airlied: gallium modesetting with r300g isn't total fail
airlied: I can see gnome-terminals, and watch a movie
airlied: as long as i don't move a window
eosie: it doesn't get much testing on non-r500
airlied: eosie: I doubt it gets testing anywhere, modesetting_drv.so that is
airlied: or radeong_drv.so as I renamed it
airlied: looks like blit issues
eosie: it asserts (crashes) right?
airlied: not anymore
airlied: just fixed an uninitialised pointer issue
eosie: I fixed a bug 2 days ago that might have caused failures, not sure... it hasn't been pushed yet
soreau: eosie: Well I recompiled and restarted compiz and Opacity seems to look better but alpha blur doesn't work still
eosie: disable independent texture fetches
eosie: and don't use the mipmap filter
eosie: uninitialized pointer... good catch
tstellar: If I build mesa with --enable-gallium-radeon does it completely install the gallium driver, or is there another step I need to take?
eosie: five r300g fixes has yet to be pushed (it's sitting in MAD's mailbox)
bridgman: airlied, I was under the impression that the xorg state tracker had replaced the modesetting_drv driver, is that not the case ?
gregorian: Hi, my GNOME takes 3.5 minutes to load with the radeon driver but vesa loads it in 5 seconds. I have uploaded the log and configuration file to pastebin. Any help?
bridgman: was radeon the first gnome startup after boot and vesa the second ?
soreau: eosie: I tried with all three filters and changing all settings including independent texture fetch off. It doen't work at all
gregorian: No, I set radeon in the xorg.conf and saw it took 3.5 minutes. I then set vesa and it took 5 seconds
soreau: eosie: Though like I said, opacity is working better now
soreau: eosie: Before, semi-transparent white windows would cause an undesirable 'bright' strange color effect, but that seems to be gone now
soreau: Seems it's white on white, ie. a semi-transparent white window with white window under it
bridgman: gregorian, I think someone said before that gnome was always slow to start the first time but much faster second time
bridgman: if you switch xorg.conf back to use radeon does it boot slow again ?
gregorian: bridgman - even xterm takes 1.5 minutes to start. It is always slow, not just first time. Only vesa loads it fast, but at lower resolution
soreau: eosie: I should say light on light, because it's not necessarily only white windows, it just doesn't happen with dark colored windows and the issue is still present, I spoke too soon :P
airlied: bridgman: prety much, we just had the name modesetting_drv.so still
airlied: in gallium
airlied: eosie: you don't cc patches to dri-devel anymore?
airlied: thinks the blitter is failing on overlaps
gregorian: Anyone else?
airlied: gregorian: so if you boot with vesa and change to radeon it takes 3.5 mins?
eosie: airlied: all my r300g patches now go to MAD's mailbox because he understands most of the code and noone else would review them anyway because it's a lot of new code
gregorian: If I startx with radeon 3.5 minutes. if I startx with vesa 5 seconds but lower resolution
gregorian: startx starts GNOME
amarsh04: is the firmware for the radeon loaded, gregorian?
eosie: airlied: the only patches I sent to dri-devel were the kernel ones
amarsh04: dmesg|grep firmware for me reports "platform radeon_cp.0: firmware: requesting radeon/R200_cp.bin"
gregorian: after starting gnome, I can see radeon listed in lsmod if that's what you mean. I'm using a framebuffer console with radeonfb at the moment
eosie: and I used to send r300g patches to mesa3d-dev which I don't do anymore as you can see
gregorian: let me startx, otherwise radeon driver won't get loaded. back in 4 minutes.
eosie: wants a FDo account
airlied: eosie: did you apply for one?
gregorian: amarsh04: I get no output for dmesg|grep "firmware", but my lsmod shows me radeon
airlied: gregorian: did you build your own kernel? enable kernel modesetting?
gregorian: I configured and compiled the kernel using menuconfig
amarsh04: gregorian, what distribution are you running? I am running Debian, where I have to install package linux-firmware to get the non-free firmware
eosie: airlied: yes, I did
airlied: eosie: did it get approved by Brian?, though finding admins over Xmas might be a pain to kick someone
amarsh04: ah, gregorian, I haven't run gentoo, just that the slowdown sounds like the firmware isn't loaded
eosie: airlied: it was approved by both brian and alex
airlied: eosie: send me the bug num and I'll see if I can kick someone
airlied: though at this point of the year it'll be lucky
gregorian: amarsh04: So what would you advise me to do?
gregorian: I have logs at pastebin
amarsh04: I was about to google for "gentoo radeon firmware" and see if you get a hit
airlied: it should be in dmseg if the firmware fails
gregorian: I followed the Gentoo handbook completely and also its instructions for installing X and GNOME
airlied: eosie: you wrote the u_blitter? how does it handle overlaps when blitting to/from same surface?
gregorian: Can the driver load even if there is no firmware? Because it does
amarsh04: what does "grep -i glx /var/log/Xorg.0.log" show?
airlied: eosie: it looks like it doesn't deal and I shouldn't be using it for those cases
amarsh04: yes, I've run radeon without firmware
amarsh04: Debian removed the firmware from the kernel before releasing the firmware in a separate package
gregorian: amarsh04: The first line shows that GLX is loaded. There are few more line
gregorian: No, warnings or errors
amarsh04: ok, I'm probably out of ideas then gregorian
gregorian: I have uploaded the conf, radeon log and vesa log to pastebin if you think you can use them.
amarsh04: I just run a radeon 9200se on a PII and a radeon 3200hd on an AMD64
airlied: dmesg is the only useful thing I'd guess
eosie: airlied: yes, I did.. it doesn't handle overlaps, it just implements clear, surface_copy, and surface_fill gallium functions, other things should be xorg-st's job
airlied: eosie: AFAIK surface copy doesn't specify the two surfaces are different
airlied: though it doesn't say much
amarsh04: feel free to post the URL's of the patebin if you want gregorian, I'm only familiar with what I had to deal with to get my setups working
tstellar: gregorian: Are you using a gentoo kernel and did you run "make firmware_install" after building your kernel?
gregorian: I did not run make firmware_install since I did not find it in any of the Gentoo guides online. amarsh: I'll post them right after if tstellar can't get it to work.
gregorian: I'm using a gentoo kernel
tstellar: gregorian: What happens when you run: ls /lib/firmware/radeon
gregorian: It does not exist
tstellar: Ok, so go to /usr/src/linux and run make firmware_install, and then try ls /lib/firmware/radeon again.
eosie: airlied: the blitter always uses a colored or textured quad, so even though surface_copy is not expected to blit into a different format, the blitter allows it if it's accelerated
gregorian: I'll be back. Thank you!
eosie: airlied: I think the visual artifacts with xorg-st might be unrelated to the blitter module
airlied: eosie: wierd I only see them when I move windows at least
airlied: and the surfaces are the same
airlied: it looks like classic overlap failure
gregorian: tstellar: I still don't find radeon
airlied: eosie: I'll play around a bit more see if I can narrow it down to that
tstellar: gregorian: What happened when you ran make firmware_install ?
gregorian: It gave a lot of lines as expected, but it /lib/firmware does not contain the radeon firmware
eosie: airlied: the state tracker probably provides its own blitter capability as the blitter module is just an optional driver module not seen by the state tracker
gregorian: Do I need to install it the way you usually install wireless firmware? Or is it something in the kernel?
tstellar: gregorian: Ok, it looks like gentoo doesn't distribute the radeon firmware, so what you should do is download kernel source from kernel.org
gregorian: Is there an alternative way? I'm new to using the kernel and all and I don't want to break my Gentoo distribution.
tstellar: gregorian: You could try emerge linux-firmware
gregorian: I will do that.
gregorian: emerge-pv gave me something called masked packages. Is it safe to proceed?
airlied: eosie: it uses surface copy if its there for overlaps
airlied: eosie: and radeon uses the blitter
eosie: I see
tstellar: gregorian: Yes
amarsh04: back in a few, rebooting for new kernel
gregorian: emerge: I can't do it.
gregorian: tstellar: "All ebuilds that could satisfy sys-kernel/linux-firmware have been masked" There must be a reason for that right?
tstellar: gregorian: You need to add sys-kernel/linux-firmware to your /etc/portage/package.keywords file.
gregorian: It's emerging. I'll wait
gregorian: tstellar: There is still no radeon in firmware
gregorian: It's also given messages about possible file collisions.
gregorian: I've uninstalled the package since it did not contain the radeon firmware and ran make firmware_install once again.
gregorian: Does "Radeon XPRESS 200M 5955 (PCIE)" require firmware?
tstellar: gregorian: I actually have that exact same card and I am using gentoo.
gregorian: But you have the firmware?
gregorian: By card, did you mean integrated chipset?
tstellar: gregorian: Yes, but I am not using the gentoo kernel, my source is from kernel.org and it contains the necessary firmware.
