#tyndur · 2010-01-18

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

  1. kwolfMoin
  2. pcmattmanHi kwolf
  3. kwolfFinally some activity on the mailing list again :-)
  4. pcmattmanYes, good to see
  5. pcmattmanI'll be back soon, dinner
  6. FreakyPenguinMoin
  7. kwolfGuten Appetit, pcmattman (don't think that exists in English...)
  8. kwolfMoin FreakyPenguin
  9. kwolfFreakyPenguin: What about you? I haven't seen a mail from you on cdi-devel for a while. You're just going to take what we decide?
  10. FreakyPenguinAs long as it looks reasonable ;-)
  11. kwolfYou should know my code well enough... ;-)
  12. FreakyPenguin*g*
  13. FreakyPenguinI'm just waiting for the discussion about cdi/fs ;-)
  14. kwolfDo we need to discuss the rewrite or should we just do it? ;-)
  15. FreakyPenguinMaybe we should discuss _how_ the rewrite should look like ;-)
  16. kwolfProbably something that is easy to implement...
  17. pcmattmanBack
  18. FreakyPenguinkwolf: Well, what is the problem with the current implementation?
  19. FreakyPenguinThe problems I see: No Hardlinks, permissions aren't usable and maybe we should do something about the cache
  20. FreakyPenguinwb pcmattman
  21. kwolfHm, I haven't implement it yet. I guess it's too much abstraction. And wasn't the cache a problem?
  22. kwolfAnd then we need to support translation tables
  23. FreakyPenguinExactly ;-)
  24. kwolfpcmattman: In my mail I'm actually coming from a monolithic kernel, too
  25. pcmattmanWell, yes - in your mail
  26. pcmattmanI was thinking more from the perspective of CDI in general, where we need to cater for micro and monolithic kernel architectures
  27. kwolfAbsolutely
  28. pcmattmanThat didn't really come across right in the mail though :)
  29. kwolfBe sure that I would never neglect the needs of a microkernel system ;-)
  30. kwolfI'm not sure why interrupting a driver function is easier in a monolithic kernel. I mean, in a microkernel the problem I described can never happen.
  31. pcmattmanWell, each driver function can be called by an individual thread: there's no real problem with needing to send an IPC request to an application or anything
  32. pcmattmanApplication, driver, whatever your terminology is for your microkernel drivers
  33. kwolfBut you get the concurrency issues in a monolithic kernel, whereas in the microkernel model things are usually queued anyway
  34. pcmattmanThis is true
  35. pcmattmanUnfortunately I'm quick to ignore the concurrency issues and go with the solution which is easiest to write. It's certainly a flaw :)
  36. kwolfIgnoring concurreny issues and later wondering what the cause for the crash might be is always easier ;-)
  37. pcmattmanOr adding "/// \todo locking" ;)
  38. kwolfFreakyPenguin: We should add these to kernel2 ;-)
  39. kwolfpcmattman: Exactly this is why currently I've completely disabled IRQs while drivers are running. Or basically while any kernel code is running, with the exception of the idle task.
  40. kwolfIt nicely solves any concurrency issues (in exchange for some other problems, of course)
  41. pcmattmanRight
  42. pcmattmanI still think working IRQs are an important requirement for a proper CDI implementation. Concurrency issues are something we need to solve at some point, after all
  43. kwolfIn the end I'm going to move driver code into their own kernel threads, though. IRQs wouldn't be a problem then any more.
  44. kwolfBut I guess many OSes that start using CDI are in an early stage
  45. kwolfOr at least that's what I hope
  46. pcmattmanRight, so providing concurrency constructs may be as simple as a volatile char in their initial implementation
  47. pcmattmanIRQs are usually one of the first thing people get working, after all
  48. pcmattmanConcurrency however is usually one of the last things people think about
  49. kwolfRight
  50. kwolfI think if we go forward with the suggestion from the mail thread, we're fine
  51. pcmattmanIf you define the interface to only allow one driver function per device to be running (and make that an OS-controlled thing), you offload a bit of effort from CDI driver writers
  52. kwolfWe just need to find a good definition for cdi_wait_irq
  53. FreakyPenguin/// Does the right thing. ;-)
  54. kwolfpcmattman: I don't think your interruptible flag does really gain us anything. When would a driver set it and when not?
