#tyndur · 2010-02-13

276 Nachrichten · Seite 1 von 2 · Original-CSV herunterladen

  1. quokthen I will have achieved world domination! :P
  2. taljethTo be accepted, cdi/os needs a clean CDI implementation, of course
  3. quokhow utterly recursive
  4. quokI'll include a symlink from cdi/os/cdi to ../../, so it'll point to cdi/
  5. taljethquokos still doesn't implement any part of CDI, does it?
  6. taljethHeh, sounds like a good approach ;-)
  7. quokquokos still doesn't have paging :)
  8. taljethOh, you haven't copied it yet? ;-)
  9. quokhey my name isn't TroyMartin. I don't copy everything :)
  10. quokgeesh, just because I ended up copying some of tyndur's pmm stuff :P
  11. taljethCopying paging, too, would just be consistent ;-)
  12. quokwell I did change some of the pmm stuff
  13. quokfor instance, I made it all use uintptr_t instead of void *
  14. quokhmm, I'd diff it and send it in as a patch but right now I cba to test tyndur aside from a compile test
  15. taljethHm, if you have it compiled, running qemu really isn't that hard
  16. taljethUsually I would run images but hesitate to compile things...
  17. quokwell I haven't compiled tyndur with my pmm changes.
  18. quokI'd also have to write an email in german :P
  19. quokI'll just have it say "this patch implements jelly donuts in tyndur. yum!" ;)
  20. taljeth*g*
  21. pcmattmanAh, I go and write a prototype and documentation for the prototype, and all of a sudden everyone talks lots ;)
  22. taljethUsually that doesn't indicate quality ;-)
  23. pcmattman:(
  24. taljethQuality of what is talked, I mean
  25. pcmattmanAhh, right
  26. pcmattmanI was going to say, I'd presume my documentation would be pretty good, as I didn't look at IRC for a whole 10 minutes
  27. taljethIt must have been interesting to write, at least
  28. taljethWhen you're done with it, send it to quok so he can translate it
  29. taljethIt will be a good exercise for him, so he can post on tyndur-devel in future ;-)
  30. quokwait, now I'm translating *to* german as well?
  31. quok:P
  32. taljethLooks like it ;-)
  33. quokwell, at least it's not french
  34. pcmattmanHeh, poor quok
  35. pcmattmanGood luck is all I'll say to whoever translates this
  36. quokhahahaha
  37. pcmattmanHurrah, it all compiles
  38. pcmattmanI'd hope so, seeing as though I merely added a prototype and documentation :)
  39. quokmissing semicolons always mess that up though :)
  40. pcmattmanIt's true
  41. pcmattmanAhhhhh, crap
  42. pcmattmanI shouldn't have rebased master, now I need to implement the mem.h stuff
  43. pcmattmanOh well
  44. pcmattmanYeah, now Pedigree doesn't build :P
  45. taljethYou can use your commit as the submodule head
  46. taljethThe mem.h patches are on top, so this would work
  47. pcmattmanWell, rebasing put my commit at the top
  48. pcmattmanAs it rewinds HEAD, applies the fetched commits, then applies your commits on top
  49. taljethOh, you rebased master onto your development branch?
  50. pcmattmanI needed to implement the mem.h stuff eventually anyway ;)
  51. pcmattmanRight
  52. taljethOuch.
  53. taljethSo your master is out of sync now.
  54. pcmattmanBy one commit, yes
  55. pcmattmanThe one I just sent to the list
  56. pcmattmanIt's alright, it'll resolve itself
  57. taljethHm, looks like a nice translating exercise ;-)
  58. pcmattmanI'm sorry :)
  59. pcmattmanI should make it clear right now that I'm not translating that ;)
  60. taljethhehe
  61. pcmattmanquok can though
  62. pcmattmanHe'll do a good job!
  63. pcmattmanAlso, taljeth: where are the different device-type-specific cdi_bus_data structs defined?
