ptr20r2: hello all
dougmencken: Faulting instruction address: 0xf2445ed0
dougmencken: Oops: Kernel access of bad area, sig: 11 [#1]
dougmencken: http://fpaste.org/vyNw/
ptr20r2: am i allowed to ask about fglrx drivers here
hnsr: fglrx is never really discussed here, might have more like in #ati if you haven't tried asking there already
hnsr: s/like/luck/
ptr20r2: do i need to have the chrome os source in order to install the fglrx module?
ptr20r2: dw
ptr20r2: leaving now
dougmencken: what about my error? dmesg: http://fpaste.org/vyNw/
mjt: ok, here's something... else.
mjt: i've an.. interesting issue here. At bootup, right after radeon module gets loaded, the screen ("text" console) gets filled with "0" chars, as if the "0" key on the keyboard is stuck.
mjt: i can hit "backspace" to compensate, but once i release it the zeros continue
mjt: in X such thing is not happening
mjt: and when exiting X, zeros on the console stops too
mjt: in 2.6.32 kernel such problem didn't exist. But it happens _almost_ at every boot with 2.6.33
mjt: anyone else seen this?
KotH: hmm.. sounds like a broken kbd driver
KotH: or something writing into the mem range of the kbd driver
mjt: i wonder how it's so strongly relates to radeon :)
mjt: maybe it's not radeon but a bunch of related modules/susystems, i dunno
mjt: subsystems even
mjt: when i don't load radeon at bootup the zeros does not come
mjt: but they start coming right after `modprobe radeon'
jcristau: you switch from vgacon to fbcon when radeon loads
mjt: yeah
mjt: hence i said i'm not sure it's radeon or something else :)
mjt: i can unplug all input devices after this (all keyboards etc), but zeros are still coming
mjt: it does not happen after suspend/resume
mjt: but it DOES happen when I first load X and next switch to console (without exiting X)
mjt: only exiting an X session fixes this
mjt: heh. Hitting "Ctrl" stops the zeros. Releasing Ctrl continues them. As if it "thinks" I pushed Ctrl+0 ;)
mjt: ok, my console is completely unusable now. Rebooted twice, exited/restarted X numerous times, even suspended/resumed - the zero's are still coming...
mjt: ok, mystery solved
mjt: it was my tv card with a remote control
mjt: its receiver, or some other part of it, apparently behaves badly under some conditions
mjt: and loading radeon triggers these conditions
mjt: i disabled loading the driver for it (which was quite difficult to do), and now there's no more zeroz coming (and no more tv, either)
mjt: (i never ever used the remote controller of this card ;)
evil_core: mjt: I had such problem with mouse emulation for imon-lirc, I was backporting old mouse code to new lirc driver and event subsystem, basing on ati wonder
evil_core: mjt: and one of their devs forgotten mine patch, and wrote it themselves with getting similar prolem for mine hw(but not his)
evil_core: mjt: I guess you got all in wonder, brooktree or imon controller, but maybe theres some new hw
evil_core: and one of them was ATI all-in-wonder, the second some microsoft remote protocol(for their windows media-center)
mjt: it's a separate add-on card, nothing to do with radeon
mjt: ultimedia controller: Philips Semiconductors SAA7134/SAA7135HL Video Broadcast Decoder (rev 01)
evil_core: never heard about analog TV philips problems, only brooktree some, but devs dont like stagnation and probably converted it to new subsystem to bring new features and problems for users ;)
evil_core: anyway good choice, I dont like brooktree(conexant) blured and stretched picture, but peoples like that because of 'lack of black stripes' and lack of pixels
evil_core: and I like scaling in radeon driver, its much more better than nvidia, which really sucks bluring everything
mjt: it has nothing to do with philips: it's an additional chip on the card
mjt: this tuner is here for about 5 years already
mjt: i were searching a cause to replace it with somthing more adequate :)
evil_core: scaling sint built-in SAA... chips?
mjt: no idea
mjt: it worked after some tweaking (i had to add support for this particular device to the saa7134 driver at that time, and for the remote too)
evil_core: I liked stereo philips tuners, and definitely every SAA I saw wasnt bluring and conexant was, there were even sites about that
mjt: but the pic quality is rather poor, and especially become poor after local tv cables where replaced
GoGi: can gallium on R600 already be tested?
evil_core: hmm...didnt I talked with you in lirc channel in the past?
xming: it's in glisse's git
mjt: evil_core: sure we talked before, but i don't remember where :)
mjt: evil_core: maybe in #debian
evil_core: no, I am using PLD
mjt: .j v4l
mjt: er
evil_core: mjt: werent you adding mouse suppot to imon devices also?
mjt: no
evil_core: hmm..maybe I should drink less ;)
TCW: nown that more or less recent X + xorg-video-radeon when starting compiz it crashes X?