gregorian: By any chance does the gentoo kernel have some option that needs to be enabled?
tstellar: You might want to ask over at #gentoo, but I don't don't see the radeon firmware included in the gentoo kernel sources.
gregorian: I don't see it either under firmware drivers. Does the option have radeon anywhere in it? Because I don't see anything
soreau: eosie: fwiw, with those patches, attempting to rotate cube crashes compiz
tstellar: gregorian: There is an option to build firmware into the kernel: FIRMWARE_IN_KERNEL, but it won't make a difference if you don't have the firmware you need.
tstellar: gregorian: I would recommend downloading a kernel from kernel.org and then you can copy the firmware files you need.
gregorian: Copy the files? Or replace the kernel?
tstellar: gregorian: I think you only need to copy the firmware files.
gregorian: Alright, but how will the gentoo kernel know to include the files?
gregorian: It's not part of it right?
gregorian: I don't mind compiling the new kernel. Is it possible to save the gentoo kernel configuration into a file and then load this file into someother kernel?
tstellar: gregorian: Yes you can do that.
eosie: soreau: I can't reproduce it here, could you bisect please?
wirry: the file is called .config and placed in /usr/src/linux/
soreau: eosie: Sure
gregorian: I will do just that. Be back after a loooong while. :) Thanks for all your help
eosie: airlied: this is my account request: http://bugs.freedesktop.org/show_bug.cgi?id=25585
soreau: eosie: Hmm.. not sure how to use git bisect with patches that have been 'git applied' :P
soreau: Might have to do it manually
soreau: No worries, just 14 patches ;)
soreau: eosie: What about branches like r300-glsl.. are these of any interest?
soreau: I usually just stay on master
eosie: soreau: well, there is no reason for the development to happen on a branch if it has never been stable, officially released, and shipped
soreau: eosie: I don't understand what you just said, but I will take that as a 'no' :)
eosie: I guess my english is not so good
Reilithion: I'd like to understand better about the radeon driver. Why does it work for some things and not for others? I think last time I tried to understand, someone told me about it needing a decent memory manager?
soreau: Reilithion: In short, there are missing features that are in the process of being implemented
Reilithion: soreau: These missing features -- Is there somewhere I can read about them? Keep up to date on their status? When they're done, will my older card (a Radeon X850, I think) be supported and be able to run, say, Danger from the Deep or Planeshift?
Reilithion: And is there anything I, as a very novice programmer, can do to help?
soreau: Reilithion: http://www.x.org/wiki/RadeonFeature
soreau: Reilithion: The best thing you can do is test the latest code and report any regressions or brokenness you might find. If you would like to get familiar with the code, read it :)
eosie: soreau: where is r300-glsl?
soreau: eosie: git branch -a
soreau: eosie: From git mesa dir
soreau: eosie: I see it last on the bottom
eosie: soreau: I don't
Reilithion: Perhaps this is a n00bish question, but I keep seeing r200, r300, r600, etc. What are these, and how do they relate to my x850?
soreau: Reilithion: http://en.wikipedia.org/wiki/Comparison_of_AMD_graphics_processing_units
Reilithion: Thanks. I'll read up.
soreau: According to that chart, x850 is an r4xx
wirry: yes it is
soreau: eosie: I'm bisecting and I'm all the way back to when airlied fixed mesa the other day and it still crashes so I don't think it has anything to do with your patches now
soreau: I am now curious as to what did break it, because iirc, it wasn't like this after I built from the other day
soreau: But I might have built it and never restarted compiz or something
soreau: digs deeper
Reilithion: Hmm. I think GLSL is the culprit that is keeping me from my precious games.
soreau: GLSL is the opposite of a culprit, it's something that is on the workbench :)
soreau: The lack of glsl is likely your problem
Reilithion: Yeah, that's what I meant.
soreau: I know :D
Reilithion: Ah, what is it about smartasses that I find so lovable?
Reilithion: Oh wait. It's that I am one myself.
Reilithion: I'm also trying to figure out why I had to disable the Gallium use flag in order to compile the x86-driver-ati package on my lovely Gentoo box.
soreau: Have you tried recently?
Reilithion: A day or two ago
soreau: It should work if you have -9999 versions of libdrm, mesa and xf86-video-ati
Reilithion: Maybe more, I did spend several very sleepless days reinstalling the whole system from scratch.
Reilithion: Is that something that'd show up if I went equery list -po libdrm ?
Reilithion: Bit of a n00b, as always
Reilithion: The versions installed are libdrm-2.4.17, mesa-7.7, and xf86-video-ati-6.12.4
soreau: Even with the gallium flag, you'd still be using classic mesa by default
gregorian: Hi, I just downloaded the latest kernel but I can't find any specific options to enable the Radeon firmware. It looks just like my gentoo kernel.
soreau: gregorian: Have you downloaded your firmware?
gregorian: When I tried to emerge it for the gentoo kernel, it was a masked package but I went ahead and did it anyway. Since I still didn't have it, I was told to download the latest kernel and enable the support for radeon's firmware. I'm supposed to download it now?
gregorian: Doesn't the kernel ship with it?
Reilithion: soreau: I'm reading up on version -9999 now, so let me know if this next question makes sense: Is there an overlay providing a "live" ebuild for those packages you mentioned?
soreau: Reilithion: Yes, the x11 overlay
Reilithion: w00t, I got it right!
soreau: gregorian: The no was for your second inquiry :)
soreau: gregorian: Seems tstellar was helping you earlier?
gregorian: Before that, I want to know if the right way to compile this new kernel is copy the .config file from the gentoo kernel to this one; point /usr/src/linux to this new one; make and make modules_install; and then configure lilo accordingly.
soreau: People still use lilo? :P
Reilithion: I know.
gregorian: I couldn't install grub easily :(
gregorian: Is what I said correct though?
soreau: gregorian: To compile a kernel with another .config, everything you said was right except you need to do 'make oldconfig' before the other make commands
soreau: after copying the old .config file to the new kernel
gregorian: I don't get it. If I overwrite the .config file with gentoo's, how will the kernel know it's been replaced?
soreau: gregorian: What I usually do is make the symlink to the new kernel, then cp .config and run make oldconfig, make, make modules_install, make install then configure the boot loader
gregorian: I've never used make install.
gregorian: so now it's copy; make oldconfig; make && make menuconfig; make install ?
gregorian: I mean make modules_install instead of menuconfig
Reilithion: How do I unmask single packages?
soreau: gregorian: Yes, that sounds right
gregorian: I'll do it.
Reilithion: (or rather, get single packages that are masked to install)
soreau: Reilithion: That is a bit beyond the scope of this channel, but I happen to know you should put packages to be unasked in /etc/portage/package.unmask
Reilithion: Thanks. I was pretty sure it was simple, otherwise I wouldn't have asked here.
amarks: tries vimperator
Reilithion: I'm starting to think I ought to comment my package.use file. :-/
Reilithion: soreau: It says I should rebuild xorg-server. Should I unmask xorg-server as well, and use the git version of it, or should I stick with the 188.8.131.521-r1 release from Gentoo?
soreau: Just stick with the release
gregorian: Hi, I just ran make oldconfig. I'm getting a set of questions. Do I say yes to everything?
soreau: gregorian: The capitalized letter is the one it will use by default, so you should be able to press enter for most if not all of them
Reilithion: Wow. Mesa is quite large...
gregorian: I'm not sure what you mean by capitalised letter, but if it means no harm. I'll just press enter.
soreau: Upper case
wirry: [Y,m,n] for example means "yes" is default
Reilithion: gregorian: Have you customized your kernel before? If you did, and the .config file in that directory is the one you used, you should probably be able to basically hold enter down the whole time. If you might need to make changes to the config, that's when you have to be paying attention.
gregorian: After pressing enter till the end, I can just re-run menuconfig and set what I need right?
wirry: or you just enter the menuconfig afterwards and change it there
wirry: dammit -.-
gregorian: Thanks :)
Reilithion: Someone, somewhere, is being mean to Gentoo users and devising a software project whose version numbers count down sequentially from 10000.
hnsr: what the ... :D
Reilithion: Oh boy, here we go trying to compile mesa...
wirry: use the x11-overla
Reilithion: Wow. Compiled ok.
Reilithion: Moving on...
Reilithion: Uh, hmm. It just occurred to me that X is running while I'm compiling this stuff...
Reilithion: Gosh, I hope nothing breaks...
Kurko: I can't get DRI working, I only get this error:
Kurko: (EE) RADEON(0): [dri] RADEONDRIGetVersion failed because of a version mismatch.
Kurko: [dri] radeon kernel module version is 2.0.0 but version 1.17.0 or newer is needed.
