_word(pvt->branch_0, AMB_PRESENT_1,
  1041. &pvt->b0_ambpresent1);
  1042. debugf2("\t\tAMB-Branch 0-present1 0x%x:\n", pvt->b0_ambpresent1);
  1043. /* Only if we have 2 branchs (4 channels) */
  1044. if (pvt->maxch < CHANNELS_PER_BRANCH) {
  1045. pvt->b1_ambpresent0 = 0;
  1046. pvt->b1_ambpresent1 = 0;
  1047. } else {
  1048. /* Read and dump branch 1's MTRs */
  1049. debugf2(" Branch 1:\n");
  1050. for (slot_row = 0; slot_row < NUM_MTRS; slot_row++) {
  1051. decode_mtr(slot_row, pvt->b1_mtr[slot_row]);
  1052. }
  1053. pci_read_config_word(pvt->branch_1, AMB_PRESENT_0,
  1054. &pvt->b1_ambpresent0);
  1055. debugf2("\t\tAMB-Branch 1-present0 0x%x:\n",
  1056. pvt->b1_ambpresent0);
  1057. pci_read_config_word(pvt->branch_1, AMB_PRESENT_1,
  1058. &pvt->b1_ambpresent1);
  1059. debugf2("\t\tAMB-Branch 1-present1 0x%x:\n",
  1060. pvt->b1_ambpresent1);
  1061. }
  1062. /* Go and determine the size of each DIMM and place in an
  1063. * orderly matrix */
  1064. calculate_dimm_size(pvt);
  1065. }
  1066. /*
  1067. * i5000_init_csrows Initialize the 'csrows' table within
  1068. * the mci control structure with the
  1069. * addressing of memory.
  1070. *
  1071. * return:
  1072. * 0 success
  1073. * 1 no actual memory found on this MC
  1074. */
  1075. static int i5000_init_csrows(struct mem_ctl_info *mci)
  1076. {
  1077. struct i5000_pvt *pvt;
  1078. struct csrow_info *p_csrow;
  1079. struct dimm_info *dimm;
  1080. int empty, channel_count;
  1081. int max_csrows;
  1082. int mtr, mtr1;
  1083. int csrow_megs;
  1084. int channel;
  1085. int csrow;
  1086. pvt = mci->pvt_info;
  1087. channel_count = pvt->maxch;
  1088. max_csrows = pvt->maxdimmperch * 2;
  1089. empty = 1; /* Assume NO memory */
  1090. /*
  1091. * TODO: it would be better to not use csrow here, filling
  1092. * directly the dimm_info structs, based on branch, channel, dim number
  1093. */
  1094. for (csrow = 0; csrow < max_csrows; csrow++) {
  1095. p_csrow = &mci->csrows[csrow];
  1096. p_csrow->csrow_idx = csrow;
  1097. /* use branch 0 for the basis */
  1098. mtr = pvt->b0_mtr[csrow >> 1];
  1099. mtr1 = pvt->b1_mtr[csrow >> 1];
  1100. /* if no DIMMS on this row, continue */
  1101. if (!MTR_DIMMS_PRESENT(mtr) && !MTR_DIMMS_PRESENT(mtr1))
  1102. continue;
  1103. csrow_megs = 0;
  1104. for (channel = 0; channel < pvt->maxch; channel++) {
  1105. dimm = p_csrow->channels[channel].dimm;
  1106. csrow_megs += pvt->dimm_info[csrow][channel].megabytes;
  1107. dimm->grain = 8;
  1108. /* Assume DDR2 for now */
  1109. dimm->mtype = MEM_FB_DDR2;
  1110. /* ask what device type on this row */
  1111. if (MTR_DRAM_WIDTH(mtr))
  1112. dimm->dtype = DEV_X8;
  1113. else
  1114. dimm->dtype = DEV_X4;
  1115. dimm->edac_mode = EDAC_S8ECD8ED;
  1116. dimm->nr_pages = (csrow_megs << 8) / pvt->maxch;
  1117. }
  1118. empty = 0;
  1119. }
  1120. return empty;
  1121. }
  1122. /*
  1123. * i5000_enable_error_reporting
  1124. * Turn on the memory reporting features of the hardware
  1125. */
  1126. static void i5000_enable_error_reporting(struct mem_ctl_info *mci)
  1127. {
  1128. struct i5000_pvt *pvt;
  1129. u32 fbd_error_mask;
  1130. pvt = mci->pvt_info;
  1131. /* Read the FBD Error Mask Register */
  1132. pci_read_config_dword(pvt->branchmap_werrors, EMASK_FBD,
  1133. &fbd_error_mask);
  1134. /* Enable with a '0' */
  1135. fbd_error_mask &= ~(ENABLE_EMASK_ALL);
  1136. pci_write_config_dword(pvt->branchmap_werrors, EMASK_FBD,
  1137. fbd_error_mask);
  1138. }
  1139. /*
  1140. * i5000_get_dimm_and_channel_counts(pdev, &nr_csrows, &num_channels)
  1141. *
  1142. * ask the device how many channels are present and how many CSROWS
  1143. * as well
  1144. */
  1145. static void i5000_get_dimm_and_channel_counts(struct pci_dev *pdev,
  1146. int *num_dimms_per_channel,
  1147. int *num_channels)
  1148. {
  1149. u8 value;
  1150. /* Need to retrieve just how many channels and dimms per channel are
  1151. * supported on this memory controller
  1152. */
  1153. pci_read_config_byte(pdev, MAXDIMMPERCH, &value);
  1154. *num_dimms_per_channel = (int)value *2;
  1155. pci_read_config_byte(pdev, MAXCH, &value);
  1156. *num_channels = (int)value;
  1157. }
  1158. /*
  1159. * i5000_probe1 Probe for ONE instance of device to see if it is
  1160. * present.