TCW: known even
evil_core: btw, can anybody review mine r500 pageflip patches, and tell whats causing hardlockup?
evil_core: http://carme.pld-linux.org/~evil/radeon/r5xx_pageflip/
mjt: TCW: the whole thing has many bugs still, and the more you throw at it, the more bugs gets triggered.
mjt: TCW: not to say it is _expected_ to crash or that it crashes for everyone else.
mjt: TCW: It works for me for example -- xserver-xorg-video-radeon 6.12.191-1 (debian experimental), 2.6.33 kernel
TCW: mjt, debian sid current on rv530 here, no experimental packages. It did work the last months rather smooth... but since the most recent upgrade (I think)... at least Xv still works. But as soon as I try to start compiz it dropsme back to gdm login.
TCW: mjt, in other words, it had less /tragic) bugs then now... at least from a users pov
mjt: i tried compiz yesterday for the first time and it worked
mjt: that's basically why i updated to experimental
TCW: mjt, what chip?
mjt: um, not rv530 :)
mjt: rv600 i think -- built-in amd780g
TCW: mjt, you did upgrade packges (which btw.?) to experimental to get compiz?
mjt: that too, or else stuff crashed left and right
TCW: mjt, lspci |grep VGA
mjt: that won't tell the chip
mjt: it's Radeon HD 3200
TCW: her it does (on all radeon chips I use)
TCW: here even
TCW: 04:00.0 VGA compatible controller: ATI Technologies Inc RV530 [Radeon X1600]
TCW: mjt, anyway, what packages did you upgrade? How often do I have to ask? :p
TCW: +to experimental
TCW: and I rather _don't_ expect that rather old chips like rv5xx are still (very) buggy... as said it did work ok almost a year.
TCW: I know regressions happen,, I know the older chips may not get so much testing as newer ones, but I don't _expect_ it to happen :)
zhasha: airlied: ping
TCW: mjt, btw. xserver-xorg-video-radeon 6.12.191-1 is older then the package in sid. Sid has 6.13.0-1
TCW: xserver-xorg-video-radeon_6.12.191-1_i386.deb
TCW: Seen in debian on 2010-03-03 21:51:13 in /pool/main/x/xserver-xorg-video-ati.
TCW: Size: 614406
mjt: well, at least mine works.
mjt: but it's different chip
TCW: mjt, another one (on another channel) has the same issue as me and also a different chip
dougmencken: ATi Rage 128 questions goes here?
dougmencken: (ATi Rage Orion for Mac OS with OF PEF code in it exactly)
dougmencken: startx stderr: http://pastebin.com/8YDhUfZw
dougmencken: strace Xorg : http://pastebin.com/RDdxxby0
dougmencken: especially this line: "write(0, "(II) R128(0): Bad V_BIOS checksu"..., 34) = 34"
Xaratas: ok, next chan, next chance ^^ i have this http://www.x.org/wiki/radeonBuildHowTo all done and a kernel mode setting enabled environment build. The xorg log http://kde.privatepaste.com/6fd3590c7c also looks good, but glxinfo says, that swrast is used
tormod: Xaratas, use LIBGL_DEBUG=1 glxinfo
Xaratas: interesting
Xaratas: i have hardware acceleration
Xaratas: but why?
Xaratas: aha
Xaratas: the xterm has as root hardware acceleration, nut not as user
Xaratas: how strange is this?!
Xaratas: libGL error: failed to open drm device: Permission denied
jcristau: wrong perms on /dev/dri/card*?
Xaratas: which should there be?
Xaratas: i have messed with this place cause the os make some thing wrong
jcristau: iirc it should have an acl to give perms to the user of the active session
Xaratas: have hat a file /dev/dri not a folder
Xaratas: *had
Xaratas: ok, as root i get 2 additional errors
Xaratas: libGL: Can't open configuration file /etc/drirc: No such file or directory.
Xaratas: libGL: Can't open configuration file /root/.drirc: No such file or directory.
jcristau: doesn't matter.
Xaratas: ok, will Section "DRI" Mode 0666 EndSection in xorg.conf solve this?
BioTube: Xaratas: only if you put yourself in the video group
otaylor: Wonder if GLX_INTEL_swap_event has ever been tested with the radeon drivers
jcristau: i think not.
jcristau: Xaratas: modern libdrm lets udev create the device node and set its permissions
Xaratas: ok, and what permissions shold be set on /dev/dri/card0 ?
jcristau: i already answered that
Xaratas: a, there could be the problem, udev maby does not know how to set it up correctly
marienz: I don't know what mechanism set it, but here it's 0660 root:video with an additional acl rule giving my current user access
evil_core: BioTube: wrong! 666 means accvess for all, Group video+0660 would mean access only for video members
marienz: if that's not the case for you I'd prefer figuring out why the acl rule is missing instead of adding myself to the video group (and I definitely wouldn't make it world-accessible)
BioTube: evil_core: thought it said 0660
evil_core: *it*
Xaratas: ok, is an acl rule a file? something a "find" could look for? maybe it got not packed in my install package
BioTube: evil_core: ?
