| Sarvatt | RAOF: got a nouveau-kernel-source that works on 2.6.32 on a PPA by any chance? | 01:22 |
|---|---|---|
| Sarvatt | or does the one in edgers work? | 01:22 |
| Sarvatt | something tells me this 185.36 nvidia blob isn't new enough to support xserver 1.7+.. | 01:24 |
| Sarvatt | yep we're gonna need to upgrade the nvidia blob to 190.xx at least for xserver 1.7 | 01:29 |
| === mac_v_ is now known as mac_v | ||
| unggnu | hi all | 13:16 |
| unggnu | Does anyone know if someone is working on a KMS vesa driver? | 13:16 |
| tjaalton | ask ajax on #xorg-devel | 13:16 |
| unggnu | This would make KMS available for everybody, at least without acceleration :) | 13:16 |
| unggnu | thx | 13:16 |
| tjaalton | but don't hold your breath | 13:16 |
| unggnu | Isn't it supposed to be the easiest integeration since it has no acceleration? | 13:17 |
| unggnu | Does current fglrx not work with 2.6.32 or is it a problem with the Ubuntu kernel? | 13:22 |
| tjaalton | it never is a kernel problem :) | 13:33 |
| unggnu | Because I have read somwhere else that 2.6.32 is supported through fglrx | 13:34 |
| unggnu | I just was interested how fglrx deals with radeon KMS in lucid | 13:36 |
| unggnu | *were | 13:36 |
| jcristau | i guess badly, but then it's fglrx so that was an easy guess | 13:37 |
| tjaalton | I don't think the blobs will/can ever support kms | 13:38 |
| unggnu | Yes, I know, I just wanted to know if there is a problem when using radeon and switching to FGLRX | 13:39 |
| unggnu | I am only using FGLRX because of Powerplay | 13:39 |
| unggnu | 2D acceleration is very poor and XV still tears | 13:40 |
| unggnu | I have to use OpenGL video output, this driver is a mess | 13:40 |
| unggnu | but OpenGL doesn't tear only if vsync is forced and no composite is used | 13:40 |
| unggnu | ciao | 14:01 |
| tormod | unggnu had some good points here though, about fglrx | 17:12 |
| tormod | proper radeon KMS support needs early loading of radeon, and fglrx does not like radeon I guess | 17:13 |
| tormod | so it has to decide in early boot whether to for radeon or fglrx | 17:13 |
| tjaalton | same thing with nvidia | 17:14 |
| tjaalton | don't know how to deal with that.. | 17:16 |
| jcristau | the closed drivers already deal with moving libGL.so.1 and libglx.so away | 17:17 |
| jcristau | they can add radeon.ko and nouveau.ko | 17:17 |
| tjaalton | hmm, right | 17:17 |
| tormod | does fglrx install its own libGL? | 17:18 |
| tjaalton | of course | 17:18 |
| tjaalton | and diverts the mesa one | 17:18 |
| tormod | so they should just install a do-not-load-radeon modprobe blacklist then, or move the module but that's difficult since it's at a new place for every kernel | 17:20 |
| tormod | and blacklist probably don't stop modalias loading | 17:21 |
| jcristau | it does | 17:22 |
| tormod | wasn't there some talk about a fglrx.ko which could coexist with radeon.ko? | 17:22 |
| jcristau | that's its whole point actually | 17:23 |
| jcristau | also, modinfo radeon|awk '/^filename:/ { print $2 }' | 17:23 |
| tormod | the point of blacklists? | 17:23 |
| tormod | jcristau, your command lists the current location of the radeon.ko. but if you upgrade the kernel, fglrx will have to move the new one. | 17:25 |
| jcristau | yes, the point of blacklists is to prevent automatic loading | 17:25 |
| jcristau | true re: the modinfo thingy | 17:25 |
| tormod | I thought the blacklist was dealt with by modinfo | 17:26 |
| jcristau | modprobe | 17:26 |
| tormod | yes :) | 17:26 |
| jcristau | udev uses modprobe -b | 17:26 |
| tormod | oh does it | 17:26 |
| jcristau | which doesn't do anything if the module is blacklisted | 17:26 |
| tormod | so that's why it is so slow :) | 17:27 |
