=== CShadowRun is now known as cshadowrun === cshadowrun is now known as CShadowRun [01:41] yo [01:56] heya LLStarks [01:57] karmic kernels are in the repos. [01:57] kms disabled by default... [01:57] D: [01:57] brb [02:53] finally got kms to boot to x [03:23] LLStarks: how'd it go? [03:23] nice. [03:23] plymouth doesn't want to play nice. [03:27] LLStarks: would you mind writing up notes on what you had to do to get it going? it'd give us a headstart on getting it all enabled and set up by default [03:27] i915 modeset=1 [03:27] in /etc/modules [03:28] 2.6.30-1 kernel [03:28] did X pick up the mode properly? [03:28] latest edgers ppa drivers [03:28] i can't tell. the fb console looked right [03:28] but i still saw blinking [03:28] also, the boot is 75% done by the time it gets there [03:29] *gets to enabling kms [03:30] also, the karmic forums mention this: http://ubuntuforums.org/showpost.php?p=7166503&postcount=8 [03:30] CONFIG_FB_UVESA=y [03:30] CONFIG_FB_VESA=y [03:30] can't be modules [03:30] have to be compiled in [03:30] mm [03:30] yeah those look like kernel config params [03:30] sarvatt is doing the intel drivers for the ppa now. [03:30] kms enabled. [03:31] as opposed to tormo disabling them [03:31] cool [03:31] which ppa? [03:32] edgers [03:32] ok [03:34] for manual kms. modprobe i915 modeset=1 [03:34] also, /etc/modprobe.d/i915-kms is now present. [03:35] option i915 modeset=1 [03:35] is the only line [03:35] also, the intel team is forcing the issue with uxa-only [03:35] latest commits have stripped out exa and xaa [03:36] yeah I know about that [03:37] also, xv overlay is gone. [03:37] I'm concerned that's going to paint us into a corner if we want to ship something newer than 2.7 [03:37] for kms+uxa [03:37] hrm [03:41] and i just froze... [03:41] lovely [03:42] bryce. uxa is not ready for karmic as things stand. [03:42] btw, is there anyway to unfreeze x if it happens? [03:42] not usually [03:42] i'm using exa right now. [03:42] however there have been cases where I've been able to unfreeze by killing the process which causes it [03:43] mplayer in my case. [03:43] but i don't have a spare machine for ssh [03:43] in one case, that was alarm-clock, in another case you could kill compiz and get back [03:43] in that case unless vt switching works, you're power cycling [03:44] what's causing these crippling crashes? [03:44] or play with sysrq combinations [03:44] careful to distinguish between "freeze" and "crash" - different things [03:44] what causes freezes is usually something poked into the GPU improperly [03:44] alt+prntscrn+b is the only thing that works [03:44] and that reboots. [03:45] I wrote a lengthy guide about freezes in wiki.ubuntu.com/X/Troubleshooting/Freezes if you're interested in the gory details [03:46] also you can install the intel-gpu-dump tools and get gpu dump info that presumably reveals why it froze [03:46] however my experience so far with the tools has been a mixed bag === jldugger is now known as pwnguin === imbrando_ is now known as imbrandon [13:49] Know how a few days ago I said X kept doing the thing where it freezes but the mouse still moves? Turned on greedy, and it's perfectly stable now. === crevette__ is now known as crevette === CShadowRun is now known as cshadowrun === cshadowrun is now known as CShadowRun [16:53] bryce: you may be able to just override pScrn->virtualX & Y in preinit instead [17:43] jbarnes: ok [17:47] is there any particular reason not to always make the virtual be its maximum size for that hardware? [17:48] Ng: yes, on some hardware it's huge [17:50] so that'd be pointlessly stealing RAM? [17:50] I thought I read somewhere that the maximum on Intel hardware is only 4096x4096 on newer chips [17:50] Ng: that's a mesa bug [17:51] oh :) [17:51] Ng: there's a patch floating around that never made it into mesa that bumps up the limiter to 80somethingx80something [17:51] heh [17:52] eh wait [17:52] hmm [17:52] was it 4096 or 2048 now [17:52] lemme get that patched mesa out and give it a go [17:53] 5120x3200 would be the size of 4 30" monitors by my calculations, so that would seem to my naive guesswork, to be a reasonable maximum at the moment [17:53] (I don't know how you'd connect more than 2 to most intel driven hardware, but at least it means you can arrange them in either axis) [18:08] LVDS + VGA + HDMI? [18:08] that's 3 [18:08] i don't see how you could get a fourth [18:08] wait, does intel even have hdmi? [18:09] or was HDMI for audio? i can't remember [18:09] Ng: you can connect more, but you have 2 crtcs, so the rest has to be cloned. [18:09] bah [18:18] 2 crtcs? [18:25] I only own three things with intel chips, but none of them