Xaratas: but if it is on disk i could place it where it belongs
marienz: Xaratas: getfacl /dev/dri/card0
marienz: Xaratas: /dev/ is normally on a tmpfs these days (managed through udev), manually messing with it is not normally a good idea
Xaratas: getfacl: Removing leading '/' from absolute path names # file: dev/dri/card0 # owner: root # group: video user::rw- group::rw- other::---
jcristau: /lib/udev/rules.d/70-acl.rules:SUBSYSTEM=="drm", KERNEL=="card*", ENV{ACL_MANAGE}="1"
marienz: ah, it helps if you grep for drm instead of dri
jcristau: and then udev runs udev-acl on devices with ACL_MANAGE
Xaratas: i have no /lib/udev/rules.d folder
marienz: Xaratas: it may be /etc/udev/rules.d/ on your distro
Xaratas: yes
marienz: and udev-acl is also run off (and presumably interacts with) consolekit
marienz: does ck-list-sessions correctly show you as logged in? which distro is this anyway?
jcristau: yeah
Xaratas: i use slitaz
marienz: if it doesn't run consolekit adding yourself to the video group may actually be the preferred solution
Xaratas: ok, i added the udev rule, i think a reboot could show if it works
marienz: err
Xaratas: ok, the udev rule changed nothing
Ronis_BR: can anyone explain me why xf86-video-radeon is so slow with SDL applications? I also cannot use DGA
marienz: yes, randomly adding udev rules by hand isn't going to do much
marienz: Xaratas: does that distro actually use consolekit?
ptr20r2: test
ptr20r2: ah ha hell oall
marienz: hmm, there went compiz. I'm assuming "[drm:radeon_cs_ioctl] *ERROR* Failed to parse relocation -12!" is too generic to be useful?
marienz: and there went the entire system again: netconsole gets "[drm:radeon_fence_wait] *ERROR* fence(d60d7940:0x000DF682) 504ms timeout going to reset GPU" followed by "[drm:r300_ga_reset] *ERROR* VAP & CP still busy (RBBM_STATUS=0x8400C100)" followed by a locked system. r300 (9600xt), lucid's X and kernel (which is 2.6.32 with a backport of .33's drm and 6.13.0 driver)
evil_core: r300/r200 is quicker than r500 with free driver?
glisse: marienz: gpu lockup ...
marienz: I'm also getting the occasional trace of corruption (usually in urxvt, with it going away if I force a redraw somehow)
glisse: marienz: agp ?
marienz: yes :(
glisse: which bridge ?
marienz: nforce2, lemme just pastebin my entire dmesg
marienz: http://pastie.org/925698
marienz: I can't reproduce this, but it happens frequently enough to be annoying
marienz: that is: I can't reproduce it on demand
glisse: marienz: try adding : ".needs_scratch_page = true," at line 312 of drivers/char/agp/nvidia-agp.c
marienz: and it's annoying because I haven't gotten the system to respond to anything other than the reset button when it does happen (those two netconsole messages are the last sign of life)
glisse: and see if it works better
marienz: that'll be fun, since this is a stock ubuntu kernel :)
glisse: oh thought you did build it your self
marienz: my bad, I'm in ubuntu where I use a stock kernel. I guess I could switch over to gentoo for a while and see if it happens there too
glisse: will more than likely happen there too
marienz: it's odd though, I think this only started happening recently
glisse: i will redo agp patch tomorrow maybe some ubuntu folks can pull them for testing
marienz: I can build kernels, it's just that I'd expect a rebuild of a kernel using ubuntu's normal config to take a few hours
jcristau: make localmodconfig is not too bad
marienz: ah, that's a thing?
marienz: that'd help
marienz: I shall look that up, thanks :)
jcristau: takes your current config, and enables just the modules that you have loaded
jcristau: instead of everything
marienz: ideally I'd patch and rebuild just that one module, but I'm not sure if the build system lets me do that conveniently
marienz: hmm, this probably won't actually be that painful
Xaratas: marienz: re and no, slitaz has no "consolekit"
marienz: Xaratas: then using the video group is probably sane, and I'd be surprised if messing with udev was desirable
marienz: that rule you added doesn't do anything interesting on its own, it tells the consolekit integration to act on the dri device, afaik
marienz: no consolekit -> no consolekit integration -> adding that rule is a noop
Xaratas: i had once a dvb-t usb stick which has no propper udev rule, so i have to add this per hand
Xaratas: it worked on the xandros eeepc
Xaratas: restart x
Xaratas: works :)
ROKO__: vyrgroznik
ROKO__: ko staa wa
ROKO__: :D
dougmencken: anybody here? to listen to my problems I posted 12 hours ago and got no response at all?
dougmencken: (or maybe it's some kind of ignoring)
bridgman: usually just means there was nobody around, or that the people who were there were in a specific discussion already
bridgman: what's your issue ?
dougmencken: wait, I'm reading logs
dougmencken: Faulting instruction address: 0xf2445ed0 \n Oops: Kernel access of bad area, sig: 11 [#1] \n http://fpaste.org/vyNw/
dougmencken: now I'm on 1024x768 (much smaller than display's native res)
dougmencken: without any chance to change it
bridgman: aw crap, I was hoping for something like "how do I get 3D ?" ;)
bridgman: looking at the log now
dougmencken: bridgman, mine is "how do I get 2D?"
bridgman: yeah...