Kurko: [dri] Disabling DRI.
gregorian: You need to enable DRI in the kernel
gregorian: Once you do that, the error will go away
Kurko: how I can enable DRI in the kernel?
gregorian: I recently posted a thread in LinuxQuestions.org. Could you wait here? I'll try to find the link.
soreau: Kurko: You need to rebuild mesa and ddx against your current version of libdrm
Reilithion: Wow. Very nice. Wish I'd run some benchmarks before, 'cause it seems slightly slower.
eosie: soreau: I forgot to mention my patch set improves performance of blending a little bit, it might be interesting to test e.g. openarena and see if it improves anything
soreau: eosie: I am still trying to bisect mesa starting with the libdrm api change, but mesa woont build so I'm having to skip all of them :P
soreau: eosie: If I remember I will try OA, but I usually don't test it so I don't have too much to compare it with
soreau: I think last I tested OA was a few weeks ago
soreau: and it worked fine, just a bit laggy
eosie: do you know where I can find anholt's openarena timedemo files?
gregorian: I compiled the new kernel and booted from it. Before starting X, dmesg|grep firmware shows nothing. After starting X, the same command gives: platform radeon_cp.0: firmware: requesting radeon/R300_cp.bin
eosie: it's not on anholt's site nor anywhere else
gregorian: I did not download the radeon firmware because the folder was already present
amarsh04: ok gregorian, that's a good sign because it should mean that the firmware was loaded
gregorian: It doesn't come on boot though.
gregorian: I have to startx
gregorian: My GNOME takes 3.5 minutes as usual.
soreau: eosie: http://dri.freedesktop.org/wiki/Benchmarking ?
eosie: soreau: the links don't work
amarsh04: yes, that's what happens with me gregorian, the X server loading process loads the firmware for the radeon card
soreau: eosie: So ping anholt?
eosie: soreau: I suspect you don't have the files..
soreau: eosie: No I do not
Reilithion: Ok, is the GLX_EXT_texture_from_pixmap extension supposed to work for my card with the versions of software I now have?
soreau: eosie: It is possible adamk might have them
soreau: eosie: I am beginning to have a suspicion they may have gotten nuked when users home folders were lost the last time fdo went down..
soreau: eosie: http://people.freedesktop.org/~cbrill/dri-log/index.php?page=dri-devel-2009-11-11.log
eosie: soreau: the links from adamk don't work either
soreau: eosie: I know, but adamk might still have them
gregoria1: If the firmware's working, what do I do now?
soreau: eosie: Also, I'd ping anholt and see what he says about it
soreau: gregorian: Enjoy?
gregorian: I mean my problem's not gone?
soreau: What is your problem?
gregorian: GNOME takes 3.5 minutes to start
soreau: What makes you think it's a radeon driver issue?
gregorian: vesa loads it in 5 seconds
soreau: Not sure then
soreau: Have you filed a bug report already?
gregorian: No, I didn't think it was a bug.
gregorian: The person who helped me earlier said he was running Gentoo on the kernel downloaded from kernel.org and he had the same graphic chipset.
soreau: Did you say simply loading the module takes that long?
gregorian: I'm now the same as him
gregorian: No, I don't know how to detemine that.
Reilithion: tail -f /var/log/Xorg.0.log ?
gregorian: I have the whole log at paste bin if you want
soreau: Well the first thing I would do is eliminate any gnome stuff by stopping gdm and X then starting X very basically with 'startx' or so
gregorian: I can't do that.
gregorian: soreau: xterm takes 1.5 minutes to start
gregorian: Reilithion: Keyboard won't respond when X loads
soreau: Then try unloading radeon and drm modules then modprobe radeon
Reilithion: gregorian: I was suggesting you do this from a virtual terminal
gregorian: You mean rmmod radeon, drm and then modprobe radeon?
gregorian: Reilithion: How do I access a terminal when nothing responds?
Reilithion: uh oh
Reilithion: If he doesn't know about that, and tries it, he won't know about Ctrl-Alt-F7 to get back...
Reilithion: Or just Alt-F7
gregorian: My caplock won't work Reilithion
gregorian: There is nothing I can do in those 3.5 minutes
gregorian: when GNOME loads
Reilithion: You've tried switching to a virtual terminal and it doesn't work?
gregorian: soreau: I can't unload drm because something's using
gregorian: Reilithion: I type startx; I can't do anything for 3.5 minutes; GNOME loads. I can do what I want now
Reilithion: gregorian: I'm asking specifically if you've tried to switch to a virtual terminal in those 3.5 minutes.
gregorian: soreau: ERROR: Module drm is in use by ttm,drm_kms_helper when I try to rmmod drm after rmmod radeon
gregorian: Reilithion: I've tried to switch, but seeing how caps lock also won't respond during that instance, it's as good as the keyboard isn't there
Reilithion: How about Magic SysRq? Does that work during those 3.5 minutes?
gregorian: I don't know about "Magic SysRq"
gregorian: But CapsLock should always work right?
Reilithion: Look it up. Every Linux user should know about Magic SysRq.
chithead: gregorian: no
chithead: magic sysrq works in some cases where caps lock does not trigger keyboard led
gregorian: That's cool. Please give me sometime I'm reading about it.
chithead: if only there were some way to find information on the internet :p
gregorian: Which command do you want me to use Alt + SysRq + ? . I don't know what those commands mean, and I don't want to crash my system. :o
chithead: alt+sysrq+r should give you working caps lock
spreeuw: alt+sysrq+b for better
gregorian: Switching the keyboard from raw mode?
Reilithion: gregorian: If you ever need to reboot a non-responsive system, remember: REISUB
chithead: if caps lock works then you can cleanly reboot with ctrl+alt+del
chithead: or force reboot as Reilithion said
gregorian: Ctrl + Alt + Del doesn't respond in that time. But I haven't tried it after Magic SysRq
gregorian: alt + sysrq + b?
chithead: alt+sysrq+r then e then i ... you get the idea
chithead: note that if system does not respond to r, then it will probably not respond to the others either
Reilithion: take back keyboard from Raw mode, send all processes the tErm signal, send all processes the kIll signal, emergency Sync all filesystems, remoUnt filesystems readonly, reBoot
Reilithion: Best done slowly
gregorian: So right now startx, alt sysrq b test for capslock. if it works, I let it continue loading otherwise C A D if it works. If that fails REISUB?
Reilithion: No, you want Alt SysRq R.
Reilithion: B would immediately reboot your system.
gregorian: oh sorry
Reilithion: Which would be bad.
gregorian: otherwise it's correct right?
soreau: and all this only works if magic sysrq is enabled in the kernel.. right?
gregorian: Is it on by default?
Reilithion: So you'll want to test Magic SysRq itself beforehand.
Reilithion: On most systems, it is.
gregorian: I'll check
chithead: /etc/sysctl.conf also has sysrq setting
gregorian: thanks. I'll be back soon
Reilithion: I think Ubuntu had it off in their kernels for a while, didn't they?
gregorian: Well CapsLock works in raw mode
Reilithion1: This suggests your kernel is still chugging away happily. ^_^
chithead: you will probably still be able to ssh in
Reilithion1: Which would allow you to see what's going on while it sits there for 3.5 minutes.
gregorian: Secure Shell login?
gregorian: I don't think I was able to switch terminals
gregorian: Shall I try to switch?
gregorian: I think I did, but don't remember
gregorian: Surely you didn't mean connecting to another computer and trying to login to this one?
Reilithion: Yeah, that's what we meant.
gregorian: I have only one computer :(
Reilithion: Okay, who wants to log into gregorian's computer!
gregorian: If Xorg could insert time stamps....
gregorian: What do I do?
Reilithion: ... cron a script that copies Xorg.0.log to a new file every time it runs, and make that script run as often as cron can make it?
Reilithion: Personally, I would rather see if any of my neighbors had a laptop for the borrowing before descending into magic like that.
gregorian: Well I'm not familiar with scipting :(
gregorian: When I startx, can I by pass the part where I'm automatically switched to the seventh terminal?
gregorian: startx& won't work right?
Ivanovic: is there some problem with the xf86-video-radeon git?
Ivanovic: the webview shows exactly the same status already for 6 days
Ivanovic: (the commit "ATOM: add new power table defs" was 6 days ago)
Ivanovic: (but it is marked as "Age: 15 hours"
gregorian: startx& won't work. Is there a way to stick to the current terminal without going away?
Reilithion: Unfortunately, I have no idea. Did you try switching virtual terminals after Magic SysRq R?
gregorian: I don't remember. I will try it again. Will be back
Reilithion: As for me, I'm still getting the White Screen with compiz. :-(
soreau: Reilithion: Pastebin your X log
soreau: white screen with compiz typically means you are using swrast
Reilithion: It did say something funky about falling back to indirect rendering.
soreau: Look at the output of 'glxinfo|grep rednerer'
soreau: t will likely say you are using software rasterizer
soreau: Your X log should say why
Reilithion: I'ma take a look myself first...
soreau: Just don't post portions of it
gregorian: Reilithion: Can't switch terminal in raw mode
Reilithion: DRIScreenInit failed. Disabling DRI. hmm.
