#tyndur · 2010-02-04

409 Nachrichten · Seite 1 von 3 · Original-CSV herunterladen

  1. taljethgn8
  2. pcmattmanHmm, taljeth isn't around yet
  3. FreakyPenguinMoin
  4. kwolfGood morning everyone
  5. pcmattmanHello kwolf, FreakyPenguin
  6. pcmattmankwolf: is there any way to access the PCI configuration space in CDI?
  7. kwolfYes, there are optional functions for it, guarded by some define. I don't think anyone implements them yet. ;-)
  8. kwolfOr Paloxena probably does
  9. kwolfcdi/pci.h, the part after #define CDI_PCI_DIRECT_ACCESS
  10. pcmattmanOkay
  11. pcmattmanJust wondering, because I'm looking at writing my AHCI/SATA driver soon and I'll need access to that space
  12. pcmattman(CDI, naturally)
  13. FreakyPenguinHm how bad is SATA? ;-)
  14. pcmattmanDepends on how you define "bad" ;)
  15. kwolfpcmattman: So you decided to do it in CDI right from the beginning now?
  16. pcmattmankwolf: sure did
  17. pcmattmanIt'll be a bit harder as I'm actually learning a lot about how to write CDI drivers as I go
  18. pcmattmanI know the interface, just not how to write drivers for it ;)
  19. XanClicYes, Paloxena does
  20. kwolfI think nobody has written a new driver for the current interface yet ;-)
  21. pcmattmanThe hardest thing conceptually at the moment that I'm finding is that I'm unsure how to handle the AHCI -> disk link
  22. pcmattmanIn Pedigree I would be able to have drivers use functions in other drivers and therefore be able to link them that way
  23. pcmattmanCDI is different, should be fun
  24. kwolfHm, is this the same kind of problem as USB has?
  25. pcmattmanWell, you have AHCI as the controller interface
  26. pcmattmanIt tells you about the SATA devices, but you still need a driver to talk to the SATA devices *through* AHCI
  27. pcmattmanSo AHCI provides an MMIO area that is used to access the controller and also the disks themselves
  28. kwolfAnd you have multiple different SATA drivers and/or AHCI drivers?
  29. pcmattmanWell, AHCI is standardised by Intel
  30. pcmattmanI know for a fact though that there's six or seven sata-xyz.c files in Linux's tree
  31. pcmattmanI still need to read the specification further, but it seems so far that the SATA drives end up acting like conventional SCSI drivers, accessed through the AHCI interface
  32. pcmattmanerr, SCSI *drives*, not drivers
  33. kwolfLooks like we need to get IDC up and running first, then
  34. pcmattmanFor example, on Windows XP you merely install an AHCI driver and it can use SATA disks
  35. pcmattman(in fact I did that yesterday)
  36. kwolfHm, so AHCI is actually enough if you already have a SCSI driver?
  37. pcmattmanIt could well be
  38. kwolfWhich you probably have to some extent because of ATAPI
  39. pcmattmanThe AHCI specification is 100 pages, and the SATA 3.0 specification is almost 600
  40. pcmattmanI own but am ignoring pre-3.0 specifications for now
  41. pcmattmanWith a quick glance of the specification I see DEVICE RESET, PACKET, READ/WRITE DMA QUEUED
  42. pcmattmanIDENTIFY PACKET DEVICE... "implemented by _native_ Serial ATA devices"
  43. pcmattmanSo yeah, looks like SCSI to me
  44. pcmattmanHmm, maybe rapid prototyping using Pedigree's native interface then porting will be more efficient
  45. kwolfHm, we should get a scsi-disk driver some time. In tyndur, that code is in the CDI lib currentlyv to avoid the need for IDC.
  46. kwolfpcmattman: If it is, we need to improve CDI. It shouldn't be tedious to write drivers for it.
  47. pcmattmanIt's not so much that it's tediou
  48. pcmattmans
  49. pcmattmanIt's more that it's much easier for me to add hacks to Pedigree to get my way than it is to do the same to CDI for testing ;)
  50. kwolfIn the end you need to get the CDI part right anyway
  51. kwolfUnless you want to keep using your native version
  52. pcmattmanYup
  53. pcmattmanThe idea is that it's easier to get native right first because I don't need to wait for IDC (or hack around it) or anything like that
  54. kwolfAnd your motivation might be higher to do it if it doesn't work otherwise ;-)
  55. pcmattmanI can get it written first and then find out what's missing in CDI
  56. kwolfWell, yes, you could do it in CDI plus some Pedigree specific calls. I think this is an entirely reasonable way to start out.