have more than 3 connectors === CShadowRun is now known as cshadowrun [19:21] hi [19:21] is there any issue with Ubuntu 9.04 having 2 nVidia card ? [19:21] the installer doesn't come up, if we have 2 cards [19:21] it gets struck when the X comes up [19:21] later, tried removing a card and installed Ubuntu 9.04, after installation, when I plugin the card, the second card is not being detected [19:22] both the cards are same nVidia chipset [19:22] I mean the lspci is listing only one card [19:46] nagappan: in general multi-cards is not supported in ubuntu [19:46] nagappan: however in theory the -nvidia driver can do it, but I've never messed with it myself. [19:47] bryce, ok [19:55] bryce, I also noticed this in /var/log/messages [19:55] Apr 28 20:07:53 xerox-1-2-dhcp77 kernel: [ 21.552051] Xorg[3413]: segfault at fffffffffffffff8 ip 00007f0804551894 sp 00007fff0c75f280 error 4 in ld-2.9.so[7f0804543000+20000] [19:56] dunno that one [19:58] hi all [19:59] fffffffffffffff8 is a fun address. [19:59] Does anyone know a howto to test kms with Jaunty? Karmic doesn't seem to have the new packages. [19:59] nagappan: maybe try with "sudo nvidia-xconfig --multigpu=On" (type man nvidia-xconfig for further information) [20:00] tseliot, sure, let me try now, thanks :) [20:00] bryce, will this be of any use ? http://pastebin.com/d79e70d92 dmesg output [20:00] I noticed this line [20:00] [ 10.674454] ck804xrom ck804xrom_init_one(): Unable to register resource 0x00000000ffb00000-0x00000000ffffffff - kernel bug? [20:00] nagappan: in general debugging a blob will be pretty hard. [20:01] jcristau, ok [20:01] (read: impossible) [20:07] tseliot, there is no such option with nvidia-xconfig, atleast on this Ubuntu 9.04 64-bit machine [20:08] nagappan: "--sli=Auto" should work [20:08] note, the same card with same configuration is working with Ubuntu studio rt kernel ! [20:08] tseliot, let me check [20:09] tseliot, option --sli not recognized [20:09] nagappan: what driver are you using? [20:09] tseliot, nvidia driver [20:09] what version [20:10] * nagappan checking [20:10] tseliot, nvidia-installer: version 1.0.7 (buildmeister@builder58) Fri Apr 17 00:40:22 PDT 2009 [20:11] tseliot, nvidia-settings: version 1.0 (buildd@crested) Sun Feb 1 20:25:37 UTC 2009 [20:11] nagappan: didn't you install it using Ubuntu's repositories? [20:11] tseliot, no [20:12] tseliot, I tried with the NVidia installer, let me try with Ubuntu version now [20:12] nagappan: no, please don't [20:12] tseliot, :) [20:12] or you'll break your system [20:13] sudo sh name_of_the_installer --uninstall [20:13] tseliot, sure [20:13] and uninstall the driver from the installer first [20:13] tseliot, ok [21:00] jbarnes: hrm, i've been poking the virtualX/Y values in preinit but seems to not affect things [21:00] jbarnes: at least, I see no change in xrandr's maximum values [21:00] oh hm [21:06] bryce: what happens? [21:06] with the virtual values [21:06] tseliot: xrandr reports 1280 1280 (the stock defaults for this hw) [21:06] bryce: are you setting it before xf86initialconfiguration gets called? [21:06] instead of 2048 2048 that I'm trying to poke [21:07] jbarnes: I've tried it several places in the function, including up at the top right after i830_kernel_mode_enabled() [21:07] looks like you might actually have to poke pScrn->display->virtualX/Y too [21:07] one thing I've found is I can only do the I965 test after about halfway through the routine, otherwise it segfaults, but that's a secondary issue [21:08] tseliot, sorry was away for lunch, just back, trying now [21:08] yeah the pci stuff has to happen first otherwise you won't have a pdev to check [21:15] jbarnes: no go [21:16] bryce: arg I guess there's not an easy way to do it [21:16] bryce: another way of achieving the same thing might be to round up the memory allocation though in the 965 case [21:17] where we allocate the front buffer you could make sure it's at least 2048 wide [21:17] that's in i830_mem.c? [21:17] i830_allocate_framebuffer ? [21:17] wouldn't change the virtual or xrandr output but would affect memory laytout the same way [21:17] yeah [21:34] i've just built mesa with a custom patch that upstream wanted me to try... to install it I tried doing "dpkg -i *.deb" but since all the generated DEBs depend on each other I think they have to be installed in some special order or something... when I do "dpkg -i *.deb" I get this error --> http://pastebin.com/m4337f2a9 [21:34] some are conflicting with each other [21:34] which makes dpkg -i *.deb sort of a problem [21:34] ahh, so I should just install a subset then? [21:34] so, err, don't do that. [21:35] which ones do I want for DRI