#tyndur · 2009-11-03

65 Nachrichten · Seite 1 von 1 · Original-CSV herunterladen

  1. pcmattmankwolf: ping
  2. kwolfpcmattman: pong
  3. pcmattmanI'm looking at the CDI e1000 driver at the moment, as it doesn't initialise properly on non-QEMU hosts
  4. pcmattmanDoes the driver work on anything other than QEMU in tyndur? Should I look at Pedigree's code or the driver code to try and fix it?
  5. kwolfIt's probably a driver issue
  6. kwolfI think it was reported to work in some other emulator, but never on real hardware
  7. kwolfI don't even have one in hardware, unfortunately
  8. pcmattmanOk, I'll see what I can do in the driver then
  9. kwolfAre you testing on real hardware?
  10. pcmattmanNot quite. Trying it in VirtualBox
  11. kwolfOk
  12. kwolfSo I should be able to reproduce it
  13. pcmattmanHopefully, yes
  14. kwolfMaybe you can tell me your finding before you leave tonight so I can have a look later
  15. pcmattmanIt actually appears to be a simple fix
  16. kwolfOh, already found it?
  17. pcmattmanhttp://slexy.org/view/s214Dqk530
  18. pcmattmanThat's the modified e1000_eeprom_read that is quite successful
  19. pcmattmanWith that modification I now successfully browse to Google using the driver
  20. kwolfHm... So you're actually failing the read or is it the sleep that makes it work?
  21. pcmattmanThe original code ran this in a tight loop
  22. kwolfI saw that, yes
  23. pcmattmanTo my understanding the 50 ms delay is enough for the card to complete the command
  24. kwolfStill don't understand why it wouldn't complete the command in a tight loop
  25. pcmattmanIt may well be that it's actually breaking out of the loop
  26. pcmattmanWhich would make sense, if it's persistently failing, it'll break out
  27. kwolfWhich means that you would return 0 instead of the right value
  28. pcmattmanBack to the drawing board :)
  29. pcmattmanYup
  30. kwolfThere is even the right printf in your code to check what happens
  31. pcmattmanIt looks like it's returning 0
  32. kwolfHm, 00:00:00:00:00:00 isn't really a desirable MAC address...
  33. pcmattmanYup
  34. pcmattmanIt seems other drivers attempt the EERD read and then if that fails attempt what looks like a "uwire" read
  35. kwolfMaybe the VBox e1000 doesn't even support this EEPROM interface? I think there is a second one
  36. kwolfA more complicated one which doesn't just involve reading out a register
  37. pcmattmanYup
  38. pcmattmanI'm going to try and obtain the specification and try and make sense of some driver code and see if I can port it across
  39. kwolfWhich is why I didn't bother to implement it ;-)
  40. kwolfThe e1000 sourceforge project has the spec, iirc
  41. pcmattmanThe question is though, which spec? :)
  42. kwolfIf in doubt, take the right one :-)
  43. pcmattmanComplex protocol
  44. kwolfFor EEPROM access?
  45. pcmattmanEERD (which is what the current driver uses) is dead easy
  46. pcmattmanThe 4-wire interface is a whole new story
  47. pcmattmanI'll be sure to submit a patch when I'm done :)
  48. kwolfGuess why I did it the EERD way...
  49. pcmattmanI do have e1000 driver code to refer to, thankfully
  50. kwolfAppropriately licensed code, I hope?
  51. pcmattmanYup, although the key word is "refer"
  52. pcmattmanAs far as I know the algorithm can't be protected by a license because it's a driver: the algorithm is defined by the specification
  53. kwolfI'm actually careful with even looking at Linux code
  54. kwolfSure
  55. kwolfafk lunch
  56. kwolfback
  57. pcmattmanWelcome back
  58. kwolfThanks
  59. kwolfWhat's the EEPROM doing?
  60. pcmattmanStill perusing code and the specification
  61. kwolfStill? This method sounds really tedious
  62. pcmattmanI just want to make sure I understand what I'm doing
  63. kwolfRight, it helps sometimes
  64. pcmattmanNight
  65. kwolfNight pc[away]