  64. taljethIn pci.h ;-)
  65. taljethIt's the only one so far I think
  66. pcmattmanOkay
  67. pcmattmanI'm trying to figure out how I'll do Pedigree's cdi_provide_device is all, so I need to know what my assets are :P
  68. taljethI assume that every device type will have its own header, so putting these structs in the respective header should work
  69. pcmattmanRight
  70. pcmattmanI'm planning on changing the PCI detection code as well to actually call cdi_provide_device rather than cdi_list_add(the_device)
  71. taljethLuckily, implementing a compliant cdi_provice_device is easy. ;-)
  72. pcmattmanJust to really test out the idea
  73. pcmattmanThe easiest cdi_provide_device is one line: return -2; ;)
  74. taljethRight. ;-)
  75. taljethHm, I'm just wondering why the bus driver needs to know if a driver was found
  76. pcmattmanWell, say it does a whole heap of setup to get the device ready for detection
  77. pcmattmanIt might need to do a whole lot of teardown if the device isn't actually used by the OS
  78. pcmattmanWe could just merge -1 and -2 into just "something bad happened", I just thought "No driver found" warranted a special error
  79. taljethThe OS might just be installing the driver and could use it later on. But at that point we already told the bus that there is no driver.
  80. taljethI think "no driver found" should be success
  81. pcmattmanTechnically cdi_provide_device should block until the device is provided - driver loaded, etc...
  82. pcmattmanAt least, that's what I think
  83. quokhmmmmm not sure I like that, but can't really tell you why right now either
  84. taljethSo if you install a driver, you need to be able to trigger a rescan on all buses at least
  85. pcmattmanquok: don't like blocking or not blocking?
  86. quokI don't like the blocking idea, I think
  87. pcmattmanI just think the controller needs to know if providing the device didn't succeed because there was no driver found, anywhere
  88. pcmattmanI would think it wouldn't attempt to start a driver load, store the device to give to the driver later, and then return "it worked"
  89. pcmattmanIf the driver failed during loading, or the driver didn't like the device, the controller has no way of finding out
  90. taljethEither the device is present or not
  91. taljethIf it's present the bus should provide it
  92. taljethIf it's not present any more, the bus should remove it
  93. taljethAnything that happens in between is the job of the OS
  94. pcmattmanprovide_device may entail a driver load, which in my opinion is a blocking concept
  95. pcmattmanIf the driver is already loaded, then absolutely provide_device shouldn't block
  96. taljethBut that's not what this call is doing but rather the OS's reaction to the call
  97. pcmattmanHm, okay
  98. pcmattmanI see what you're saying
  99. taljethThe bus doesn't care if a driver is loaded or not
  100. pcmattmanRight
  101. pcmattmanSo, cdi_provide_device could return 1: "Driver being loaded", and if the bus needs to know for sure if the device was successfully provided, it can call cdi_provide_device again and wait for success?
  102. pcmattman(The OS would need to check for an in-progress driver load, naturally)
  103. taljethIs there any reason for the bus to need this information?
  104. pcmattmanAt this stage I'm unsure
  105. pcmattmanI'm sure one arcane device exists that would need this kind of treatment though :)
  106. taljethHm, which one?
  107. pcmattmanI have no idea
  108. pcmattmanIt's up to you if we want to add this just to potentially support some random hardware device which we know nothing about yet, or if we ignore it for now and add it later if we need to
  109. taljethquok: What do you think?
  110. quokHmmm... could cdi_provide_device just return 0 or 1, and we can have another function, say cdi_query_device_state (or driver_state) to get the state of the driver? Like "no such driver", "driver loading", etc
  111. pcmattmanThat could work
  112. quokI figure that way cdi_provide_device would just return "device exists" or "device does not exist" and some other function could get a more exact state if needed
  113. quokIMO provide_device is just a boolean type of operation. either it works and the device exists, or it doesn't and the device doesn't exist. there's no in between :)
  114. quokevery now and then I have a good idea ;)
  115. quokjust not very often, haha
  116. taljethHm, I'm not sure what the state would look like, but I think I like this better
  117. taljethI absolutely agree that cdi_provide_device should be boolean
  118. taljethAnd false should only be returned if something went wrong internally, but normally there should be no way to get this
  119. pcmattmanI'm happy with that
  120. pcmattmanWant a revised patch?
