#tyndur · 2010-02-12

536 Nachrichten · Seite 2 von 3 · Original-CSV herunterladen

  1. eddybhow can a->b = c; (c being a #define'ed constant) to write to the c address?
  2. kwolfHm, Pedigree uses a completely different set of magic numbers. I couldn't interpret those. ;-)
  3. pcmattmankwolf: Elite Cafe ;)
  4. pcmattmaneddyb: If b overlaps c
  5. eddybs/to write/write/
  6. kwolfpcmattman: Sure, but isn't it tied to a certain error condition anyway? ;-)
  7. pcmattmanYou can try and blame everything else if you want, but those problems haven't come up anywhere else but the code you're writing now
  8. pcmattmankwolf: We sometimes use more than one magic number ;)
  9. pcmattmankwolf: I mean, one magic number more than once
  10. kwolfNot sure if it's clever to do this...
  11. eddybpcmattman: try and run ./Configure
  12. eddybwith that don't ask flag
  13. pcmattmankwolf: Usually the context of the error is the key, the logs are pretty verbose
  14. eddybat some point you'll get a deterministic 0x1337cafe
  15. pcmattmaneddyb: what on earth are you doing running ./Configure again?
  16. pcmattmanYeah, a deterministic 0x1337cafe right before it runs out of RAM
  17. eddybis the only deterministic way of getting that bug...
  18. eddybi get it sometimes from ext2
  19. pcmattmanUnless you can mathematically prove to me that your code is not at fault, your code is at fault to me
  20. kwolfpcmattman: It really helps to know "0xd00fc0de => use after free" and such things. So you can't defer that from your numbers?
  21. eddybi mean, while allocating from ext2
  22. pcmattmankwolf: Well, the allocator magic numbers are very well-defined
  23. pcmattmanThe important ones, anyway
  24. eddyb(NN) KERNELELF: Module network-stack finished executing
  25. eddyb(EE) Page Fault Exception at 0x1337cb02, error code 0x00000002, EIP 0xff42697b
  26. pcmattmanu comes after n
  27. pcmattmanuhci is being executed, most likely
  28. eddybnope
  29. eddybis usb-hcd
  30. eddybso comes after usb
  31. eddybwhich wasn't loaded yet
  32. pcmattmanThen your dependencies are completely wrong
  33. pcmattmanSeriously, *you're doing something wrong*
  34. eddyband it's in finaliseModule
  35. pcmattmanIt's not Pedigree!
  36. eddyband you want me to belive that?
  37. pcmattmanI'm not really giving you a choice
  38. eddybare you afraid of a 0xb00b1e55 clone?
  39. XanClicHm, thx
  40. pcmattmanI'm afraid of you wasting more time tracking down a bug when your code is at fault
  41. XanClicNice constant for niwohlOS
  42. eddyb-Policy_Binding(not demangled)->String::free->delete[]->SlamAllocator::free
  43. kwolfXanClic: Can't you create your own instead of stealing them all? ;-)
  44. eddybSlamAllocator::free is where *all* 1337cafe bugs end up
  45. XanClickwolf: No ;-)
  46. XanClicCreating constants would mean we still have some Niwoh
  47. eddyb0xb00bbeef :P
  48. pcmattmaneddyb: all the 1337cafe bugs end up in SlamAllocator::free because that's where it asserts that the magic value in the header is correct
  49. eddyband somehow it writes to the magic value... or an offset in it..
  50. pcmattmanAnd "it" in this case is your code
  51. pcmattmanNow fix it
  52. eddybif you don't want, i'll close the window and boot again, it won't appear
  53. eddyb=))
  54. pcmattmanYour code is corrupting the heap, which is the exact cause of this error
  55. eddybi never touched SlamAllocator::free and there it writes....
  56. pcmattmanI know it's your code, because my working copy boots fine on every piece of real hardware I own as well as in VPC, VMWare, VirtualBox and QEMU
  57. eddybit's not deterministic....
  58. pcmattmanHeap corruption bugs aren't deterministic
  59. pcmattmanBut I haven't run into anything like this in my working copy since before Foster came out
  60. eddybat bootup from ext2 it happened only 2 times
  61. eddybbut it's deterministic in ./Configure...
  62. eddyband btw, is right after the "Perl devs(i think) greet you. Do you greet back?"
  63. eddybso i don't think it runs out of memory *that* soon
  64. pcmattmanAgain, I haven't received that, just an "out of memory" error in 4 GB of RAM
  65. kwolfpcmattman: Now I know why you were thinking about user space drivers ;-)
  66. pcmattmankwolf: :)
  67. pcmattmaneddyb: also, I honestly do not recommend using ext2 at all
  68. pcmattmanAnyway, I'm off. Night all
  69. kwolfWhich reminds me that we might need to revise cdi/fs
  70. kwolfGood night
  71. eddybargh
  72. eddybthere he went
  73. eddybbecause i found out that no code touched by me(that isn't in the master) is runned until that error
  74. eddybalso, have you considered double free?
  75. kwolfWho is "you"?
  76. eddybmatt
  77. eddybbad timing XD
  78. kwolfWell, I'm not sure if you're interested in my opinion, but...
  79. kwolfYou are the one who gets the error. You should be the one to find the cause and fix it
  80. eddybhe wouldn't let me to...
  81. eddybhe doesn't belive the bug exists
  82. kwolfIf you can point to a bad line of code I'm sure he'll believe you. But you can't until now.
  83. eddybwell
  84. eddybi added a bunch of lines to test a condition...
  85. eddybnext time, if i'm right, i can come with a theory...
  86. eddybwhich mostly means double free
  87. eddyband if that object was freed then something else allocated, it's possible that the magic value overwrites the cache pointer, and then i get a call in the object referenced by the pointer...