  57. kwolfHm, does AHCI even access the SCSI disk or is it only the other way round?
  58. pcmattmanThe AHCI controller provides the port ranges through which you access the SCSI disk
  59. pcmattman*SATA disk
  60. kwolfSo AHCI never needs to access the SATA disk driver, right?
  61. pcmattmanTechnically no
  62. kwolfYou could AHCI in CDI then and a native SATA disk driver
  63. pcmattmanHm, defeats the purpose in my opinion
  64. kwolfIf I understand right, AHCI should be the greatest part of the whole thing anyway
  65. pcmattmanIf there was an scsi driver, the AHCI driver could tell the scsi driver (via IDC) information about each detected drive
  66. kwolfpcmattman: More than doing both native?
  67. pcmattmanWell, I'm thinking AHCI might be the only driver necessary if this is done correctly
  68. kwolfSure, that SCSI driver and IDC is the next step
  69. pcmattmanAnd of course, if I have interpreted the spec properly so far... otherwise none of this is right :)
  70. kwolfI'm just saying that there's no point in doing a native AHCI driver
  71. pcmattmanI think I'm beginning to agree
  72. pcmattmanAs long as someone else does the SCSI driver ;)
  73. FreakyPenguin*g*
  74. kwolfFreakyPenguin volunteers? ;-)
  75. pcmattmanSounds like it!
  76. pcmattmanBasically, the AHCI driver would set up the interface and pick up all the drives on the bus (up to 30, iirc)
  77. Homixi thought, ahci would use ata commands?
  78. pcmattmanThen it tells the SCSI driver "Here's a new disk device at this port, go talk to it"
  79. kwolfTo be precise, AHCI would call some cdi_provide_device() for SCSI interface and CDI would go tell the SCSI driver about it
  80. pcmattmanRight
  81. kwolfI'm feeling like starting to hack on it ;-)
  82. kwolfUnfortunately it will have to wait until the evening at least
  83. pcmattmanHeh
  84. pcmattmanWell, if an OS with a CDI implementation can run and use SATA disk drives, that's an instant selling point of CDI
  85. pcmattmanUDI can't do that yet ;)
  86. kwolfI still haven't seen the hobby OS with UDI support yet ;-)
  87. pcmattmanHeh, true
  88. kwolfCombuster claimed to have it, but when I kept asking for the code and he posted the URL of his repo eventually, I saw that it's all empty stubs
  89. pcmattmanRight
  90. pcmattmanAnd that aside, everyone likes SATA support ;)
  91. pcmattmanHomix: by the looks of it it uses a small subset of ATA commands
  92. pcmattmanIDENTIFY PACKET DEVICE, PACKET, etc...
  93. pcmattmanIt's certainly not going to happen tonight though, I'm still reading the specification and trying to make sense of what I need to do
  94. kwolfIt surely doesn't need to be done tomorrow. I'm happy as long as I see things moving.
  95. pcmattmanGive me a couple of weeks
  96. pcmattmanIn the meantime I'll probably be looking at the automated driver loading patches
  97. kwolfYup, they would be nice to have, too
  98. pcmattmanAs for now, sleep time. Too much spec reading for one night ;)
  99. pcmattmanNight all
  100. FreakyPenguinn8 pc[away]
  101. eddyb'night
  102. eddybheh, sometimes spec reading is driving me crazy :)
  103. Homixahci uses ata task files, so i think all ata commands schould be available ;)
  104. quokmoin
  105. eddybyou mean "morn" ?
  106. eddyb:P
  107. quokno, I mean moin
  108. eddybmoin as in ... ?
  109. FreakyPenguinMoin quok ;-)
  110. eddybit's a short form of morgen?
  111. FreakyPenguinYes indeed ;-)
  112. eddybit didn't sound like it :P...
  113. eddybanyway, moin quok :P
  114. eddybhmm
  115. eddybi like how it asks for keyboard...
  116. eddybi'm looking at tyndur :)
  117. eddybpc[away]: for releases it could be useful, like something in login :)
  118. eddybi choose EN at grub, why is the help in german :(
  119. FreakyPenguinNoone volonteered yet to translate it ;-)
  120. eddybany kind of i18n layer?