Reilithion: gregorian: aww
Reilithion: too bad
Reilithion: Probably you need to either ssh in, or do the script trick.
gregorian: Is there anyone who knows how to startx without switching to the graphic terminal?
gregorian: I mean stay in the same terminal
nox-: gregorian, wont work id say
nox-: gregorian, you need another console (serial, network, whatever)
gregorian: Reilithion: ssh isn't possible. I wouldn't be able to script either
chithead: gregorian: startx & ( sleep 5; chvt 1 )
Reilithion: soreau: http://pastebin.com/d71eebe0e
gregorian: chithead: startx& failed. You still want me to try that
nox-: chithead, ok thats an idea, if x didnt hang the box...
chithead: if sysrq works, then chances are chvt works too
Reilithion: What is VGA arb?
amarsh04: gregorian, do you get a root prompt back if you type "nohup startx &" without the quotes and hit enter a couple of times?
gregorian: chithead: That didn't work
Reilithion: gregorian: Hey, if you only have 1 computer, how are you talking to us exactly?
gregorian: amarsh04: Why will I get a root prompt if I'm not logged into it?
gregorian: Reilithion: If you didn't notice, there's approximately a four minute lag between the time I try something and come back.
Reilithion: Yeah, I'm just curious, are you using finch or irssi or something?
amarsh04: gregorian, the "nohup startx &" should let the X server start but still leave you in the shell that you were in
Reilithion: terminal based?
gregorian: I'm using irssi
soreau: Reilithion: Well for one thing you don't have kms enabled. Also about the dri error, did you build libdrm first, then X, mesa and ddx against that libdrm?
gregorian: amarsh04: I'm not logged in as root, but I will try it as a normal user
Reilithion: What's ddx? Aside from that, yeah.
Reilithion: soreau: Also, what's kms?
soreau: Reilithion: ddx = xf86-video-ati. kms = kernel modesetting
Reilithion: I actually built libdrm, then mesa, then X, then xf86-video-ati
amarsh04: ok gregorian
Reilithion: For kernel modesetting, how do I check on that? I bet I didn't enable it when I compiled my kernel or something.
soreau: You need to make sure to enable radeon drm in the kernel
Reilithion: Know it's kernel config name off the top of your head?
Reilithion: I'm going to grep my config real fast.
gregorian: amarsh04: It didn't work
gregorian: I forgot to mention this:
soreau: Reilithion: and DRM_RADEON_KMS ?
Reilithion: I see no such key. I grepped for RADEON, and nothing like that came out. Let me try KMS real quick
gregorian: Before I startx, radeon and some other modules are not loaded. After I startx, they are loaded. When I quit GNOME, the modules are still loaded. When I startx again, it takes the same time. Therefore the time taken to load the modules may have nothing to do with the dealy.
Reilithion: soreau: lsmod | grep kms shows that drm_kms_helper is loaded, too
soreau: Reilithion: what about radeon and drm modules?
soreau: Reilithion: Does dmesg|grep drm show anything interesting?
Reilithion: soreau: Yes and yes. dmesg | grep -i drm shows 4 lines. 3 initialized's and "radeon defaulting to userspace modesetting
soreau: Reilithion: I guess you need to boot with radeon.modeset=1 then
Reilithion: So, stick that in my grub.conf?
gregorian: Would someone like to look at my log? I've uploaded it to pastebin.
soreau: There are millions of pastebin sites...
gregorian: I mean I can provide the link if you're interested
Reilithion: soreau: Now really, I doubt there are actually millions.
soreau: I'm not interested actually, but uploading anything to pastebin is useless without providing the link
soreau: Reilithion: Millions.
Reilithion: Hundreds of thousands, at best.
soreau: That's a million, don't you know how to do dramatic math?
Reilithion: ...uh oh
Reilithion: Why is Grub taking forever to load?
gregorian: Now you have my problem? :P
Reilithion: oh dear oh dear
Reilithion: I just upgraded this system
Reilithion: Okay, that was weird. I'm just glad it seemed transient.
gregorian: Did anyone see the log?
Reilithion: Grr. Now DRI is disabled for a different reason.
amarsh04: I had a look at it gregorian... there was a problem I had way back with KDE startup taking forever that was solved by generating some missing font caches by running "fc-cache -v"
Reilithion: Who was having problems with the version mismatch earlier?
Reilithion: I have the same error Kurko had earlier.
gregorian: Thanks amarsh04, but even xterm takes very long to start. And even though the vesa log reports the same errors, it starts in 5 seconds.
gregorian: I'm trying your suggestion
Reilithion: And X just produced a backtrace.
Reilithion: Anybody want this Xorg.0.log?
Reilithion: It's very short.
Reilithion: and maybe I screwed up, myself...
Reilithion: What should my video= kernel option look like? Should I even have one?
gregorian: amarsh04: fc-cache -v didn't work
amarsh04: ok, it was just a thought gregorian
gregorian: I didn't mention this earlier, but somewhere in the middle of the startx, the screen becomes black for a second and then becomes dull gray as usual.
amarsh04: I think that is usual
twnqx: so my laptop crashed again
gregorian: Is this a bug?
twnqx: it happens since i connected a 1920x1200 external display...
gregorian: I shouldn't be this difficult to start X right?
twnqx: welcome to linux.
gregorian: lol, this must be routine for you guys.
gregorian: I'm glad radeonfb's working. At least I can watch movies on the frambebuffer console.
Reilithion: gregorian: what does your video= option to your kernel look like?
Reilithion: It doesn't even mention radeon?
gregorian: I don't even need to mention the resolution probably. Two lines from dmesg:
gregorian: radeonfb: detected LVDS panel size from BIOS: 1280x768
gregorian: radeonfb (0000:01:05.0): ATI Radeon 5955 "YU"
gregorian: If it means anything the log, keeps mentioning unknown vendor ... many times
Reilithion: soreau: Got a new one for ya. http://pastebin.com/d4a9bd44e
gregorian: Is this a bug?
Reilithion: I don't know, gregorian.
Reilithion: Does this happen on a clean install?
gregorian: I used to use Slackware 12.1. Never had a problem. Used Fedora 10. No problem. But when I inserted the Gentoo live DVD, it crashed when it tried to start the graphic manager.
gregorian: I mean KDE
soreau: Reilithion: You still need to rebuild mesa and ddx. xf86-video-ati should say kernel modesetting: yes when it's configuring
soreau: Reilithion: Also, you never mentioned you have two cards on that machine
Reilithion: Oh yeah
Reilithion: I have two cards in this machine.
Reilithion: I want to use the x850.
Reilithion: I'll worry about the dinosaur later.
soreau: anyway, I'm going to rest my neck. Some of the more knowledgeable people should be waking up soon
Reilithion: ok. Thank you for all your help.
gregorian: Thank you all for you help. Even if I couldn't sort this, I learned to compile a kernel and the lovely SysReq
soreau: Reilithion: Meh, also don't use xaa.
soreau: Use exa
Reilithion: Somebody who is not soro, how do I use exa instead of xaa? (because apparently I am using xaa)
amarsh04: Section "Device"
amarsh04: Identifier "Radeon 9200SE"
amarsh04: Driver "radeon"
amarsh04: BusID "PCI:1:0:0"
amarsh04: Option "AccelMethod" "exa"
amarsh04: # default is xaa
Reilithion: Seriously? I have to put it in my xorg.conf? There is no nicer solution?
amarsh04: so you just need the Option line in your radeon Device section and comment out any mention of XAA
amarsh04: unless you recompiled xserver-xorg-ati to default to exa
Reilithion: Okay, fair enough.
Reilithion: How about my little EE problem? Did you look at my pastebin?
amarsh04: not yet, I'll look now
amarsh04: no suggestion unfortunately, I mainly use the Debian Unstable packages
Reilithion: Well, I got another one, too. You any good with backtraces?
amarsh04: not really, my skills are limited to getting backtraces and submitting them as bug reports
fudje: I have a confusing error....
Reilithion: You too, eh?
fudje: the r600_dri module is not loading as though it wasn't built against libdrm_radeon, even though it was.
fudje: buuuuut, this only affects 64-bit. I compiled the dependencies and library for 32-bit emulation and that works fine....
spreeuw: ldd /path/to/r600_dri.so
spreeuw: maybe you lack the proper library at runtime
fudje: and now we's have a lol moment
spreeuw: it must be from git to be sure
spreeuw: and the kernel drm too
fudje: seriously, I'd checked that already
spreeuw: on my machine its linked to libdrm_radeon.so.1
spreeuw: did you build everything in the correct ordeR?
spreeuw: with the correct flags
fudje: but I just solved the problem.
