=== fabbione [i=fabbione@gordian.fabbione.net] has joined #ubuntu-kernel === doko_ [n=doko@dslb-084-059-068-086.pools.arcor-ip.net] has joined #ubuntu-kernel [03:19] fabbione: ping [07:19] dilinger: pong? [07:21] let me guess [07:21] sunfire is down === chmj [n=chmj@wbs-146-157-70.telkomadsl.co.za] has joined #ubuntu-kernel === zul [n=chuck@CPE0006258ec6c2-CM000a73655d0e.cpe.net.cable.rogers.com] has joined #ubuntu-kernel [02:59] morning [03:00] hey zul [03:12] any know problems with hci_usb on powerpc ? [03:31] dont know...i dont use powerpc :) [03:31] its yucky ;) [03:36] seem to be broken since -9.21 [03:36] or something [03:38] nope [03:38] no changes to USB in ages [04:04] ugh [04:05] yea, looks like the sunfire crashed [04:07] dilinger: yes.. i saw that :) [04:07] can you restart it before you vanish? [04:07] yep [04:07] i just need to be sure that cron doesn't run at 20/50 before i can login again [04:08] otehrwise it will restart the buildd in a mess === mkrufky [n=mk@68.160.103.77] has joined #ubuntu-kernel === lamont [n=lamont@15.238.6.221] has joined #ubuntu-kernel === dilinger twiddles his thumbs and wonders whether it's coming back up or not [04:12] no way to look at console? [04:13] ah, it's pingable [04:13] there is, but it means plugging stuff in. i wanted to avoid it if possible :) [04:13] alright, it's back up [04:17] same ip? [04:18] yea [04:18] .130 [04:18] yeah [04:18] i am in [04:18] right in time to stop cron :) [04:18] thanks dude [04:23] hppa 89.53% 5582 of 6235 [04:23] sparc 89.01% 5589 of 6279 [04:23] sparc's almost caught up. dammit [04:23] fabbione: of course, it helps that you can build kde [04:23] lamont: heheh === lamont again resists the temptation to divert gcc-4.0 to gcc-3.4 for kde packages. [04:24] on hppa, of course. === dilinger looks at the rain outside, sighs, and heads off to work [04:26] dilinger: later [04:26] lamont: well it's not that good as it looks like.. sparc install CD doesn't work thanks to apt-get being segfaulorama on file:// url [04:27] and other annoying problems [04:27] X is rarely autoconfigured [04:28] hppa X has not autoconfigured yet, in my experience [04:28] and network doesn't seem to do hotplug, etc. === clintcan [n=root@203.84.187.254] has joined #ubuntu-kernel === clintcan [n=root@203.84.187.254] has left #ubuntu-kernel [] [04:29] lamont: yeah we have a lot of glitches.. we will get them right for dapper.. specially now that David Miller is my side :P [04:30] heh [04:48] x autoconf is nice. === joh [n=joh@cD908888B.sdsl.catch.no] has joined #ubuntu-kernel [05:23] Hmm, what does a "kernel access of bad area, sig: 11 [#1] " OOPS indicate? I get this with the madwifi modules in linux-restricted-modules-2.6.12-9-powerpc 2.6.12.4-8 and my wifi card :\ [05:23] A segfault from the kernel module? [05:32] joh: yes it's an OOPS [05:35] fabbione: how can I debug this further? you see, a friend of mine has tried my card on the same type of computer with the same version of linux-restricted-modules, where he doesn't experience any problems like that. [05:36] joh: there can several different reasons why it dies and there is no way to debug it without the code [05:37] fabbione: well, there can't be any problems with the madwifi modules, as the same ones work on a different computer... [05:37] fabbione: which other reasons could it be? [05:38] fabbione: I just the problem is not in the hardware :\ [05:38] joh: that's whay you assume :) it can be everything.. also.. the hw might look the same and it's not.. like a different revision in a component or stuff like that [05:38] that might trigger the module to do something different and therefor crash on one and work on the other [05:38] again.. no code.. no way to know [05:39] fabbione: well, that would be strange as we bought the two computers at the same time and they came from the same shipment from apple :P [05:39] fabbione: what's strange is that it worked perfectly before [05:40] fabbione: I don't remember which version of linux-restricted-modules it worked on though... [05:40] joh: same shippement means nothing.. you have no prove they are the same [05:40] fabbione: true [05:40] there is nothing you can do [05:40] there is no code [05:40] unless I start debugging madwifi [05:40] if it doesn't work you can only buy a wifi cards for which there is support in the real kernel [05:41] joh: you don't have the code for madwifi :) the binary blob can/might/is the culprit [05:41] if the problem is there you are doomed [05:41] fabbione: oh :\ right [05:42] and the problem might be everywhere really.. [05:42] fabbione: the problem might not be there though... I look at the stack trace and trace