bridgman: looks like OpenFirmware is setting up a framebuffer device before drm comes along and tries to do the same
bridgman: wonder if they are interacting badly ? I don't remember seeing the OF fb messages before but then again I haven't looked at many powermacs running linux
dougmencken: yes, OF *must* have framebuffer, it's the reason for us to look at yaboot and kernel booting
bridgman: the drm init is clearly having problems (cp test failing etc.) before the segfault
dougmencken: it's with radeon.modeset=1 video=ofonly
dougmencken: (as kernel boot args)
dougmencken: there's no text mode or vesa for us
dougmencken: is it so hard to get benh to speech?
bridgman: when he's sleeping or away from the computer, yep
bridgman: ;)
dougmencken: but it's not the only one problem; the other card is ATi Rage Orion Mac Edition; bug report - http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=578269
bridgman: problem is that I don't know how openfirmware and drm are supposed to interact since they are both trying to own the GPU
bridgman: looking at bug report
dougmencken: I have 3 videocards; none of them works well
dougmencken: (yep, all of them are mac-edition-special-non-bios)
bridgman: did they ever work, ie are we looking at regressions, or is this a case of trying to get that hw working for the first time with these drivers ?
glisse: it should work
glisse: i need to dust of my mac
dougmencken: never worked; always something bad; like blank black screen etc.
dougmencken: (I tried since 2005)
bridgman: I guess the magic words are "video=ofonly", presumably that tells drm to use open firmware calls or something ?
dougmencken: bridgman, airlied told me to avoid using of radeonfb drived there's magic of "video=ofonly"
dougmencken: drivers*
bridgman: sounds right; don't think radeonfb and kms drm get along
bridgman: they both want to own the hardware IIRC
dougmencken: can it be "normal"; i.e w/o any kernel args and such?
glisse: yeah avoid radeonfb at all cost
glisse: with kms
dougmencken: just plaing recognizing
dougmencken: default kernel booting chooses radeonfb
dougmencken: so *why*?
dougmencken: >> how openfirmware and drm are supposed to interact <-- OF is handled by OS when OF execution is passed to OS; so OF is "helper", not foe
dougmencken: but x86-bios-based-mind thinks "it's all due to parallel universe's wrong formulas" ;)
glisse: dougmencken: kms replace radeonfb
glisse: if both are loaded bad thing happen
dougmencken: ? type it as root?
glisse: because they will step on each other
dougmencken: but, as an advocate: radeonfb works w/o blueish-yellowish artifacts and w/o SIGBUS errors due to "bad checksum in V_BIOS"
glisse: i am not saying radeon kms works on ppc
glisse: i am saying it should work
dougmencken: cool, it doesn't
glisse: and i know people who successfully run it
dougmencken: on which cards?
glisse: mostly powerbook laptop so r3xx
dougmencken: ah, notebooks
dougmencken: on my side kms doesn't works producing only errors
dougmencken: (and unchangable 1024x768)
dougmencken: any solution then?
glisse: well open a bug
glisse: i guess it will wait until someone has time to look into it
dougmencken: is http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=578269 enough?
dougmencken: is opengl rendering like http://img706.imageshack.us/img706/4884/hedgewarsshot.png enough?
dougmencken: I worked with airlied, he said that there's smth wrong with sysfs on my side
glisse: dougmencken: rage 128 is not supported by kms
dougmencken: yeas, but my current card is (it is original 7000 Mac-OpenFirmware edition)
glisse: then open a bug for it on bugs.kernel.org
glisse: if kms doesn't work with it
dougmencken: okay, time to
glisse: btw what is the game in the screen shot ?
dougmencken: hedgewars
dougmencken: #hegdewars
glisse: has likely found another addictive game after bubble :)
dougmencken: IceCat can't find the server at bugs.kernel.org.
glisse: bugzilla.kernel.org then
dougmencken: Resolving bugs.kernel.org... failed: Name or service not known.
dougmencken: wget: unable to resolve host address “bugs.kernel.org”
dougmencken: so you kidding?
dougmencken: yes, you do
glisse: each project has different url for bugzilla i can remember all the time
glisse: https://bugzilla.kernel.org/
glisse: works here
dougmencken: ah, okay; but none of 3 videocards (all of them are ATi's) don't work << why?
jcristau: because the developers hate you
dougmencken: jcristau, fine
glisse: give asic of the 3 card please
dougmencken: bash: asic: command not found... current one: "0001:10:15.0 VGA compatible controller: ATI Technologies Inc Radeon RV100 QY [Radeon 7000/VE]"; wait a minute for others (need to switch monitor)
glisse: asic->model like radeon 7000
glisse: or radeon 9200
glisse: ...
dougmencken: it's original 7000 rv for mac with DriverLib for OS <= 9.2.2
bridgman: IIRC the second was a rage 128
dougmencken: yeah, trying to boot it now
dougmencken: rage orion
dougmencken: the 3rd is rage 128 pro agp mac edition
glisse: kms doesn't support rage
glisse: so for them stick to what ever was working previously
dougmencken: nothing ever worked for rages
dougmencken: I'm about 7000 now
marienz: ok, I really need to go try that agp module change (system just locked up again)
marienz: and rumor on the internet has it I can just compile and install the one module
marienz: well, that worked (added a print which was triggered), now I'll just have to wait and see if it locks up again
glisse: marienz: let me know
otaylor: glisse: have you looked at all at GLX_INTEL_swap_event?