  1161. * return:
  1162. * 0 for FOUND a device
  1163. * < 0 for error code
  1164. */
  1165. static int i5000_probe1(struct pci_dev *pdev, int dev_idx)
  1166. {
  1167. struct mem_ctl_info *mci;
  1168. struct edac_mc_layer layers[3];
  1169. struct i5000_pvt *pvt;
  1170. int num_channels;
  1171. int num_dimms_per_channel;
  1172. debugf0("MC: %s: %s(), pdev bus %u dev=0x%x fn=0x%x\n",
  1173. __FILE__, __func__,
  1174. pdev->bus->number,
  1175. PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
  1176. /* We only are looking for func 0 of the set */
  1177. if (PCI_FUNC(pdev->devfn) != 0)
  1178. return -ENODEV;
  1179. /* Ask the devices for the number of CSROWS and CHANNELS so
  1180. * that we can calculate the memory resources, etc
  1181. *
  1182. * The Chipset will report what it can handle which will be greater
  1183. * or equal to what the motherboard manufacturer will implement.
  1184. *
  1185. * As we don't have a motherboard identification routine to determine
  1186. * actual number of slots/dimms per channel, we thus utilize the
  1187. * resource as specified by the chipset. Thus, we might have
  1188. * have more DIMMs per channel than actually on the mobo, but this
  1189. * allows the driver to support up to the chipset max, without
  1190. * some fancy mobo determination.
  1191. */
  1192. i5000_get_dimm_and_channel_counts(pdev, &num_dimms_per_channel,
  1193. &num_channels);
  1194. debugf0("MC: %s(): Number of Branches=2 Channels= %d DIMMS= %d\n",
  1195. __func__, num_channels, num_dimms_per_channel);
  1196. /* allocate a new MC control structure */
  1197. layers[0].type = EDAC_MC_LAYER_BRANCH;
  1198. layers[0].size = 2;
  1199. layers[0].is_virt_csrow = true;
  1200. layers[1].type = EDAC_MC_LAYER_CHANNEL;
  1201. layers[1].size = num_channels;
  1202. layers[1].is_virt_csrow = false;
  1203. layers[2].type = EDAC_MC_LAYER_SLOT;
  1204. layers[2].size = num_dimms_per_channel;
  1205. layers[2].is_virt_csrow = true;
  1206. mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, sizeof(*pvt));
  1207. if (mci == NULL)
  1208. return -ENOMEM;
  1209. kobject_get(&mci->edac_mci_kobj);
  1210. debugf0("MC: %s: %s(): mci = %p\n", __FILE__, __func__, mci);
  1211. mci->dev = &pdev->dev; /* record ptr to the generic device */
  1212. pvt = mci->pvt_info;
  1213. pvt->system_address = pdev; /* Record this device in our private */
  1214. pvt->maxch = num_channels;
  1215. pvt->maxdimmperch = num_dimms_per_channel;
  1216. /* 'get' the pci devices we want to reserve for our use */
  1217. if (i5000_get_devices(mci, dev_idx))
  1218. goto fail0;
  1219. /* Time to get serious */
  1220. i5000_get_mc_regs(mci); /* retrieve the hardware registers */
  1221. mci->mc_idx = 0;
  1222. mci->mtype_cap = MEM_FLAG_FB_DDR2;
  1223. mci->edac_ctl_cap = EDAC_FLAG_NONE;
  1224. mci->edac_cap = EDAC_FLAG_NONE;
  1225. mci->mod_name = "i5000_edac.c";
  1226. mci->mod_ver = I5000_REVISION;
  1227. mci->ctl_name = i5000_devs[dev_idx].ctl_name;
  1228. mci->dev_name = pci_name(pdev);
  1229. mci->ctl_page_to_phys = NULL;
  1230. /* Set the function pointer to an actual operation function */
  1231. mci->edac_check = i5000_check_error;
  1232. /* initialize the MC control structure 'csrows' table
  1233. * with the mapping and control information */
  1234. if (i5000_init_csrows(mci)) {
  1235. debugf0("MC: Setting mci->edac_cap to EDAC_FLAG_NONE\n"
  1236. " because i5000_init_csrows() returned nonzero "
  1237. "value\n");
  1238. mci->edac_cap = EDAC_FLAG_NONE; /* no csrows found */
  1239. } else {
  1240. debugf1("MC: Enable error reporting now\n");
  1241. i5000_enable_error_reporting(mci);
  1242. }
  1243. /* add this new MC control structure to EDAC's list of MCs */
  1244. if (edac_mc_add_mc(mci)) {
  1245. debugf0("MC: %s: %s(): failed edac_mc_add_mc()\n",
  1246. __FILE__, __func__);