Reilithion: wot was it?
fudje: it's linked against the wrong libdrm.so
fudje: now I just have to work out how I acquired a /usr/local/lib/libdrm.so.2, or why it wasn't linking with it (and causing the problems) earlier....
Reilithion: fudje: Ooo, how do you check what libdrm.so it links against? That might help me...
fudje: it's part of the ldd output
fudje: so for example I did `ldd /usr/lib/dri/r600_dri.so' and it had a line like ` libdrm.so.2 =>/usr/local/libdrm.so.2'
Reilithion: Ok, how do I make sure that the kernel modesetting module is loaded before X starts?
fudje: libdrm.so.2 => /usr/local/lib/libdrm.so.2
fudje: the modesetting is built into the radeon module
fudje: so if you run 'modprobe radeon modeset=1' ...
Reilithion: So, when the output says, "If using kernel modesetting, make sure your module is [dri] loaded prior to starting X, and that this driver was built [dri] with support for KMS.", it means the radeon module?
Reilithion: 'cause as far as I can tell, I'm doing everything being asked of me, and it's still complaining.
Reilithion: What is the kernel module responsible for modesetting actually called? Is it drm_kms_helper?
amarsh04: drm 107705 5 radeon,ttm,drm_kms_helper
Reilithion: And this is the thing that needs to get loaded /before/ X starts, right?
amarsh04: not sure, Reilithion, I'll boot the other machine without X and try it
amarsh04: on the other machine, X is not loaded, but radeon,ttm,drm_kms_helper are all loaded
Reilithion: Perhaps I should try removing gdm from my default runlevel and see if they're loaded on mine too...
Reilithion: none are...
Reilithion: Okay, we're getting somewhere...
Reilithion: Holy crap, yeah it's working.
Reilithion: Okay, now I need to do something about this: The image wiggles
Reilithion: It looks as if the refresh rate might be off or something.
Reilithion: wait, that fixed itself.
Reilithion: Ok, how do I make sure the radeon module is loaded before X starts?
idletask: Reilithion: well, your distribution probably has a mechanism to load modules on boot
idletask: Use it
Reilithion: Oh, ok
Reilithion: I'll ask there
fudje: Reilithion, what distro are you using?
Reilithion: Ah, there it is. modules.autoload
Reilithion: modules.autoload.d is not working. :-(
[Enrico]: Reilithion: gentoo? are you using baselayout 2? couse with bl 2 the file to auto load modules is changed
Reilithion: yes, where did it move?
[Enrico]: Reilithion: /etc/conf.d/modules but be aware it has a different syntax
Reilithion: Ok, thank you very much
[Enrico]: Reilithion: read the migration guide next time ^^
Reilithion: [Enrico]: I did not upgrade -- I only installed Gentoo within the past month.
[Enrico]: Reilithion: you have, the default baselayout installed is version 1
[Enrico]: not version 2
Reilithion: I see. I was not aware.
[Enrico]: you updated stable -> testing
idletask: With nexuiz, when setting one specific option (I don't know which one yet, haven't really tracked it down), there's graphics garbage
idletask: Is that Mesa?
twnqx: yey for another X lockup
uyf: hello, could someone please why i get error 151(GLX) when running glinfo, i have struggled to get dri2 working for a few days now. could it be some remnant of the fglrx driver that's making trouble?
uyf: omg so stupid, forgot to load radeon kernel module :-D
penguin42: ubuntu lucid+xorg-edgers is not a happy bunny on radeon; I get a solid hang starting X with just a cursor at the top left - the Karmic version was fine
soreau: kms or non?
penguin42: I think none; it happens both with the lucid supplied kernel and my 2.6.33-rc2 with kms in
tomislavos: What about problem with Arch Linux + Radeon Mobility 7500 - only EXA seems to work with 2D desktop, (XAA made strange things with cursor sometimes) however 3D games end with drmRadeonCommandBuffer. KMS off with radeon.modeset = 0. Any idea? System updated 2 days ago.
soreau: penguin42: Try booting with radeon.modeset=1
soreau: tomislavos: You want exa with any recent version of X
soreau: tomislavos: What is the output of glxinfo|grep renderer ?
penguin42: soreau: Same - it's a solid kill, stops pinging and if I hit reset it actually powers the machine off
tomislavos: OpenGL renderer string: Mesa DRI R100 (RV200 4C57) 20090101 AGP 4x x86/MMX/SSE2 TCL
soreau: penguin42: I dont know then
soreau: tomislavos: Looks like you have 3D working, not sure why its failing
penguin42: soreau: on my kernel booting with radeon.modeset=0 gets me a VESA low res X server
soreau: tomislavos: Keep it in the channel please
penguin42: (which ain't pretty, but it's better than a solid hang)
soreau: penguin42: Have you tried disabling compiz?
penguin42: soreau: I don't run compiz, this hangs very very early in the X server startup - handful of lines of debug
soreau: Check dmesg?
penguin42: soreau: Hangs too quickly to get a chance
penguin42: soreau: I'm running netconsole and I don't think that's seeing anything
soreau: Should be stored in syslog
penguin42: nah, it's dieing too solidly
penguin42: ok, that's odd; I figured out it was missing the radeon and ati X org packages after upgrade, so using the vesa drivers it was dieing - and that's odd
penguin42: right, that's better - with mode switching on a 184.108.40.206 X.org
edgecase: penguin, i have Thinkpad here, working, certain 3d apps crash but not as bad as you're saying
edgecase: penguin42 that is, it's a T30
penguin42: edgecase: Yeh it seems to be OK now; It's not clear to me how those two packages disappeared; and even less clear as to why it killed the system solid when running the standard X server
cxo: Word of warning if any of you guys are following linux-2.6.git. There is some weird ext4 regression in 33-rc2 that causes frequent random lockups. So avoid that one
penguin42: cxo: Hmm I've not seen anything that would suggest I'm seeing that
cxo: i'm using raid, lvm and ext4, so maybe its more involved than just ext4. but it hangs in a raw spin lock everytime,
penguin42: cxo: I'm running lvm+ext4, no raid though
idle_task: cxo, I use those three features too and haven't encountered any lockup so far
Pallokala: if I compile mesa with gallium, will I be able to switch between gallium and non-gallium without recompiling?
Pallokala: is that an xorg.conf-option?
Dr_Jakob: Then again it is still recommended to use the non-gallium drivers for all radeon hardware.
Pallokala: ok, will not enable
michal_: hello how should I install drivers to radeon X2100 on acer aspire 5720G
michal_: I have only terminal (artefact's on desktop). System is reinstaled (fglrx don't work) I wand run compiz on opensorce drivers, but I don't know how.
michal_: Sax2 -r ISaX: could not import file: var/cache/sax/files/config at /usr/sbin/isax line 199
soreau: michal_: Which distro?
michal_: suse 11.2 gnome
michal_: sorry i forgot about it
soreau: Well first thing you would have to do is completely purge all fglrx from the system. The open drivers wont work with it installed
michal_: i reinstall system
soreau: Hopefully after you do that (and reboot), they will justwork(TM) but if they don't you can come back here and pastebin your X log
michal_: is reboting now
michal_: reboot don't help
soreau: Did you remove/purge all fglrx from the system first?
michal_: i see desktop but i don't see text
michal_: i install system agani
soreau: You mean this is a clean install?
soreau: Then pastebin the output of 'compiz-manager'
michal_: i don't see text
michal_: i write from other computer
soreau: What do you mean you dont see text?
soreau: You're not in an X session?
michal_: x run
soreau: So open a terminal?
michal_: but there are erros in dispaly
soreau: You mean like corruption?
soreau: Well can you ssh into the machine and get the X log from it?
michal_: triangle to up left corner of window
michal_: one moment
michal_: fire i have to run ssh
seb_: hi everyone
michal_: wher i can find turn on ssh
seb_: i try to get KMS working with a RV710 chip and mesa never compil :-(
seb_: i have r600_cmdbuf.c:456: warning: initialization from incompatible pointer type
seb_: make: *** [r600_cmdbuf.o] Erreur 1
soreau: seb_: Did you compile libdrm first?
BioTube: seb_: are you sure you're compiling against the new libdrm?
seb_: git pulled it and configured it with --prefix=/opt/xorg
BioTube: did you set CPPFLAGS?
seb_: not at all BioTube
seb_: hm gonna give another try, thx you guys
soreau: Well if it's in nonstandard prefix, you probably have to set some env vars to tell it where to find libdrm
seb_: yes the xorg wiki taught me that
seb_: just to be sure mesa has to be build in 2nd or in third (after ddx) ?
soreau: I don't think it matters. Both just need to be built after libdrm
seb_: BioTube: i use libdrm 2.4.17 compiled with radeon API, it seems OK ?