glisse: otaylor: no, not yet
glisse: it's somethings we need to do for buffer swapping on kms
otaylor: glisse: well, right now it's being advertised for radeon, but the events don't arrive
otaylor: glisse: which makes all clutter based stuff very unhappy
glisse: weird it shouldn't be advertised
otaylor: It's just unlaterally turned on in the DRI2 code in mesa
glisse: oh sweet
glisse: thanks intel
glisse: i should pay more attention to mesa commit list
glisse: i guess the bugs mailling list distract me too much
oga: who's responsible for classic mesa these days?
oga: found a bug
airlied: oga: which chipset?
oga: r1-500
dougmencken: airlied, hi, can you help?
oga: specifically, radeon_cs_legacy uses DRM_RADEON_IRQ_EMIT completely wrong
oga: firstly it calls it as drmCommandWrite, when it is R/W (openbsd checks this and fails the ioctl if it is wrong)
airlied: otaylor: we haven't got page flip support yet whichI think it needs
oga: seconly it is trying to priovide an irq cookie to wait for, when neither linux or openbsd pays attention, the kernel gives you the cookie
oga: I can provide a patch for the former issue, i need to read more code before I can deal with the latter
glisse: airlied: i think too we need page flip/buffer swap or whatever we name it today for this extension to make sense
airlied: glisse: me too
airlied: just have to write it :-(
airlied: oga: cool let me have a look
dougmencken: airlied, look at this please: Faulting instruction address: 0xf2445ed0 \n Oops: Kernel access of bad area, sig: 11 [#1] \n http://fpaste.org/vyNw/
glisse: yeah and it's somethings we should have done time ago ...
airlied: dougmencken: oh thats not nice
glisse: adding it to my todo list
glisse: so we can get it in 2.6.35
airlied: glisse: I think suokko started it maybe
oga: airlied: sure, let me make a patch. mesa-dev a good place?
glisse: oh will ping him
oga: or dri-devel?
dougmencken: yep, it's weird;
airlied: oga: mesa is best
dougmencken: airlied, also, who to blame for bad ATi Rage support
oga: dougmencken: the sands of time.
dougmencken: hmm?
dougmencken: oh, if you all are on the nevest radeon 9000 watts 32 coolers 256 TiB of RAM...then sorry
airlied: dougmencken: r128 is 10 years old I would gues
airlied: so the number of people who test them in powerpc configs is probably 1, which is you
DanaG: Rage 128 is ancient...
dougmencken: airlied, I bought ATi Rage Orion in 1998
airlied: though we haven't touch r128 in a long time
airlied: we split it out into its own driver and let it sit there
dougmencken: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=578269
airlied: dougmencken: http://fpaste.org/XU8B/raw/
airlied: can you apply that, no way we should get into that function on the macc
dougmencken: is it for my Radeon 7000 or for Orion?
BioTube: besides, the radeon 9000 is old, too
airlied: oga: that seems fine to me the passing in pending_age
airlied: it passes in a pointer to it and the kernel copys to user into the pointer
dougmencken: airlied, where to apply it? on kernel?
jcristau: dougmencken: yes
oga: ah, yes. misread that.
oga: the drmCommand type is definitely wrong though
oga: i'll send a patch for that to mesa later today
Ke: oga: so you are working on GEM/TTM ofr OBSD?
Ke: for
oga: correct.
airlied: oga: the Write should be fine as in theory the kenrel copies nothing back
airlied: but it might be wrong value in-kernel
oga: not if you check ioctl types as we do
oga: that ioctl is read write.
oga: it always has been
airlied: yeah I suspect that is wrong, but yeah easier to fix userspace, probably an abi break to touch kernel onw
oga: ok, i'll mail the patch when I sort out a mesa git tree.
oga: found that with 7.8.1
dougmencken: so I need to rebuild a kernel due to 1 log line?
dougmencken: okay, will try
airlied: dougmencken: oh I thought you had it built already
dougmencken: airlied, yeah, but again... okay okay
airlied: there is no way it can oops there unless memory corruption
airlied: dougmencken: this won't be fixed without kernel compiles a lot more I'd guess
airlied: nobody has that old hardware anymore :-)
airlied: though I think I have to fix ppc/rv100 at some point I just haven't got hw to do it on yet
airlied: I tihnk some of the ppc servers use it
dougmencken: well, you claim rv100 is supported on your dri site; many guys told me my card *must* be supported, because it's listed as supported on your site
airlied: dougmencken: its not an ati rv100, its an apple fucked with rv100,
airlied: flas it with an x86 bios it'll work
dougmencken: it's ATi in its good ages made a good card (not bios-x86-one)
airlied: its an apple card, apple cardss require hardcoding in the drivers
airlied: since apple hardcode in their drivers
airlied: so we can't add support for them unless we actually have them
airlied: it sucks, but Apple doesn't want people running Linux and made it hatd
dougmencken: do you really think apple wrote a bit of openfw-drivers for ati?