the error in the part of the code in madwifi which is opensouce... [05:42] you load the binary blob that fucks up a kernel pointer [05:42] on one machine is it is not used (no crash) [05:42] on the other it is (boom) [05:43] define 'everywhere'... [05:43] perhaps you see something else oopsing [05:43] once you load a binary blob of anykind [05:43] the kernel OOPS can't be trusted anymore [05:43] Yeah [05:44] so you might see an OOPs in madwifi [05:44] But there were a stack trace with debugging symbols in the kernel logs. Maybe it's possible to debug it from there? [05:44] when perhaps is foobar-usb that's crashing [05:44] (the OOPS occured when doing iwlist scan) [05:44] joh: that OOPS information are not reliable [05:44] that's what i am telling you [05:45] fabbione: it's not reliable because it might come from the binary blob? [05:45] it is not reliable because you have no idea what the binary blob is doing [05:45] it might as well mangle kernel pointers [05:45] and since you don't know that [05:45] there is a chain reaction of events of which you have no control [05:46] but isn't madwifi divided into two parts, one proprietary binary blob and one opensource? [05:46] yes and? [05:47] if the error occurs in the opensource part before it actually communicates with the binary blob... [05:47] you are not listening to what i say [05:47] maybe I've misunderstood..? [05:47] once you just load a module [05:47] that contains a binary blob [05:48] you have lost [05:48] the OOPS become unreliable [05:48] just loading it, make the kernel unhappy [05:48] ok, I see [05:48] but how do you explain the fact that it worked before? change in the module? [05:48] when you load a module there are already a set of functions called [05:49] like initmodules and similae [05:49] similar [05:49] yeah [05:49] perhaps change in the module.. could be due to a change in the binary blob or in the general code.. [05:49] from my point of view it's the same [05:49] ok [05:50] at the end you still get one kernel module for which you don't have full source [05:50] they might as well have broken their internal ABI by mistake [05:50] sure, I just want to get it working at this point :P [05:50] buy better hw :) [05:51] what are the chances that there actually is an error in the computer hardware? [05:51] no idea.. [05:52] it's not necessarely an error [05:52] for example.. you get laptop A and B [05:52] they come from the same manufacturer [05:53] but perhaps one has a different resistor to cope with pcmcia version foo [05:53] and the other has a different one because foo was broken [05:53] you card works on A but on B [05:53] what is at fault? [05:55] Sure, so a change in the ABI could result that it won't work on B but still on A. [05:56] perhaps [05:56] that's a possibility [05:57] An oops is an error IMHO though, either caused by incompatibility between driver and hardware or the ABI itself. [05:59] Are there any package archives of the restricted repository? [06:01] no [06:01] there was morgue.ubuntu.com but it's temporary down [06:02] ok [06:03] temporary down as in packages are not archived, or temporary down as in server? [06:03] *server down [06:05] oh, I see what you mean :p [06:12] fabbione: one question... if some PCI hardware is unrecognized by the kernel (i.e. no modules to handle it), a "PCI: device has unknown header type X, ignoring." message will be printed, right? [06:13] hmmm i am not 100% sure about it.. i tend to buy hw that actually works [06:13] grep for that string inside the source and see what code prints it [06:13] fabbione: my card worked before :P I just downgraded the drivers to -7 but it doesn't seem to recognize the card now :\ [06:14] i should add an openafs BOF [06:14] to the ubz wiki [06:15] dilinger: are you going to be there? [06:15] if so i am up for it [06:15] i like this kind of challenges :) [06:16] but you will also need to lead it [06:16] not that many have afs experience [06:16] that's find [06:16] fine [06:16] at the very least, openafs clients would be useful for lots of schools [06:16] i haven't really tried touching our afs fileservers yet [06:17] isn't afs supposed to distribute everything in such a way that there is no real central server? [06:17] i wonder if we can write a torrentfs to achieve that.. [06:17] well, there are fileservers for each cell [06:17] so that everybody has everything :) [06:17] openafs is neat :) [06:17] for example, our homedirectories are spread out across 3 machines [06:17] dilinger: #define cell ? [06:18] oh, cell is just an organizational unit [06:18] afs has a global namespace [06:18] /afs// [06:18] sounds ldpaish [06:18] where for us is 