BioTube: that's what you need to compile against
soreau: Well.. read the topic though
soreau: newer mesa wont compile against old drm
Ghworg: 2.4.17 is after the api change though isn't it?
Ghworg: Yes, I just checked. The version got bumped to 17 just after
seb_: i have KMS working at boot (its awesome!) issues appear when running 'startx'
seb_: mesa build still fail with [r600_cmdbuf.o] Error 1
seb_: google doesn't come up handy for troubleshooting that
mmichal_90: im back
mmichal_90: i have vnc so waht you wand whith log
mmichal_90: my last nick was michal_
mmichal_90: is hire someone???
soreau: mmichal_90: You can just pastebin the X log somewhere and show us the link
soreau: I find it suspicious you have a garbled screen though you're able to vnc ok
soreau: Almost sounds like bad vram
soreau: mmichal_90: While the machine is booting, do you notice any corruption? like on the bios splash screen or anything?
soreau: michal_: pastebin /var/log/Xorg.0.log file
mmichal_90: i prefare normal desktop :)
seb_: hey guys whats the better way to fix a (EE) RADEON(0): [dri] RADEONDRIGetVersion failed because of a version mismatch. please
knoppers: didnt I have the same error yesterday
knoppers: seb_, what does it say about drm in dmesg?: dmesg | grep drm
adamk_: seb_: Sounds like your radeon Xorg driver wasn't built with KMS support.
airlied: eosie: http://people.freedesktop.org/~airlied/scratch/
airlied: eosie: files with anholt in them
adamk_: seb_: You should build libdrm with with experimental radeon api option and then rebuild xf86-video-ati against that new version of libdrm.
soreau: mmichal_90: michal_: You still have fglrx installed. You need to completely remove it if you want to use the open source driver
soreau: michal_: If you want to use fglrx, you will have to ask in #ati
michal_: i format disc and reinstal suse
seb_: knoppers: plenty og good things: [ 3.743152] [drm] Initialized drm 1.1.0 20060810
seb_: [ 3.771187] [drm] radeon kernel modesetting enabled.
soreau: michal_: that x log shows fglrx is being loaded
michal_: i don't know how
eliasp: does anyone have an idea, how to get more debugging information when my screen stays black after waking up my box from s2ram but it doesn't even react on SysRq... although it seems the disk spins up, the fan starts rotating etc.
soreau: michal_: Try searching your package manager for fglrx
michal_: i instal suse reboot and do vnc this is all
seb_: adamk_: i thought i did that ... ./configure --enable-experimental-radeon before building ....
adamk_: soreau: I'm not convinced that fglrx is installed.
Ghworg: michal_: You might want to try using radeon instead of radeonhd too
seb_: ./configure --prefix=/opt/xorg --enable-radeon-experimental-api --disable-intel
adamk_: soreau: While the built in Xorg configuration does try to load fglrx (probably something stupid suse does), it's not actually loading the driver and the glx module is clearly Xorgs.
adamk_: seb_: But did xf86-video-ati build against that new version of libdrm?
soreau: adamk_: Ok, well you saw he is having garbled/corrupted/artifact/unusable screen?
seb_: how figure that of myself ?
adamk_: soreau: Yeah, but that's not necessarily caused by fglrx :-)
soreau: adamk_: Ok
adamk_: In any case, I agree with Ghworg... michal_: You should use radeon, and you'll need to create an xorg.conf file to do that.
michal_: i don't have rpm with fglrx
adamk_: michal_: Because, at the moment, you are using radeonhd and this is the wrong channel for radeonhd help.
adamk_: seb_: Well when you run ./configure in xf86-video-ati, does it show that kernel modesetting is enabled?
michal_: i don't have xorg.conf
soreau: michal_: Seems I was wrong about fglrx, its just that suse does things differently than some other distros :)
adamk_: Yeah, I came to the opinion, a few years ago, that suse is stupid :-)
michal_: so where i should chnge it
adamk_: michal_: Create an xorg.conf
eliasp: adamk_: ACK, did this decision ~9 years ago ;-)
soreau: adamk_: Poor davidr ;)
michal_: what should by in xorg.conf
adamk_: michal_: Just use 'Xorg -configure' to have Xorg create an xorg.conf. Then edit the configuration file to load 'radeon'
seb_: adamk_: no, kernel modesetting sounds disabled in /xf86-ati-video :-(
cxo: idletask, weird, i couldnt even keep a few minutes uptime. The box locks up as soon as i look away. I reverted back to 2.6.32. Anyhow...
adamk_: seb_: You may need to set the PKG_CONFIG_PATH to /opt/xorg/lib/pkgconfig (or whatever directory the libdrm* pc files were installed to when you installed libdrm).
airlied: eosie: fix gpg key ;-)
michal_: X dont start
michal_: i see only black screan
mmichal_90: Section "Device"
mmichal_90: ### Available Driver options are:-
mmichal_90: ### Values: : integer,
mmichal_90: ### [arg]: arg optional
mmichal_90: #Option "NoAccel" # [
mmichal_90: #Option "SWcursor" # [
mmichal_90: #Option "Dac6Bit" # [
adamk_: mmichal_90: STOP
mmichal_90: #Option "Dac8Bit" # [
mmichal_90: #Option "BusType" # [
mmichal_90: #Option "CPPIOMode" # [
mmichal_90: #Option "CPusecTimeout" #
eliasp: mmichal_90: don't flood!
mmichal_90: #Option "AGPMode" #
mmichal_90: #Option "AGPFastWrite" # [
mmichal_90: #Option "AGPSize" #
eliasp: mmichal_90: use nopaste!
mmichal_90: #Option "GARTSize" #
mmichal_90: #Option "RingSize" #
mmichal_90: #Option "BufferSize" #
mmichal_90: #Option "EnableDepthMoves" # [
mmichal_90: #Option "EnablePageFlip" # [
mmichal_90: #Option "NoBackBuffer" # [
mmichal_90: #Option "DMAForXv" # [
mmichal_90: #Option "FBTexPercent" #
mmichal_90: #Option "DepthBits" #
mmichal_90: #Option "PCIAPERSize" #
mmichal_90: #Option "AccelDFS" # [
mmichal_90: #Option "IgnoreEDID" # [
mmichal_90: #Option "DisplayPriority" # [
mmichal_90: #Option "PanelSize" # [
mmichal_90: #Option "ForceMinDotClock" #
mmichal_90: #Option "ColorTiling" # [
mmichal_90: #Option "VideoKey" #
mmichal_90: #Option "RageTheatreCrystal" #
mmichal_90: #Option "RageTheatreTunerPort" #
mmichal_90: #Option "RageTheatreCompositePort" #
mmichal_90: #Option "RageTheatreSVideoPort" #
mmichal_90: #Option "TunerType" #
_moep_: more output!!!!
mmichal_90: #Option "RageTheatreMicrocPath" #
mmichal_90: #Option "RageTheatreMicrocType" #
mmichal_90: #Option "ScalerWidth" #
mmichal_90: #Option "RenderAccel" # [
adamk_: What makes this particular idiotic is that he knows what pastebin it and used it before.
Ghworg: No anti-flood bot on here then I guess
eliasp: a proper anti-flood bot would have kicked him
soreau: holy cow
adamk_: mmichal_90: All but the very first line is commented out. Which makes it even more idiotic that you would pastebin all that.
adamk_: mmichal_90: Now PASTEBIN your /var/log/Xorg.0.log file.
MostAwesomeDude: we need moar ops in this channel.
mmichal_90: http://pastebin.com/mc4976a0 xorg
adamk_: mmichal_90: By all appearances Xorg started up just fine.
adamk_: What exactly happened when you started X?
michal_: i see black screen
seb_: adamk_: following your advice, mesa built successfully again the right version of libdrm :-)
adamk_: seb_: xf86-video-ati also needs to be built against that version.
adamk_: michal_: How did you start X? By running 'startx' or something else?
michal_: it was my mistake
michal_: startx and X strat screen
adamk_: So it's working now?
michal_: and now how run compiz
jwm: anyone tried firmware loading in 2.6.33?
jwm: my 3200 crashes it
adamk_: michal_: http://en.opensuse.org/Compiz_Fusion And I suggest you take this over to #compiz now.
michal_: for everything
idletask: jwm: I tried it and it works
idletask: jwm: r790
idletask: rv790, sorry
jwm: it identifies mine
jwm: but it crashes heh
seb_: adamk: please can you remind the way to create a default xorg.conf file ? (at my last attempt i messed up the whole display :-)
idletask: seb_: Xorg -configure
idletask: It will create an xorg.conf.new in your current working directory
idletask: I run without an xorg.conf
idletask: Autodetection works great :p
seb_: then just have to tweak a bit the file to load "radeon" and ModulePath "/opt/xorg/lib/xorg/modules,/usr/lib/xorg/modules"
Ghworg: Mine just has a device section with Driver "radeon". For some reason autodetect doesn't work correctly without it
idletask: Ghworg: do you have any other ddx installedN
Ghworg: I haven't tested in a while though, I should see if it has magically started working
Ghworg: idletask: No. I did have radeonhd on at one point but I've removed all traces of it that I know of
seb__: running startx freezes the laptop, Xorg.log remains empty
eliasp: does anyone have an idea, how to debug this: running latest DDX and libdrm, xorg-server-220.127.116.111, mesa-7.7, kernel 2.6.32, KMS disabled, R500/x1600 .... s2ram works, but waking up doesn't ... the disk + fan spins up, LEDs turn on... but the laptop doesn't react at all.... not reachable via SSH, no reaction on SysRq... any ideas how to get further in this situation?