airlied: dougmencken: yes
airlied: or ati wrote them for apple
dougmencken: airlied, I mailed ati; ... yeah, ati wrote them for apple
airlied: either way we aren't ati or apple, and its an 10 year old card
airlied: the ati developers who did it are probably long gone and afaik their apple team is a world onto itself
dougmencken: but you claim support for rv100 (at least update your site saying "only x86-special-for-ibm-pc-100%-compatible-bios-machines")
airlied: wonders which site
glisse: dougmencken: because 99.99% of people are x86
airlied: we support the chip fine, we don't support the crap in the bios on it
dougmencken:
dougmencken: and a lot of others told me "your card *is* supported well"
dougmencken: glisse, wrong: 99.9999999999999999999999999866745% are
dougmencken: yes, I'm frustrated
dougmencken: " the card is listed as supported but i can't really say i have any experience with ati cards that old"
dougmencken: "according to freedesktop it should give you 3d accel"
dougmencken: more?
airlied: I changed the wiki to say powerpc cards are best effort
airlied: though I'd suspect you'd rather we got the card working , than bitch about what a wiki says
dougmencken: please don't call anybody "bitch", okay?
airlied: I didn't
airlied: grammar matters
dougmencken: a lot of guys (on other channels, on e-mail) told me my crad *is* supported
dougmencken: card*
dougmencken: so it's all about that
dougmencken: ]but...
dougmencken:
dougmencken:
dougmencken: non-standard firmware! oh no, then sigbus
airlied: the firmware contains all the useful info about the card, not having means someone needs to write support specifically for that card
mjt: oww
dougmencken: bios == firmware
airlied: nobody has because it seems like nobody who cares owns the hw, or has been made to fix it
mjt: dougmencken: you're still fighting that??
DanaG: or rather, VBIOS. video BIOS.
dougmencken: mjt, yes
mjt: oh well... :(
dougmencken: I can put this card into my G5 and report another bugs...
dougmencken: because r128 driver as well as radeon driver can't deal with mac cards
glisse: it can deals only with mac cards it knows of
otaylor: airlied: I don't see why, offhand, it has anything to do with page flip support
glisse: which are some r2xx,r3xx last time i check
dougmencken: glisse, which mac cards do you support?
dougmencken: the ones with of pef drivers in it
dileX: [22:29:34]
airlied: dougmencken: no the ones people have added support for
glisse: one in powerbook and one in G5, i am sure they used to work and i heard of people having them working in recent month
dileX: means substitute OF with X86 bios
glisse: one in G5 being radeon 9600
dougmencken: glisse, may they post their openfirmware device tree dump?
glisse: what good it will be ?
otaylor: airlied: the extension just tells you that the swap happened - now, it's not that interesting if you don't have asynchronous swaps - if glXSwapBuffers is synchronous and blocks, it doesn't provide extra information, but asynchronous swaps could be either copy or page flip, and even with synchronous swaps, you could still send the events
dougmencken: glisse, to see is it real mac card or is it pc-bios one
glisse: it's a real mac card i have one
glisse: i used to have some of them
airlied: dougmencken: http://pastebin.com/4yPXBYQd is the current list
glisse: i only got 1 left
glisse: and it was working 1 year ago
Dice-Man: RISE OF NATIONS
airlied: otaylor: hmm have to ask krh/jbarnes whats needed then
airlied: otaylor: I've been waiting for dri2 to settle down again before I tackle it
otaylor: airlied: well, if it requires more, then we at least need to not advertise it...
DanaG: https://bugs.launchpad.net/ubuntu/+source/linux-ti-omap/+bug/541030 -- that was the most relevant one I could find.
dougmencken: glisse, can you post the output of this script: http://ftp.osuosl.org/pub/manulix/scripts/dump-device-tree.sh
airlied: otaylor: I pinged krh/jbarnes on #dri-devel
glisse: my G5 is dusting, i might bring it back to life at one point
glisse: but it's definitly a native apple GPU as it did came with the G5
UnNamed: MostAwesomeDude: hey, got the glxinfo?
MichaelLong: my G5 has a big hardware bug
airlied: my g5 has nvidia ;-)
MichaelLong: it is shutting down when running a few seconds at 2 Ghz :(
dougmencken: glisse, well G5 does nothing with DeviceKit drivers bundled with *non-flashable* radeons
glisse: it was back when you coould choose btw nvidia or ati
dougmencken: for OS <= 9.2.2
Dr_Jakob: airlied: I'm sure you are making sure that nouveau runs on it ;-)
dougmencken: I tried to put random values into r100.bin << nothing changed, it's non-flashable
glisse: i know people that were ready to send marcheu G5 but he was affraid that i would dump him in the middle of a lac tied to the G5 ;)
jcristau: glisse: now he ran away to california maybe he's safe...
dougmencken: Dr_Jakob, nouveau is x86(_64) only
airlied: dougmencken: it works on poewrpc
airlied: it works on mine at least, though th 3D driver needs work
dougmencken: airlied, I love *ati*
glisse: jcristau: i am trying to get nvidia to contract someone to get rid of him ;)
jcristau: haha
glisse: after all nvidia headquarter are few miles away of his place
airlied: dougmencken: blame apple, you bought it from them
dougmencken: the mac os [x] support in ati cards were much and much better than nvidia ones; remember quickdraw3d and 3DMF
dougmencken: and that Rage(tm) exclusive stuff (like exclusive windows-only non-needoed direct3d today)
dougmencken: airlied, what can we do?