  88. eddybbbs, food
  89. eddybi'd like more Bughunter :P
  90. XanClicSorry, but there's no /ns bug command ;-)
  91. eddyb/ns?
  92. XanClicabbreviation for /msg nickserv
  93. eddybaah
  94. eddybnice
  95. eddybirc or just konversation?
  96. XanClicI think it's konversation
  97. kwolfIf there was this command, I would have NickServ bug eddyb all the time ;-)
  98. XanClicBut www.euirc.net uses this abbreviation, too
  99. eddybhmm
  100. eddyband that Ghosthunter is your secondary nickname?
  101. XanClicYes
  102. eddybi have a /ghost
  103. eddybbut the problem is...
  104. eddybit sends an empty message
  105. eddybbefore it does it's job
  106. eddybactually, containing one space
  107. eddyb
  108. eddyblike^^^
  109. eddybXanClic: if i give you a pastebined part of my code(the one that setups the TD and the frame pointer), can you tell me what i'm missing?
  110. XanClicHm, dunno
  111. eddybif i set the frame pointer to 0, i get 0x30(HC process error+HC halted)...
  112. eddybbut if i change the PID field, i don't get any error
  113. XanClicGive me a week, maybe I'll be writing a new driver by then ;-)
  114. eddyba week?
  115. eddybfor what?
  116. XanClicpaloxena3 isn't far enough yet
  117. XanClicFirst of all, I need some driver infrastructure
  118. XanClicAnd I mean good infrastructure
  119. XanClicCDI doesn't have buses until yet (afaik?)
  120. eddybhttp://codepad.org/GJ7mD7sE
  121. eddyband http://codepad.org/llS8OhaY
  122. eddybit's small
  123. eddyband i don't see why it doesn't get to it
  124. eddybkeep it mind that it works in qemu
  125. XanClicHm, you set status to 0x80
  126. XanClicIn my TD structure that bit is reserved
  127. eddyblol?
  128. eddybit doesn't work in qemu
  129. eddyb(NN) USB: Device: QEMU USB Keyboard, class=0x3, subclass=0x1, proto=0x1
  130. eddybaah
  131. eddybtakes a little while
  132. eddyband i always get USBSTS 1
  133. XanClicI set status to 0x01
  134. eddybTD status?
  135. XanClicYes
  136. eddybfunny
  137. XanClicHm, wait a moment...
  138. XanClicLittle Endian is confusing...
  139. XanClicmom
  140. eddybXD
  141. eddybit's not like in datasheet
  142. XanClicOK, I set it to 0x80, too ;-)
  143. eddybeach dword had to be written from low to up(i mean the bit thingies in struct)
  144. eddyblike reading from right to left :P
  145. eddybeven so, it *does* work in qemu
  146. XanClicI have just the same thing as you
  147. XanClicThough I don't reset the frame counter
  148. eddybi do async by resetting frame counter, starting HC, waiting for irq, stopping it :P
  149. eddyblol, why's always one port "connected"?
  150. eddybi got one device and two ports with connected bit set...
  151. eddybif i have no device, i still have one with connected...
  152. eddybanyway...
  153. eddybwhat am i missing?
  154. kwolfXanClic: We can get cdi_provide_device and IDC in this weekend if you're finally interested in it ;-)
  155. XanClicHm, let's see ;-)
  156. quokmoin
  157. XanClicMoin, quok
  158. kwolfMoin quok
  159. kwolfOr if anyone is interested in it... I think I'll still be busy with the mem.h stuff
  160. eddybwtf is wrong with my code?
  161. kwolfThe semantics, I guess
  162. eddyb??
  163. kwolfWell, if it was syntax, the compiler would tell you. And I think you wouldn't care about style. So it must be semantics that is wrong.
  164. eddybi meant "what i forgot to do?!"
  165. eddybcan someone point me to the differences between newlib and glibc and how could glibc be "emulated" by newlib?
  166. eddybXanClic: any idea about my code or should i give up
  167. XanClicGive up *g*
  168. XanClicnewlib is crap
  169. XanClicThat's it ;-)
  170. eddybnewlib?
  171. XanClicYes
  172. eddybwe *use* newlib in the posix subsystem
  173. XanClicDoesn't change my opinion about it ;-)
  174. eddybwhat's teh difference?
  175. eddyb*s
  176. XanClicNewlib contains many stubs before you replace them
  177. XanClice.g., the directory functions
  178. XanClicIt doesn't tell you they don't work if you don't replace them explicitly
  179. eddybi know(think) avr-libc is based on newlib...
  180. XanClicI think that's not very good ;-)
  181. eddybwtf?
  182. XanClicFurthermore, the printf was broken
  183. XanClicAt least in my version of it
  184. XanClicafk
  185. eddybre:my code...
  186. eddybbefore i'm giving up, i'll try on that old machine
  187. eddybmaybe the changes will do something
  188. eddybif you want to contribute to the progress of humanity, i can give you a 1.8MB ISO
  189. eddyb:P
  190. quok humanity be damned
  191. quok;)
  192. eddybhttp://en.wikipedia.org/wiki/Embedded_Linux
  193. eddyblol, does your plasma got latest kernel?
  194. XanClicIf I got the latest kernel?
  195. eddybno, i was talking of that wiki page
  196. eddybLCD and plasma TVs from LG, Panasonic, Samsung, Sony [2]
  197. quokGerman phrase of the day: " Ich möchte Sie zum Abendessen einladen.
  198. quok"
  199. quokcouldn't forget my closing " :) Wouldn't want anyone to return EPARSE ;)
  200. XanClicHm, what's the actual sense of those phrases?