  121. kwolfMoin quok
  122. FreakyPenguinquok: Would be a nice exercise for you *g*
  123. eddybfor me, too :P
  124. FreakyPenguinFeel free ;-)
  125. eddybif you gimme some files, i can translate, but just if i don't have to learn the source tree of tyndur :)
  126. eddybit took me some while with pedigree, but now i can find a file with no problems :)
  127. kwolfActually I think help doesn't suport any kind of i18n yet
  128. eddybyou would really get more people to test tyndur if it would have some sort of localisation
  129. eddybkwolf: it's a xml file, right?
  130. stultusuhm
  131. stultushelp is somewhat similar to docbook
  132. stultusat least i've tried to use that format. although the parsing probably really couldn't be considered "conforming"
  133. eddybwhat about <lang:de>stuff</lang><lang:en>english stuff</lang> ?
  134. eddyb:)
  135. kwolfI would use different files for each language
  136. eddybor that
  137. kwolfSo you can choose which languages to install
  138. eddybfile:/system/lpt-doc/en/tyndur.xml :)
  139. stultuswould be the best, actually. Hacking language support into the program would be tedious.
  140. kwolfIt could simply check $PAGENAME_$LANG.xml first and failing that it could fall back to the current $PAGENAME.xml
  141. eddybor that :)
  142. kwolfI like eddyb's suggestion more for its cleanness but it's a bit more effort to implement, iirc
  143. eddybthat reminds of my dream of having pedigree in romanian :P
  144. eddybi can get maghyar translations, too
  145. eddybactually, i think matt tried to get gettext working on pedigree
  146. kwolfeddyb: But who would use Hungarian?
  147. eddybmy friend who would translate it, LOL :P
  148. kwolfgettext is easy to port. I think on tyndur it just worked without hacking anything.
  149. eddybi think on pedigree it was something like not compiled it into the ports....
  150. eddybwhere can i browse the source?
  151. kwolfFor the ports?
  152. kwolfAnyway, start at http://git.tyndur.org/
  153. eddybs/ports/ported apps/
  154. kwolfhttp://git.tyndur.org/?p=lbuilds.git;a=tree;f=lbuilds;h=ee6770ff138e648a51440fd18fd606b4e8742dee;hb=HEAD
  155. eddybah, and, kwolf: what do you think about german keymaps for pedigree :P ?
  156. kwolfOh, right, I think I promised at some point to do them
  157. eddybheh
  158. eddybi can give you instructions and it's really simple...
  159. kwolfPing me later this evening and I'll do it
  160. eddybwe already have Us and Uk sets, so it'd be easy to do it, just looking at current ones :)
  161. eddybokey
  162. kwolfYup, and it will improve usability for me a lot ;-)
  163. eddybkwolf: those links almost forkbombed my system
  164. kwolfwtf?
  165. kwolfThis is a normal gitweb frontend
  166. eddybi left it loading and after 5 min, the system was slowed down really bad
  167. FreakyPenguin:-/
  168. eddybafter closing the tab, it's almost okey
  169. kwolfSomething is wrong with your system...
  170. kwolfI was already thinking that when you said that building an image takes three minutes, but even more so now
  171. eddybnow i see a chome zombie was eating HALF A GYGABYTE
  172. eddybmuch better now after i killed it
  173. eddybwell, that page wasn't even loading, so i was thinking it's because of it
  174. eddyb130MB when i stopped it
  175. eddybit was tyndur git
  176. eddyband i have sandbox enabled....
  177. eddybkwolf: does it load for you?
  178. kwolfYes, without any problem
  179. eddybin what browser?
  180. kwolfEh, no, it doesn't work any more
  181. kwolfFreakyPenguin: "403 Forbidden - No projects found" - what are you doing?
  182. eddybtold'ya :P
  183. eddybit's insane for a chrome process to use more than 32MB
  184. eddybwell, except flash and youtube
  185. eddyb42 and 34
  186. FreakyPenguinkwolf: Now it should work again
  187. kwolfWhat was it?
  188. FreakyPenguinI installed the debian gitweb-Package to have a look at the diff, and it installed a config-File ;-)
  189. FreakyPenguinIn our gitweb I just changed the default settings in the source ;-)
  190. eddybit's still doing something like while(1);
  191. eddybanyway, its process eats all the memory fast
  192. FreakyPenguineddyb: The diff to the original debian script is almost empty
  193. kwolfeddyb: Have you tried with a different browser?
  194. eddybi still think chrome can't be the problem :P
  195. eddybthe simple link seems to load
  196. kwolfHm, need to restart the X server...
  197. FreakyPenguinHas qemu eaten your cursor? ;-)
  198. eddybcan qemu even do that?
  199. eddybhi, XanClic
  200. XanClicUh, hi, eddyb *g*