Kaapa: hey guys. I have my toshiba laptop connected to an external source. While everything works, the quality is not very good. it sort'a blinks like it had a really really low refresh rate
Kaapa: is this issue familiar to anyone? It's on 1920x... something :p
dougmencken: Kaapa, glxinfo | grep renderer
Kaapa: I don't have the external monitor here - does that work even with lvds only?
dougmencken: lvds is so much well known... what's it?
Kaapa: sorry, I didn't understand that last sentence
airlied: dougmencken: we can figure out why it oopses, fix that, then make it work ;-)
dougmencken: Kaapa, what is your question?
dougmencken: airlied, it's promising
Kaapa: dougmencken: if there's somerthing I can do to stop the flickering on the external monitor
Kaapa: dougmencken: the command gives: OpenGL renderer string: Mesa DRI R600 (RV630 9581) 20090101 x86/MMX/SSE2 TCL DRI2
dougmencken: Kaapa, your card? your setup?
Kaapa: it's a toshiba laptop: ATI Technologies Inc M76 [Radeon Mobility HD 2600 Series]
dougmencken: Kaapa, cool, and your renderer string is okay
UnNamed: Kaapa: does it flicker in all monitors?
Kaapa: UnNamed: didn't test.
Kaapa: I tested with a mac connected to the same monitor and the display is ok
Kaapa: I can work with it - nothing THAT bad, but a bit annoying when the mac guy starts being ironic :p
dougmencken: any Xorg.log maybe?
Kaapa: "buy a mac and that won't happen" :)
dougmencken: Kaapa, please don't mock on me
Kaapa: no, that's what he tells me :p
DanaG: Kaapa: try the thing without power cable plugged into the laptop.
UnNamed: Kaapa: if the hw is wrong, it could happen (go hammer their macs, or electroshock them, and see who laughs ;] )
DanaG: I've heard of ground loops going from ac -> laptop -> monitor -> ac.
dougmencken: Kaapa, I'm on mac, the one which apple dropped in trash
Kaapa: http://kaapa.pastebin.mozilla.org/716347 <- this is my xorg.log
Kaapa: DanaG: same with power cable unplugged
Kaapa: that I use often
DanaG: ah.
Kaapa: any tips or settings that I could look for?
UnNamed: Kaapa: i would test the laptop output with more monitors, to discard hw issues
dougmencken: Kaapa, "vesa"... are you on ibm 100% compatible pc?
dileX: Kaapa: why do you use radeon.new_pll=0 in kernel-command-line?
dileX: Kernel command line: auto BOOT_IMAGE=Linux ro root=803 radeon.modeset=1 radeon.new_pll=0
Kaapa: dougmencken: hum, not sure how to answer that question. It's a standard laptop...
Kaapa: dileX: when I upgraded to .33, suspension stopped working
UnNamed: dileX: maybe he read https://bugs.launchpad.net/ubuntu/+source/linux/+bug/543045 :]
Kaapa: I had to put that flag there
Kaapa: someone suggested here that I turned it off
Kaapa: actually, no, that wasn't the issue
Kaapa: far worst :S
Kaapa: with pll, the external monitor detected 59Hz as the output frequency
Kaapa: and *that* was unusable
Kaapa: at least unti pat gives me a new kernel in -current, I won't take out that option
evil_core: airlied told me I needpageflip for vsync, but its not true, I get vsync even without it
evil_core: or maybe its notr real vsync?
airlied: we have tearing avoidance at the momen
airlied: but its not real vsync
airlied: as in the client app can't request/control it
evil_core: I am using UMS
airlied: this is all KMS btw
airlied: UMS is dead to me
evil_core: and I got ~60fps in glxgears when enabled
evil_core: maybe to you, most glide games(in wine) doesnt run in KMS :/
airlied: in UMS driconf vsync stuff should work fine
airlied: do they run too slow?
evil_core: foI got r500
evil_core: no, they usually dont run at all
evil_core: like Dethkartz
UnNamed: speaking of not running, is FBO complete in r300c+kms?
evil_core: and Ignitio ngot problems
evil_core: UnNamed: is it special branch?
UnNamed: i got one case in which the program crashes, and only obvious thing i can nothing is warnings about FBO
UnNamed: evil_core: no, i mean running r300 classic over kms
evil_core: http://carme.pld-linux.org/~evil/radeon/r5xx_pageflip/
airlied: UnNamed: fbo may not work in all cases, though I think it got most of them, r300g has more complete fbo stuff in theory
evil_core: anybody can tell me whats wrong there?