Ghworg: eliasp: Blindly enter a command that creates disk activity like find / and see if disk led flashes
eliasp: Ghworg: hmm, IIRC i've already tried that, but i'll retry...
eliasp: and if this would work, SysRq should work too...
eliasp: so i'm pretty sure it will not work... but let's see ;-)
Ghworg: Probably yeah
Ghworg: eliasp: Next step is boot with init=/bin/bash and see if you can suspend/resume properly with that
eliasp: Ghworg: yeah, wanted to do that too, but couldn't figure out how to trigger s2ram without KDE ... i feel a little dumb, as i'm usually better skilled on CLI than on GUI environments, but this time, it's the other way round ;-)
Ghworg: eliasp: http://www.mjmwired.net/kernel/Documentation/power/basic-pm-debugging.txt
eliasp: Ghworg: great, thx
eliasp: transfers this to the dead-tree-edition and goes for a pm-debugging-spree
seb_: hi again, i have issues with xf86-ati-video build :-(
Ghworg: seb_: That header is part of x11proto-gl-dev package (on debian), so try updating your x11protos package
seb_: Ghworg: thx i only have 1.4.10-1 version on a freshly updated debian
Ghworg: seb_: Hmm, don't know then, that is the same version I have
seb_: i am upgrading xorg-dev to 7.5
Ghworg: just compiled xserver-xorg-video-ati_6.13.0-381-g9d0f3af_i386.deb with no problems
soreau: seb_: Did you set your PKG_CONFIG_PATH like adamk suggested?
seb_: yes i did soreau
soreau: seb_: Are you sure you don't have any old code in of the prefixes that are in your PATH?
seb_: that could, you mean in /opt/xorg ?
soreau: I mean for example, other code that's left in /usr
seb_: yes yes all the /usr/include/GL/* stuff that comes out with x11proto-gl-dev package
cxo: guys... we have 2d accel with r700 right?
cxo: i'm working on some gtk and i've found the treeview to be really slow
cxo: i get graphical corruption too sometimes if i scroll too fast
cxo: here is a sample program, http://en.pastebin.ca/1728846
cxo: (you only need the gtk+2.0-dev packages)
idletask: cxo: yes, there is 2D accel
seb_: Ghworg: how did you build your xserver-xorg-video-ati_6.13.0-381-g9d0f3af_i386.deb ? please
Ghworg: 2d accel is faster than fglrx even based on the benchmarks I've seen. But some gtk widgets appear to be incredibly slow for me too
Ghworg: seb_: I have a build script that makes a tarball of the git dir, adds a custom debian dir (based on the one from the last released package but updated by me) then it compiles and makes a package
idletask: cxo: this program runs smoothly here
cxo: Are you using compiz? KMS and friends?
cxo: when i scroll, the view lags behind almost a second
idletask: Not compiz but kwin, and I run KMS
cxo: kwin has compositing?
idletask: Yes it has
adamk_: cxo: Are you sure *you* have 2D acceleration?
cxo: I dont know really. Is there a way to query the driver?
Ghworg: seb_: This is my build script http://git.trollgod.org.uk/?p=misc.git;a=blob;f=bash/build_radeon.sh;hb=HEAD
adamk_: cxo: /var/log/Xorg.0.log should tell you if acceleration is enabled.
er0x: hey, kms works fine on .32 but on 32.2 with same config i get no consoles after radeon kicks in, drm/radeon/fbcon are build in, any ideas?
idletask: "Sorry, this demo requires GL_ARB_fragment_shader" <-- this is what I get when I try to run the arbfslight demo from Mesa
idletask: Missing DRI stubs?
cxo: well here is my Xorg.log http://en.pastebin.ca/1728869
adamk_: cxo: Well you definitely have acceleration.
idletask: cxo: why both ati and radeon loaded?
eliasp: Ghworg: thx for your help, now i can s2ram successfully using "s2ram --acpi_sleep 1 --force" ... do you know what's the equivalent setting in /proc to always use "1 aka s3_bios"?
cxo: idletask, isnt "ati" just a wrapper that calls "radeon"?
adamk_: Yes, it is.
idletask: Ah, OK
idletask: I thought this was refering to fglrx, my bad
cxo: well that sample program is really sluggish on my box, 2.6ghz x3, hd4870,
seb_: usually how do you get rid of old code ? (specially to succeed with xf86-ati-video build) thx
Ghworg: eliasp: "/proc/sys/kernel/acpi_video_flags, with "1" for s3_bios, "2" for s3_mode and "3" for both", from http://en.opensuse.org/Suspend_to_RAM
eliasp: Ghworg: found it... using the cmdline param acpi_sleep=s3 should work...
eliasp: Ghworg: ah, nice, thx
eliasp: although i don't like SuSE at all, the quality of the wiki increases even more these days...
adamk_: cxo: It's definitely slower here on an r7xx than on an intel i915 GPU.
Ghworg: The s2ram devs work for suse, so that is the place to go for suspend info :-)
eliasp: ah, ok.... ;-)
eliasp: another question: does anyone know, when powersaving support for ati/radeon on KMS is to be expected?
cxo: adamk, wasnt it airlied who worked the 2d accel code? maybe i should file a bug report?
eliasp: at the moment my Fan runs always on full speed when using KMS
stikonas: eliasp: there are initial patches, so framework is here
idletask: The clearspd demo from mesa makes the display crawl to a halt
eliasp: stikonas: oh, great.... in which git repo? so i'd try to apply them to my 2.6.32
adamk_: cxo: No idea who worked on it.
Ghworg: eliasp: When it works correctly. Currently it causes glitches on every clock change, until that is sorted it won't go in
eliasp: ah, ok... so it is actually worked on..
stikonas: eliasp: in dri-devel mailing list, they are not it git repo
eliasp: that#s good news at least
eliasp: stikonas: ok, thx
stikonas: strange, I've applied that patch and there are no glitches
seb_: do some users experiment errors with /usr/include/GL/glxint.h when building xf86-ati-video ?
idletask: More 3D demos from mesa have a little hiccup every second or so, where can that come from?
Ghworg: stikonas: I'm just reporting what Zajec has been saying on the dri-devel list, I haven't tried myself.
stikonas: it may be that only some cards have glitches
eliasp: is going to try that... let's see
Ghworg: has the patch in his downloads dir but hasn't got around to applying and testing yet
idletask: "Sorry, this demo requires GL_NV_fragment_program" (fplight demo)
er0x: any1 got working kms on 32.2?
idletask: Mesa's demo choppiness disappears if I do not run KMS
idletask: er0x: which chipset?
cxo: adamk_, i have a feeling its partly GTKs fault. The same program in QT is liquid.
er0x: idletask: rv635 aka mobility hd3650
idletask: er0x: you need extra firmware for that
er0x: idletask: 32.0 works fine
idletask: er0x: yes, I know :/
er0x: idletask: 33-rc2 was complaining about r600_rlc.bin, ive downloaded it from agd5f's space but it still failed to boot stops at loading r600_rlc.bin
er0x: 32.2 got same prob?
Ghworg: idletask: Could be IRQ related. Someone else reported jerkiness due to that, IRQs are only active with KMS
idletask: Ghworg: I was pondering about that, indeed
Ghworg: er0x: 32.2 doesn't need that firmware because the IRQ code isn't in it
er0x: so what im missing for 32.2?
er0x: would be cool to get kms runing on 32.2
idletask: gltestperf makes 2.6.33-rc2 hang
idletask: (no KMS)
er0x: looks like this related to my probs
idletask: er0x: well, in fact, kms also makes 18.104.22.168 hang here
idletask: No KMS and it runs
idletask: 22.214.171.124 didn't have this problem
er0x: w/o kms it runs here too
idletask: Has the problem been bisected?
er0x: and 32.0 is just fine with kms didint tryed .32.1
idletask: 32.1 works
idletask: OK, mesa's glperf test causes drm to reset the GPU
idletask: As a result, X is killed
idletask: But there's no getting back to the console :(
idletask: I had to restart X from another machine
idletask: I have a lot of messages like this one: [drm] wait idle failed status : 0xE77344A4 0x00FF0F02 (the same message each time)
idletask: Right after, it says "Resetting GPU" and X is dead
er0x: now kms works on 32.2 :)
idle_task: er0x: how did you do that?
er0x: idle_task: used airlied's patch http://marc.info/?l=dri-devel&m=126137027403059&w=2 hope he moves it to 32.3
mentor_: Hmmm, I trying to pull drm-next onto 32.2, but I'm getting lots of conflicts
mentor_: Is that the branch I shoudl use, or shuold I go for drm-linus?