radeon? [21:35] libgl1-mesa-glx and libgl1-mesa-dri [21:36] mnemo: fwiw, I exclude the swx packages when installing mesa [21:36] otherwise it bumps you into software rendering or some such [21:38] ok and then I have logout and login again? [21:38] or is there some magic way similar to what modprobe does for kernel? [21:40] I usually restarting gdm (or for the freeze issue do a fresh reboot just to be sure). dunno if there's a simpler method [21:42] mnemo: depends what your issue is. if it's related to aiglx, then you need to restart the X server. if it's direct rendering, then you don't have anything to do. [21:43] jcristau: ok, but if its direct rendering and im using compiz I need to restart compiz at least right? [21:43] compiz doesn't use direct rendering (on dri1 anyway). so restart X. [21:44] ok [21:44] jbarnes: yeah setting pScrn->display->virtualX seems to just crash X [21:44] ugg [21:45] out of curiosity I printed virtualX/Y throughout the preinit call, and it seems to never change from what I set it, so there isn't something that's resetting it [21:54] whoa [21:54] something I did just did it [21:54] Screen 0: minimum 320 x 200, current 1280 x 800, maximum 2048 x 2048 [22:01] bryce: heh cool [22:01] narrowing the patch down [22:02] I'd just started randomly moving the display->virtualX stuff at different spots in the routine, and I think I stuck it in a spot where it had the right effect [22:03] "development by a 1000 monkeys, eventually shit just has to work right" [22:03] hehe sry :) [22:03] ez true [22:04] cheep cheep [22:06] bryce: btw I think I managed to screw up my computer in exactly the way you told not to... i mean.. after install *.deb of mesa ... now "glxinfo | g direct" says "NO!" [22:06] is there anyway around that? [22:07] yeah, I think it is to do a apt-get install -f [22:07] which cleans up the mesa stuff [22:07] bryce your the guy who wrote the reply to the brainstorm on multi monitor support i think, right? === cshadowrun is now known as CShadowRun [22:07] CShadowRun: probably [22:07] cool, i'm all up for testing (quad screen user \o/) [22:08] been trying to prod gnome into fixing some long-standing multi X screen bugs recently lol [22:09] Federico Mena-Quintero has done some really nice work on xrandr fixes in GNOME lately as well [22:10] CShadowRun: what graphics card do you have? [22:10] mnemo i have a pair of 8800GT [22:10] one X screen per card, so 2 twinviews. [22:11] CShadowRun: link to the brainstorm page plz? [22:11] bryce i'll try and find it again, it was a long time ago though [22:15] don't think i can find it, i remember it was in development for intrepid, and that it had about 2000 +'s [22:15] and that there was a reply explaining about xrandr [22:15] and that testers where needed [22:16] well, -nvidia multi-card is completely different [22:16] and afaik is not an xrandr thing [22:16] bryce http://brainstorm.ubuntu.com/idea/206/ [22:17] i heard xrandr 1.3 was going to have shared graphics memory which would allow for multi cards? [22:17] yeah [22:17] still, that's orthogonal to -nvidia [22:18] yea, i know theres no support for the nvidia binary driver [22:18] those are the 2 things i'm waiting on :D [22:19] also another thing that wasn't mentioned in the article is alot of software really doesn't work well if you use multiple X screens [22:19] gnome-panel is a good example of that [22:19] * bryce nods [22:20] but i been trying to prod the gnome-panel dev into fixing some of that, hopefully [22:20] then i can move from hardy to jaunty \o/ [22:20] a lot of the Xinerama style setups seem to have suffered regression now that xrandr is the norm [22:20] good, keep at that [22:21] yea, i'm not using xinerama [22:21] i like the seperation of X screens [22:22] anyway nice talking to you, if you ever need anything tested i have a partition dedicated to testing stuff, so just drop me a line :) [22:22] sure, usually if we need testing I announce to ubuntu-x@lists.ubuntu.com [22:23] cool, sign up for that [22:23] CShadowRun: https://lists.ubuntu.com/mailman/listinfo/Ubuntu-x [22:24] ty [22:26] ok... moment of truth, time to verify this solves the freeze... [22:27] /mne keeps fingers crossed [22:41] has passed the 10 min mark, that's a very good sign [22:43] ok, ppa updated. I'm going to snag some lunch and let this laptop finish its testing [22:44] jbarnes: http://pastebin.ubuntu.com/160964/ - open to suggestions for improvement [22:45] bryce: yeah looks like a good hack [22:46] bryce: i would mention some "workaround" plus some LP bug number in the debug output like you did with that intel i865G bug [23:11] ok [23:11] (still going strong btw) [23:12] nice [23:16] hi phoronix :)