  1247. /* FIXME: perhaps some code should go here that disables error
  1248. * reporting if we just enabled it
  1249. */
  1250. goto fail1;
  1251. }
  1252. i5000_clear_error(mci);
  1253. /* allocating generic PCI control info */
  1254. i5000_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
  1255. if (!i5000_pci) {
  1256. printk(KERN_WARNING
  1257. "%s(): Unable to create PCI control\n",
  1258. __func__);
  1259. printk(KERN_WARNING
  1260. "%s(): PCI error report via EDAC not setup\n",
  1261. __func__);
  1262. }
  1263. return 0;
  1264. /* Error exit unwinding stack */
  1265. fail1:
  1266. i5000_put_devices(mci);
  1267. fail0:
  1268. kobject_put(&mci->edac_mci_kobj);
  1269. edac_mc_free(mci);
  1270. return -ENODEV;
  1271. }
  1272. /*
  1273. * i5000_init_one constructor for one instance of device
  1274. *
  1275. * returns:
  1276. * negative on error
  1277. * count (>= 0)
  1278. */
  1279. static int __devinit i5000_init_one(struct pci_dev *pdev,
  1280. const struct pci_device_id *id)
  1281. {
  1282. int rc;
  1283. debugf0("MC: %s: %s()\n", __FILE__, __func__);
  1284. /* wake up device */
  1285. rc = pci_enable_device(pdev);
  1286. if (rc)
  1287. return rc;
  1288. /* now probe and enable the device */
  1289. return i5000_probe1(pdev, id->driver_data);
  1290. }
  1291. /*
  1292. * i5000_remove_one destructor for one instance of device
  1293. *
  1294. */
  1295. static void __devexit i5000_remove_one(struct pci_dev *pdev)
  1296. {
  1297. struct mem_ctl_info *mci;
  1298. debugf0("%s: %s()\n", __FILE__, __func__);
  1299. if (i5000_pci)
  1300. edac_pci_release_generic_ctl(i5000_pci);
  1301. if ((mci = edac_mc_del_mc(&pdev->dev)) == NULL)
  1302. return;
  1303. /* retrieve references to resources, and free those resources */
  1304. i5000_put_devices(mci);
  1305. kobject_put(&mci->edac_mci_kobj);
  1306. edac_mc_free(mci);
  1307. }
  1308. /*
  1309. * pci_device_id table for which devices we are looking for
  1310. *
  1311. * The "E500P" device is the first device supported.
  1312. */
  1313. static DEFINE_PCI_DEVICE_TABLE(i5000_pci_tbl) = {
  1314. {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_I5000_DEV16),
  1315. .driver_data = I5000P},
  1316. {0,} /* 0 terminated list. */
  1317. };
  1318. MODULE_DEVICE_TABLE(pci, i5000_pci_tbl);
  1319. /*
  1320. * i5000_driver pci_driver structure for this module
  1321. *
  1322. */
  1323. static struct pci_driver i5000_driver = {
  1324. .name = KBUILD_BASENAME,
  1325. .probe = i5000_init_one,
  1326. .remove = __devexit_p(i5000_remove_one),
  1327. .id_table = i5000_pci_tbl,
  1328. };
  1329. /*
  1330. * i5000_init Module entry function
  1331. * Try to initialize this module for its devices
  1332. */
  1333. static int __init i5000_init(void)
  1334. {
  1335. int pci_rc;
  1336. debugf2("MC: %s: %s()\n", __FILE__, __func__);
  1337. /* Ensure that the OPSTATE is set correctly for POLL or NMI */
  1338. opstate_init();
  1339. pci_rc = pci_register_driver(&i5000_driver);
  1340. return (pci_rc < 0) ? pci_rc : 0;
  1341. }
  1342. /*
  1343. * i5000_exit() Module exit function
  1344. * Unregister the driver
  1345. */
  1346. static void __exit i5000_exit(void)
  1347. {
  1348. debugf2("MC: %s: %s()\n", __FILE__, __func__);
  1349. pci_unregister_driver(&i5000_driver);
  1350. }
  1351. module_init(i5000_init);
  1352. module_exit(i5000_exit);
  1353. MODULE_LICENSE("GPL");
  1354. MODULE_AUTHOR
  1355. ("Linux Networx (http://lnxi.com) Doug Thompson <norsk5@xmission.com>");
  1356. MODULE_DESCRIPTION("MC Driver for Intel I5000 memory controllers - "
  1357. I5000_REVISION);
  1358. module_param(edac_op_state, int, 0444);
  1359. MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
  1360. module_param(misc_messages, int, 0444);
  1361. MODULE_PARM_DESC(misc_messages, "Log miscellaneous non fatal messages");