  121. taljethIf you think everything has been discussed now, sure
  122. pcmattmanSent patch v2
  123. taljethOk, looks good. Let's just wait for a comment from Max before I commit it
  124. pcmattmanYup, no worries
  125. pcmattmanPast midnight now for you, right?
  126. taljeth1:16 to be precise
  127. pcmattmanAh, GMT+1
  128. pcmattman:)
  129. quokUTC+1 :P
  130. quokwho uses GMT anymore? :P
  131. pcmattmanI only just got used to using GMT, now everyone wants UTC :(
  132. taljethJust say CET ;-)
  133. taljethHm, a bit more than a month to go and it will be CEST again. Then we probably won't meet so often in your morning any more.
  134. pcmattmanUnfortunately true
  135. pcmattmanThe mailing list is handy for that though
  136. quokspeaking of the mailing list
  137. quokI ack'd patch v2 that wasn't v2 :)
  138. pcmattmanThat wasn't v2?
  139. quokthe second version of your patch wasn't labeled as [PATCH V2]
  140. pcmattmanIt totally was
  141. quokOh, gmail fail
  142. quoksorry
  143. pcmattmanHehe, I use gmail as well, and it merged the two into the one thread
  144. quokright
  145. pcmattmanIt's either designed to do that, or that's a bug
  146. quokwell, it follows the thread-id header or whatever it is, which is fine
  147. taljethmutt didn't do it
  148. quokbut it hides the subject by default in replies
  149. pcmattmanThis is what we get for using webmail ;)
  150. pcmattmanSoon I plan on moving to Google Apps for Your Domain "Premium" which supports Exchange sync (ie, Outlook :P)
  151. taljethProabbly gmail puts mails with similar subjects into the same thread or something
  152. XanClicThunderbird doesn't, too
  153. taljethGood morning, XanClic ;-)
  154. XanClicMoin, moin ;-)
  155. taljethYou're early up today ;-)
  156. XanClicYeah, I plan to watch this Vancouver thing
  157. XanClicDon't know how to call it in English
  158. XanClicOlympic Winter Games?
  159. quokyup
  160. quokI saw the video of the guy that died during his practice run on the luge. Ouch.
  161. XanClicI knew they'ld send the video. But I don't appreciate such behaviour...
  162. quokI saw it on the news. Can't really believe they aired it
  163. taljethYup, it was on the news here, too
  164. taljethAnyway, it's late enough and I'm not really interested in the opening ceremony ;-)
  165. taljethgn8
  166. XanClicgn8
  167. quokgn8
  168. pcmattmanHmm
  169. taljethXanClic: You might want to have a look at the cdi_provide_device patch
  170. pcmattmanI went to purchase something on the internet, and instead of pulling out my visa card, I pulled out my "Approved to work with children" card :/
  171. XanClicAh, yes, there was a window too much open, that was Thunderbird ;-)
  172. quokhaha
  173. pcmattmanI probably should think about implementing mem.h stuff into Pedigree
  174. pcmattmanSo much effort though :P
  175. XanClicAh, opening ceremony finally over
  176. XanClicAnd it's 6 am
  177. XanClicHm, "good night" would be wrong now, so good morning ;-)
  178. taljethEin XanClic! :-)
  179. taljethMoin
  180. XanClicMoin, moin
  181. taljethSo have you decided by now? CDI or fighting Nazis? ;-)
  182. XanClicThe police helicopter says I should take the former one ;-)
  183. taljethhehe
  184. eddybmoin :)
  185. XanClicMoin, eddyb
  186. taljethMoin
  187. eddybhttp://forum.osdev.org/viewtopic.php?f=1&t=21568 << is that guy trying to look important..?
  188. eddybzeroing all unused fields doesn't work
  189. eddyb:(((
  190. eddybi simply can't see anything wrong...
  191. eddybXD
  192. quokmoin
  193. XanClicMoin, quok
  194. XanClicBut you know, I went to bed at 6 am ;-)
  195. taljethMoin quok
  196. taljethXanClic: And you still haven't worked on CDI nor fought Nazis!
  197. XanClicDAMN
  198. XanClicHm, let's see
  199. XanClicDamn, who was tranlated the comment to "int cdi_provide_device(struct cdi_bus_data* device);"?
  200. XanClicThat was my job