Ghworg: mentor_: or drm-radeon-testing if you want bleeding edge
idle_task: "Sorry, this program requires GL_APPLE_vertex_array_object"
MostAwesomeDude: APPLE_vbo should work.
idle_task: tries drm-radeon-testing without KMS, to see
cxo: I think i found the magic line that radeon doesnt like
cxo: "gtk_tree_view_set_grid_lines" take that line out and the app speeds up like crazy
cxo: like at least 100x speed increase on my box
edgecase: [107957.868933] [drm:radeon_cs_ioctl] *ERROR* Failed to parse relocation ! nexuiz shows splash screen, then crashes
SnowRaptor: Hey folks! X segfaults when I try to boot compiz. Syslog says it's related to libjpeg.so.7, but it happened after I booted a recently pulled kernel from GIT and patched it with drm
SnowRaptor: syslog also contain complains about "integrated sync not supported" and "invalid framebuffer id"
SnowRaptor: I'm running it on an xpress 115 (rs480, I guess).
SnowRaptor: Hmm, forgot to recompile libdrm/ddx/mesa
mentor: "AIGLX error: Calling driver entry point failed"
gregorian: Hi, my GNOME takes 3.5 minutes to start with the radeon driver and I was asked yesterday to find the bottlenecks by writing a script. I've written a script that 'tails' xorg log every 1 second. I think I found the bottlenecks. Here's the link:
gregorian: I'll be back later when there are more people
Demo1: hello, i have problem
Demo1: cant install radeon driver
Demo1: any help?
tstellar: Demo1: Can you be a little more specific?
Demo1: cant install
Demo1: im trying
Demo1: dont know
Demo1: what to try now
Demo1: trying install radeon driver on gentoo
tstellar: Are you trying to install mesa?
tstellar: Or the kernel driver, or the X dirver?
tstellar: This channel is for the open-source radeon driver, try #ati, if you have questions about fglrx.
Demo1: ok, then how to install open soruce driver? :D
tstellar: First, you need to enable the radeon driver in your kernel.
Demo1: in make.conf?
tstellar: In make.conf, you need to add this line: VIDEO_CARDS="radeon"
tstellar: Make sure you have CONFIG_DRM_RADEON enabled in your kernel config.
Demo1: where waas the kernel config?
tstellar: Did you build the kernel yourself, or did you use genkernel?
tstellar: Ok, then genkernel probably did that step for you.
Demo1: ok, next?
tstellar: Now you need to: emerge libdrm mesa xorg-server xf86-video-ati
tstellar: What graphics card do you have?
Demo1: can you write next steps?
tstellar: emerge libdrm mesa xorg-server xf86-video-ati
Demo1: im doing it now
tstellar: Demo1: btw, I don't think your card is supported by fglrx anymore.
Demo1: than the open source is right for me
Demo1: who cares?
amarks: hmm, video tearing even with irqs here
Demo1: tstellar what next?
cxo: amarks, when i switch into fullscreen it goes away
amarks: cxo: i thought tearing should be vanished with irqs?
tstellar: Demo1: You should be done now. Try running glxinfo and pastebin the output.
Demo1: glxinfo: command not found
cxo: amarks, i didnt know that. I thought the tearing was a sync issue with composited desktops and some gl function that wasnt implemented or something... but i could be wrong
tstellar: Demo1: emerge mesa-progs
amarks: cxo: it would be nice to get some clarity on this issue
Demo1: error http://dpaste.com/138430/
amarks: cxo: r u on kde or other?
tstellar: Demo1: Can you post the build log.
cxo: amarks, gnome, ubuntu-9.10, linux,mesa,drm,xf86-video-ati out of git
amarks: cxo: k, so both kde and gnome are tearing
hifi: btw. irq's did not enable vsync
amarks: hifi: what does?
cxo: amarks, with xine, i just get a line of distortion at the bottom, but the video is smooth, fullscreen is perfect
hifi: as I was told before, DRI2 does not have vsync support (yet)
hifi: and tearing was reduced with some kind of a hack
amarks: tearing is rather visible here
cxo: if you blink fast enough it will go away
amarks: i've seen no improvement for ... years lol
Demo1: tstellar so?
tstellar: Demo1: The build log you posted has some of the lines truncated, so it is hard to read. Can you try again.
amarks: is adding vsync support a challenge or something?
cxo: what card do you have?
cxo: I have no tearing with r700
cxo: OpenGL renderer string: Mesa DRI R600 (RV770 9440) 20090101 TCL DRI2
cxo: OpenGL version string: 2.0 Mesa 7.8-devel
amarks: i'm on drm-radeon-testing
amarks: and git mesa/ddx/libdrm
cxo: me too
amarks: kde is set to sync to vblank
cxo: 2.6.32 + drm-radeon-testing
amarks: yes same
cxo: i didnt mess with any of the settings, i just have compiz going
amarks: i ain't using compiz
cxo: maybe thats why it doesnt tear because everything is going through gl
tstellar: Demo1: I think you need to select the X11 OpenGL implementation. Try: eselect opengl list
amarks: cxo: xv tears, i'll try gl
tstellar: Demo1: There should be a blue asterix next to xorg-x11.
cxo: i have force_glx_sync enabled on compiz apparently
Demo1: the blue is next to ati
tstellar: Demo1: Change it to xorg-x11.
tstellar: Demo1: eselect opengl set xorg-x11
tstellar: Demo1: Then eselect opengl list to verify that it worked.
Demo1: again emerge mesa-progs
tstellar: Demo1: You need to remove fglrx, it can't coexist with the open-source drivers: emerge --unmerge ati-drivers
amarks: cxo: nice i tried mplayer gl and had a null ptr kernel crash
cxo: amarks, mplayer has been doing that for me too for the last couple of weeks, i've given up, using xine now
amarks: cxo: this is all still very unreliable *sigh*
cxo: amarks, xine is actually a better player. the code base is a lot more mature. its just that xlib gui that turns people off and makes it look antiquated
amarks: cxo: mplayer is just a habit, very hard to break
tstellar: Demo1: You should emerge mesa-progs, and then I think you need to restart X, because you changed the OpenGL implementation.
cxo: amarks, I always say that if you want to know how stable a piece of code is, look at its Makefile :)
tstellar: Demo1: Ok now try posting the ouput of glxinfo.
amarks: cxo: and what annoys me is that when we crash, i have _never_ seen anything on the screen, ever ever, i've got a laptop permanently waiting patiently for netconsole output. if the kernel is going to panic fine but take me back to a text console so i can see it
tstellar: Demo1: Right now you are only using software driver, what happens when you run: lsmod | grep -i radeon
tstellar: Demo1: can you post the output of xorg.log
amarks: cxo: do you have a bug open regarding mplayer gl causing null ptr crash
Demo1: where is that file?
tstellar: Demo1: /var/log/Xorg.log
cxo: amarks, nah, i couldnt be bothered. They have really ugly Makefiles. I've somewhat got cozy with xine
Demo1: i just have Xorg.0.log
tstellar: Demo1: That is fine.
Demo1: ok i will seperate it
Demo1: too long
amarks: OT: the latest Top Gear episode was good
tstellar: Demo1: Try: modprobe radeon
Demo1: FATAL: Module radeon not found.
tstellar: Demo1: So, you haven't built the kernel module. Did you remember what arguments you passed to genkernel when you built the kernel?
Demo1: check http://wiki.cchtml.com/index.php/Gentoo_Installation_Guide
Demo1: where is menuconfig
Demo1: i recompiled kernel
Demo1: Bad i idea
Demo1: i can recompile again
tstellar: Demo1: Go to /usr/src/linux
tstellar: make menuconfig
tstellar: Demo1: nevermind, quit this and do: genkernel --menuconfig all
Demo1: ok and something else
Demo1: i had some error with udev
Demo1: cant i fix it
Demo1: too, before compiling kernel?
tstellar: Yes, go to General Setup and uncheck enable deprecated sysfs...
tstellar: Demo1: Then to enable the radeon driver, go to Device Drivers->Graphics Support->Direct Rendering Manager
tstellar: Demo1: Then select ATI Radeon.