| tormod | where does udev do the autoloading? | 17:29 |
| tormod | /lib/udev/rules.d/80-drivers.rules | 17:30 |
| tormod | so why does this not load the radeon module? | 17:35 |
| jcristau | do you have CONFIG_DRM_RADEON_KMS=y? | 17:35 |
| tormod | yes | 17:35 |
| jcristau | dunno then | 17:35 |
| tormod | sudo modprobe -b pci:v00001002d00005653sv00001025sd00000070bc03sc00i00 | 17:38 |
| tormod | FATAL: Module pci:v00001002d00005653sv00001025sd00000070bc03sc00i00 not found. | 17:38 |
| jcristau | modinfo radeon should list some modaliases.. | 17:39 |
| tormod | it does list my card | 17:39 |
| jcristau | maybe check how /lib/modules/`uname -r`/modules.alias is generated | 17:40 |
| tormod | funny enough my card is in there too | 17:41 |
| jcristau | weird. | 17:42 |
| tormod | I think the kernel does tell udev about it | 17:43 |
| tormod | does not | 17:43 |
| jcristau | tormod: udevadm info --export-db shows the modalias? | 17:47 |
| tormod | yes | 17:48 |
| tormod | but that is now, after radeon has been loaded "manually" | 17:48 |
| jcristau | oh, it's not there before? | 17:49 |
| tormod | I don't know :) | 17:49 |
| jcristau | i'll try and see what happens when 2.6.32 hits sid | 17:49 |
| tormod | week old kernel release, in Debian? :) | 17:50 |
| tormod | maybe I should try this before loading radeon | 17:50 |
| jcristau | well, we're at -rc8, and 18:18 < CIA-3> waldi * r14741 /dists/trunk/linux-2.6/debian/changelog: debian/changelog: Prepare to release (2.6.32-1). | 17:52 |
| tormod | brb | 17:56 |
| tormod | must be something with modprobe's alias resolving | 18:16 |
| tormod | for instance, modprobe --resolve-alias pci:v00001389d00000002sv*sd*bc*sc*i* -> applicom | 18:27 |
| tormod | modprobe --resolve-alias pci:v00001002d00005653sv*sd*bc*sc*i* -> (nothing) | 18:28 |
| tormod | omg plz forget everything I said while I go hiding, I just discovered an old modprobe file I had put there for debugging some stuff months ago... | 18:32 |
| jcristau | hah. | 18:33 |
| Sarvatt | so many problems with intel now :( this happened 3 times today freezing x - http://paste.ubuntu.com/336069/ | 19:04 |
| tormod | ok, I have been spreading a fair amount of disinformation about radeon KMS, but the initialisation race can still be a problem. Here, kms can take up to 1.7s to initialise and nothing stops X from starting before it is ready. | 19:08 |
| tormod | I wish upstart could log events with timestamps | 19:08 |
| Sarvatt | there isnt a wait after loading it in the initramfs | 19:09 |
| Sarvatt | ? | 19:09 |
| tormod | Sarvatt, not for radeon, it is loaded later | 19:09 |
| jcristau | i don't think it should be loaded in the initramfs | 19:09 |
| Sarvatt | hooks/framebuffer packs in all gpu and agp modules and it waits for fb0, dri/card0 and fbcon to load as well before moving on from what I see, I'm probably misunderstanding though | 19:17 |
| Sarvatt | looking at initramfs-tools-0.92bubuntu55 which is waaay different than debians | 19:20 |
| tormod | this is only in use if you boot with video= right? | 19:22 |
| tormod | uh no | 19:23 |
| Sarvatt | http://paste.ubuntu.com/336085 | 19:31 |
| yofel | hi, would bug 491483 be SRUable? | 20:10 |
| ubottu | Launchpad bug 491483 in xorg "Since failsave-x was enabled in karmic it starts if gdm is disabled and kdm is used" [Undecided,Confirmed] https://launchpad.net/bugs/491483 | 20:10 |
| Sarvatt | yeaaah I give up on attempting to package i915 gallium tonight :D | 22:10 |
| johanbr | hmm... xserver-xorg-video-nouveau: Depends: linux-nouveau-modules which is a virtual package | 22:44 |
| johanbr | did I manage to hit the PPA just while it's being updated? | 22:46 |
| johanbr | (the PPA being xorg-edgers) | 22:47 |