  55. pcmattmanWell, I don't think a driver would set it at all
  56. pcmattmanIt would be for interface use only, as a lock to only allow one driver function per device to be called
  57. pcmattmanThis "flag" could easily be a mutex or a similar concurrency construct
  58. kwolfIf drivers don't use it, there's no reason to include it in the interface
  59. pcmattmanThat's true
  60. kwolfThe library can do anything it wants anyway
  61. pcmattmanI think though it would be beneficial to use something like my cdi_wait_irq example for a future implementor's reference or something, as a psuedocode example
  62. pcmattmanSure, we are looking at drivers and the interface itself, but we also need to consider programmers who want to implement the CDI interface into their OS and might be a bit confused about what some terminology means
  63. kwolfWe might want to provide some standard functions, indeed. For example I'm going to remove the partition table code from ata (it really doesn't belong there), but I'd still like to share it.
  64. pcmattmanPerhaps it could go into a partition driver?
  65. kwolfAnother driver? Hm...
  66. pcmattmanIn Pedigree our partition driver is separate
  67. pcmattmanIt means we can talk to the partitioner directly with things like disk images without going through ATA or something
  68. kwolfata is something that should be easy to implement. I wouldn't feel comfortable with requiring IDC for it
  69. pcmattmanI'm not sure how you would make it work for tyndur/cdi
  70. pcmattmanI'm merely pointing out how my project does it, so you understand where I'm coming from in this instance
  71. kwolfRight
  72. kwolfIf we put the things into a library instead of a driver you wouldn't lose anything, would you?
  73. pcmattmanI doubt it
  74. kwolfAnd things like the cdi_list_t implementation are something else to share for the library that certainly can't be a driver
  75. pcmattmanYup
  76. kwolfLooks like there's still a lot to do until end of February
  77. kwolfHm, we're getting checksum offloading for cdi/net?
  78. taljethHey, quok, XanClic, we're waiting for your comments on the list
  79. XanClicHm, there's nothing I didn't decide not to response to ;-)
  80. FreakyPenguinHm das mit dem Englisch üben wir aber nochmal ;-)
  81. mgottschlaghey, nix gegen 3x verneint.
  82. XanClicIrgendwann muss man doch mit boolescher Algebra anfangen ;-)
  83. taljethAber zumindest müsste es respond heißen, oder? ;-)
  84. XanClicAh, ja ;-)
  85. XanClics/response/respond ;-)
  86. taljeth+/ ;-)
  87. XanClic+g?
  88. taljethKommt doch nur einmal vor
  89. XanClicEgal, kann nicht schaden ;-)
  90. quokXanClic: btw, for now assume that the BIOS will work properly and support devices up to 8GB with CHS only
  91. XanClicthx :-)
  92. quokI'm not going to run bootsector tests on a 386 with a 8GB HD in it :)
  93. XanClicI thought you planned to do so?
  94. XanClicOr will you just test a 8GB imag?
  95. XanClic*image
  96. quokNo, because of BIOS limitations, the large image tests will be done on more modern machines that can support those drive sizes
  97. XanClicAh, 386 was meant literally ;-)
  98. quokyes
  99. XanClicMy machines are modern ;-)
  100. quokI'm about to post all of that
  101. quokThen you have nothing to worry about. :)
  102. XanClic*g*
  103. quokwell, SHOULD have nothing to worry about.
  104. quokI'm preparing for a job interview at the end of this week, so I haven't yet finished running all my machines through the tests I have planned
  105. quokbut next week I should be able to post them
  106. quokYou're right why I chose 8GB as the max image size, too :)
  107. XanClicbtw, one thing I have wanted to ask for a long time: One doesn't use "an" in front of letters "starting" with a vowel such as "S" which is spoken "as" (or something like that)?