2702.athenacr.com [06:18] ldapish [06:18] and another cell might be foo.stanford.edu [06:18] and so on [06:19] ok [06:19] sort of workgroups.. [06:20] yea [06:21] so basically you have N servers syncing automatically data for /afs// [06:21] and clients connect to servers indipendently.. [06:22] if i get this right [06:22] there's also client caching involved [06:22] yes i recall that [06:23] i still have all the AFS docs printed on paper [06:23] never had the time to read them [06:23] we typically set aside at least 100gb of cache for clients === calc [n=ccheney@ip70-185-4-246.ma.dl.cox.net] has joined #ubuntu-kernel === fabbione goes for dinner === calc [n=ccheney@ip70-185-4-246.ma.dl.cox.net] has joined #ubuntu-kernel === calc_ [n=ccheney@ip70-185-4-246.ma.dl.cox.net] has joined #ubuntu-kernel [10:17] jbailey: i know this isnt your thing, but in case you are interested: [10:17] i removed the dependency of video-buf-dvb from v4l-kernel cvs [10:17] and it should compile fine now on ubuntu [10:17] kernel 2.6.10 [10:17] mkrufky: Nice! [10:18] mkrufky: Is that a package in Ubuntu now? [10:18] no [10:18] it is still kernel modules ... i think you dont allow packages for those [10:19] if you like, i can provide packages to ubuntu ... but it would be silly.... much easier to install from cvs IMHO [10:29] Yeah,I guess. [10:29] We should probably find some way to make that possible at some point. [10:30] mkrufky: Might be worth suggesting as a spec - a way of forcing a build-time dependancy on another package that causes it to get rebuilt when the other package changes. [10:30] Or something like that. [10:34] hmmmm... gentoo has that in place [10:34] because packages are build from source [10:34] Well, packages are built from source in Ubuntu too. [10:34] It's just done on the server rather than on the client. [10:34] ya, thats the difference [10:35] truth is, i CAN release source to you that will build on any kernel [10:35] we maintain backwards compatability with older kernels [10:35] and i just fixed the biggest backwards-compat bug (the video-buf-dvb dependency) [10:36] so, i think i just have to learn about how baz works.... maybe it will all just fall into place? [10:48] Well, probably not for breezy. =) [10:48] Because it would be lovely if it did for Dapper. [10:49] Esp. since it will be supported for 5 years. [10:49] It would be nice if folks who need it for a little while can hope to keep it up to date [10:54] 5 years!?! [11:02] yup [11:02] 3 years on the desktop, 5 years for servers. [11:10] hmm, well, with the rate they're making these video capture cards and digital tv cards, 5 years is too long to go without v4l/dvb updates [11:10] lol... 1 month is too long sometimes [11:10] this is the reason why we maintain backwards-compat in v4l-cvs [11:14] Well, I do't imagint most servers will have dvb needs. =) [11:15] you're totally forgetting all about broadcast servers === mkrufky is the devil's advocate [11:16] =) [11:16] To forget about them assumes I knew about them in the first place. [11:19] how else do you think they get digital tv into your living room? === jbailey blinks [11:20] lol [11:20] the only digital TV I've been exposed to was something called "DirecTv" [11:20] I think it might still be around. [11:20] what country do u live? [11:20] it is still around, yes [11:20] Canada. [11:20] Ah, cool. I did the wide area network integration documents for them. =) [11:21] And the last time I saw anything relating to them was in 1993. =) [11:21] ah nice [11:21] they're still marketing in the states [11:21] Cool. [11:21] but i am anti-satellite [11:21] lol [11:21] i live in NYC [11:21] so i use cable [11:21] I lived in NYC for a year when I was working there. =) [11:21] satellite is better where cable is not available, IMHO [11:21] where in nyc? [11:22] i live in brooklyn, grew up in queens [11:22] For 9 months, midtown. 52 and Broadway. [11:22] (work in new jersey) [11:22] For a couple month, the Bronx, near Fordham station. [11:22] ah nice... [11:22] i dont know bronx very well [11:22] It was a learning experience. [11:22] I spent most of my time staying at other people's places. Much safer. =) [11:22] haha how is that?!? [11:23] Well, in the end we moved because there was a shooting outside of our appt. [11:23] omg [11:23] i lived in nyc all my life and still never was that close to a shooting [11:23] High crime, and it's a bit weird to be the only white person in the neighbourhood. [11:23] hmm... i guess thats a quirk about bronx [11:23] After the first few days it was alright, but I learned alot about multi-cultural relations there. [11:24] maybe thats why i dont know it so well :-P [11:24] We don't have anything quite like that here. [11:24] ya i know what u mean [11:24] All told, though, my experience in NYC was positive. [11:24] I still miss it sometimes. [11:24] ya, nyc is an experience [11:25] ive been to canada twice [11:26] only toronto [11:26] Toronto is really the asshole of Canada. [11:26] Are you coming up for UBZ? [11:26] UBZ... i know i SHOULD know what that is [11:26] "Ubuntu Below Zero" [11:26] ... ? [11:26] hmm [11:26] It's the Canonical conference here in Montral. [11:26] interesting [11:26] when? [11:26] We hold two conferences a year. The one in the Spring was in Sydney [11:27] No idea where next spring will be. [11:27] i see [11:27] October 30th to Nov 13th, although most people only come for part of it. [11:27] well, maybe would be fun to go [11:27] The big community day is the 30th. [11:27] The first week is spent figuring out what we want to do for the next Ubuntu release. [11:27] oh, thats in 2 weeks [11:27] The second week is spent doing Launchpad stuff. [11:27] i do have a vacation time left for this 2005 year ... hmmmm........ [11:27] Yes, very soon. =) [11:28] is there a site i can read about it? [11:28] https://wiki.ubuntu.com/UbuntuBelowZero [11:28] does it cost money? [11:29] page loading [11:29] fyi: i have some insider v4l news... i wanna spill the beans [11:29] ivtv project is merging into v4l .... [11:29] I don't think the conference itself costs money. You have to cover your lodging and stuff. There were sponsorships available, but I think those were all decided a while ago. === jbailey googles ivtv [11:29] so, im sure users complain about ivtv conflicts with kernel... worried no more [11:29] s/worried/worries [11:30] eh, i can afford lodging [11:30] damn page wont load [11:30] I'm lucky. I'm neither a kernel developer. nor do I own a television. =) [11:30] hah really? [11:30] Try using http instead of https? [11:30] k [11:30] I haven't lived in a place with a TV in like 8 years. [11:30] And then it was only fgor a year, and probably another 8 before that. =) [11:31] to be honest, you're probably better off [11:31] Oh, I'm sure I am. =) [11:31] I rent DVDs of TV shows. [11:31] No commercials. [11:31] heh ... less mush in the brain than most [11:31] ooooooh... i take it all back [11:31] Paying to have someone advertise to me seems to be the height of sillyness. =) [11:31] yes... but what about canadian news? [11:32] http://www.cbc.ca/ [11:32] i hear it's more objective than usa [11:32] true, web [11:32] I use the RSS feeds to see the top stories. [11:32] aha... i gotcha! u say u dont own a tv, yet u rent dvds? [11:32] There are also the free daily newspapers. [11:32] ...watching dvds on pc? [11:32] Sure. decss isn't illegal here. =) [11:33] Congrats on the project merger. If Conexant's video ochips are as annoying as their other chips, this should save you a bunch of work. =) [11:34] yes [11:34] the PITA is that ivtv is NOT kernel subsystem, and v4l is [11:34] so v4l is official drivers [11:34] and ivtv are directly incompatable [11:34] Ah, so still some work to do. [11:34] u must remove v4l to get ivtv to work [11:34] well, that WAS the problem [11:35] Any way to get it fast enough in userspace and just push it all out? [11:35] well, i think we'll have all the ivtv stuff ready for 2.6.17 [11:35] this cannot be userspace - its all kernel modules [11:35] (unless i am misunderstanding something) [11:35] btw, 2.6.14 comes out probably this week [11:36] you can always allocate iospace and pci bus memory to userspace. [11:36] (with a bounce buffer on really broken architectures, but it's still possible) [11:36] I don't know if Linux has a way of doing IRQ notifications. [11:36] ya this is true, but it is not what we are trying to do [11:36] however [11:37] we are planning to change the driver model such that chip support will be in kernel [11:37] but [11:37] card support will be configured form userspace [11:37] Right, but it might solve some problems if your in-kernel stuff was just enough glue to let the chip support sit in userspace. [11:37] hmmm [11:37] Someone at OLS 2 years ago managed to put the IDE driver in userspace, so what you need is likely there. [11:37] sounds like a lot of work [11:38] No idea. [11:38] Most of my kernel knowledge comes from hacking the Hurd. =) [11:38] And paying attention at OLS. =) [11:38] hurd, now that i never played with [11:39] It's a great way of learning, because there's still so much to look at that's unfinished. [11:40] hmm [11:40] is hurd used anywhere in production === mkrufky doubts it [11:41] Unlikely. [11:42] maybe one day i will mess with it [11:42] for now, linux is too much fun [11:42] lol