| Sarvatt | nope, sounds like it was synced with debian instead of ubuntu | 22:47 |
| johanbr | oh | 22:48 |
| johanbr | so the dependency should be on nouveau-kernel-source? | 22:49 |
| Sarvatt | just checked and the one waiting to build uses linux-nouveau-modules too, i uploaded that 10 hours ago.. | 22:49 |
| Sarvatt | yeah its linux-nouveau-modules in debian | 22:49 |
| Sarvatt | i'll upload a new one but it'll be a looong time until it builds at the rate ppa builds are going | 22:50 |
| johanbr | alright | 22:51 |
| johanbr | but I just downloaded the source package, which does have Depends: nouveau-kernel-source in debian/control | 22:51 |
| johanbr | weird | 22:51 |
| Sarvatt | oh? what version is it? | 22:52 |
| Sarvatt | are you on karmic or something? | 22:52 |
| johanbr | oh, never mind, wrong version | 22:52 |
| johanbr | I didn't put in the deb-src line for the ppa | 22:52 |
| johanbr | I'll see if I can fix it locally in the meantime... thank you for the help | 22:54 |
| Sarvatt | no worries, updated source should be up in about 5 minutes if not, just finished uploading it | 22:55 |
| Sarvatt | hmm, is a xbox 360 controller supposed to use -joystick? shows up as a mouse here | 23:00 |
| johanbr | hmm... the package compiled, but didn't work too well: "drmOpen failed" | 23:06 |
| johanbr | http://pastebin.com/f595ad9b5 | 23:07 |
| Sarvatt | theres no dri for it | 23:12 |
| Sarvatt | trying to remember what options i used with it, page i wrote it on has changed :( | 23:13 |
| jcristau | well nouveau needs the drm | 23:14 |
| jcristau | but he didn't paste dmesg, so.. | 23:14 |
| Sarvatt | ahh maybe he didnt force KMS on with the module option, i remember having that same problem now | 23:27 |
| Sarvatt | found my old dmesg from it working http://sarvatt.com/downloads/nouveau-kms.txt | 23:27 |
| Sarvatt | i couldnt start x without KMS but that was like 6 months ago | 23:28 |
| Sarvatt | same exact gpu though | 23:29 |
| johanbr | well... got a little further... the nouveau kernel module loads fine, but then the system locks hard when gdm starts | 23:33 |
| Sarvatt | try booting with nouveau.modeset=1 added to grub johanbr | 23:34 |
| johanbr | alright, I'll give that a try, thanks again | 23:34 |
| Sarvatt | i had the same problem on the same card 6 months ago when I was messing with it, had to use KMS to get it all working | 23:35 |
| Sarvatt | should probably ship a modprobe.d rule for nouveau with the package | 23:36 |
| tormod | Sarvatt, johanbr, if the new nouveau dep is wrong, just use my old package | 23:37 |
| tormod | that dep is most of the change anyway | 23:37 |
| johanbr | I just replaced it with nouveau-kernel-source and rebuilt | 23:38 |
| Sarvatt | heres what my logs looked like with KMS, when i tried without KMS it got stuck trying to restart gdm constantly http://sarvatt.com/downloads/nouveau-kms.txt | 23:40 |
| Sarvatt | nvidia didnt have a blob that worked on xserver 1.7 back then after the video abi changed, at least 190.xx and up works now | 23:41 |
| Sarvatt | i'll try it out right now, purging 195.22 i manually installed | 23:43 |
| johanbr | about to do the same | 23:47 |
| Sarvatt | i cant even build the kernel module | 23:51 |
| Sarvatt | /usr/src/linux-headers-2.6.32-6-generic/arch/x86/Makefile:81: stack protector enabled but no compiler support | 23:51 |
| Sarvatt | think ccache is screwed up on this machine | 23:52 |
| Sarvatt | nouveau sets a initramfs rebuild trigger before it makes the module? | 23:54 |
| Sarvatt | then it does it again after the module is built, ah | 23:55 |
| Sarvatt | yep it aint working right now: http://sarvatt.com/downloads/gallium.txt | 23:59 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!