  108. quokDo you mean like "an old laptop" vs "a old laptop"? the former would be correct.
  109. XanClicI know
  110. XanClicBut if I use "SVO" as an abbreviation, for example
  111. XanClicIs it "an SVO" or "a SVO"?
  112. XanClicBecause when spelling the letter "S" I actually start with speaking a vowel which sounds like a short "e"
  113. Homixi think it is "a svo"
  114. XanClicJust wanted to ask someone with English as his mother tongue ;-)
  115. quokHmm, I have no idea which would be correct. :) the failings of american schools is now quite clearly evident! :)
  116. XanClicHm, I think s/as/being would be better
  117. XanClic*g*
  118. quokhmm s/is/are/
  119. quokbtw, I'm also glad to see that you found you cannot truly trust the BPB's geometry information
  120. quokthat's the one thing I kept forgetting about adding to the osdev wiki's FAT article when I wrote my fat12/16 bootsector
  121. XanClicHm, wiktionary states "The article an is used before vowel sounds, and a before consonant sounds." which sounds as "an SVO"
  122. XanClicquok: Admit it, you wanted to make it harder ;-)
  123. quokI did want to make it harder :)
  124. quokI hope I didn't mis-state that the geometry information can be assumed to be correct. The rest of the FAT information can be assumed to be correct, of course, otherwise you'd have a corrupt filesystem
  125. XanClicBesides of the type field ;-)
  126. quokbut you can't ever really trust the drive geometry, except on certain hardware.
  127. XanClicWhich may be assumed to be correct but with an incorrect one the FS wouldn't be broken
  128. XanClicHa, Wikipedia ftw
  129. XanClic"an SSO"
  130. XanClic:-)
  131. XanClic"an LWR (Note: When an acronym is spelled out, rather than spelling out what it is that is represented by the letters in the acronym, the phonetics of the acronym should be used when reading the text aloud. So the use of "a" or "an" as shown in this example is correct based on the proper application of this rule)"
  132. XanClicSo it would be "an 8GB image", too ;-)
  133. HomixXanClic: link? ;)
  134. XanClichttp://en.wikipedia.org/wiki/English_articles#Indefinite_article :-P
  135. XanClicIt also says "now used before words starting with a vowel sound, regardless of whether the word begins with a vowel letter."
  136. quokbah, why does English has to be so complicated? :)
  137. XanClicEscape is useless *g*
  138. XanClicI don't think it to be complicated, that's the way I always did it ;-)
  139. XanClicRead my forum posts ;-)
  140. quokhmm, well, when I take over the world and every computing device runs QuokOS, I'll also make the official language quokish
  141. XanClic*g*
  142. quokin Quokish, the world 'quok' is used like the word 'smurf' in the world of the Smurfs
  143. quok"have a quok day!" or "I'm going to go quok your wife, you bastard" ;)
  144. XanClicAnd about greetings - "good morning" and "good night" are "official" greetings? Just want to be sure, I don't translate too literal ;-)
  145. XanClic*g*
  146. XanClicHm, wasn't there also a word like that in The Hitchhiker's Guide To The Galaxy?
  147. quok"How was your day? "Quite quok, thank you very much!" (or "Damn, it was the mother quok")
  148. XanClic*g*
  149. XanClicHm, pcmattman is already awake?
  150. XanClicThat means I should go to bed now *g*
  151. quokwhat timezone are you in, btw?
  152. XanClicUTC+1
  153. XanClicIt's 10:35 pm here
  154. quokfor some reason I was thinking UTC+2
  155. XanClicThat's in summer time ;-)
  156. XanClicI have three clocks here
  157. XanClicOne for Kansas City, one for me and one for pcmattman's city, I can never remember that name *g*
  158. XanClicThe timezone was the same of Brisbane, that clock says
  159. XanClicAh, Rockhampton
  160. XanClicGood night and thanks for making that CHS thing clearer :-)