UnNamed: airlied: if you want to look http://pastebin.com/JuiF3XcV r300c and r300g info
evil_core: it locksup GPU, once I got info in Xorg.log about infinite GPU loop
evil_core: and sending TERM to Xorg causes hardlockup, ecen syrq stops working
UnNamed: s/thing i can nothing is/thing i can notice is/
evil_core: nobody have time, or nobody wants support UMS?
MostAwesomeDude: I don't want to support UMS, and I doubt anybody else does, either.
MostAwesomeDude: You're free to hack in your spare time, though.
evil_core: shouldnt vsyncprivide much power savings?
evil_core: vsync provide*
ghepeu: airlied, I saw you were talking of tearing avoidance
ghepeu: it doesn't work on my r300
ghepeu: or, it doesn't always work
ghepeu: it's ok if I start a 3d apps (even a compiz plugin like the benchmark plugin), otherwise tearing when moving windows and in xv textured video
evil_core: ghepeu: do you now that desktop isnt double buffered?
evil_core: ghepeu: and not "when you start 3 apps, but compiz is compositing manager and windows doesnt render to desktop, but to textures which are composited using OpenGL as desktop later
ghepeu: it used to work okay three months ago, it still works with a compiz plugins, textured viseo is broken too
evil_core: it for me on r500 also broken, at 1st it uses much CPU
evil_core: I can play on UMS two 1080p simultanously, in KMS even one I cant :/
airlied: wonders do we use bicubic on Xv on ums
dougmencken: I'm on fbdev/radeonfb now: resolution is okay (it's native for display)
evil_core: r300/compiler means classic and UMS?
UnNamed: evil_core: do you ask about "r300c"?
evil_core: no, r300/compiler in commits
UnNamed: aah, ok, sorry
dougmencken: it's so cool to have 1600x1200 (instead of 1024x768), it's like 4x and even 8x times a desktop; and it's native; and it's your hated fbdev@radeonfb, no radeon.modeset=1 now
evil_core: but it sucks at suspending probably ;)
evil_core: and UXGA works fine with KMS
DanaG: why use radeonfb?
DanaG: how about fbdev on radeondrmfb?
evil_core: yay, nice speed improvement in UMS :D
TCW: mjt, it seems xserver-xorg-video-radeon is not to blame here, see http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=578270 if interested. Btw. how do you start compiz?
Ronis_BR: does radeon driver have some problem with SDL/Opengl
MostAwesomeDude: No?
Ronis_BR: MostAwesomeDude: SDL/OpenGL softwares here can't use SDL_HWSURFACE and is VEEERY slow
MostAwesomeDude: Ronis_BR: Is GL accelerated at all? What does glxinfo say?
Ronis_BR: MostAwesomeDude: yes, direct rendenring works perfectly
Ronis_BR: MostAwesomeDude: I took two examples from nehe one using SDL other GLX, with GLX it is amayzing, with SDL it is very slow
Ronis_BR: MostAwesomeDude: I'm asking to avoid installing catalyst when it is a known bug
MostAwesomeDude: Ronis_BR: Are you sure it's our bug? SDL-based stuff works fine here.
Ronis_BR: MostAwesomeDude: no I don't
Ronis_BR: MostAwesomeDude: know where the bug is
Ronis_BR: MostAwesomeDude: I'm shooting to all directions :D
Ronis_BR: MostAwesomeDude: what video driver do you use with sdl?
MostAwesomeDude: Ronis_BR: r300, r300g. Both work fine.
Ronis_BR: r600 here
Ronis_BR: MostAwesomeDude: my SDL/OpenGL app that I could run with 50 FPS on a X200 runs with 8 FPS here :D
Ronis_BR: MostAwesomeDude: maybe my programming is wrong
Ronis_BR: MostAwesomeDude: or too much old
MostAwesomeDude: airlied: IIRC the bicubic path is only on KMS these days. I could be wrong though.
evil_core: I got black screen when loading radeon with modeset=1 :/
evil_core: does mine patches affects KMS in any way?
MostAwesomeDude: Are they kernel patches?
evil_core: yes
MostAwesomeDude: Link?
evil_core: http://carme.pld-linux.org/~evil/radeon/r5xx_pageflip/
evil_core: but its custom kernel config, not tested before
evil_core: anyway it works in UMS when pageflip is disabled(otherwise GPU hung)
MostAwesomeDude: Shouldn't change KMS.
evil_core: so its another bug
evil_core: everything right in dmesg
evil_core: no errors, etc :/
airlied: evil_core: so for KMS videos you should try turning off bicubic
evil_core: airlied: bivubic in compiz?
airlied: no in xvattr
Ronis_BR: MostAwesomeDude: tell me something, should we expect a performance gain with radeon in the future? or everything is optimized and I need to change my card? :D
evil_core: its not enabled in UMS?
evil_core: airlied: currently I have black screen with KMS with backlight disabled
MostAwesomeDude: Ronis_BR: r600 is so far away from optimization that it's not even funny.
evil_core: missed fbcon ;)
evil_core: airlied: can you give me command?
evil_core: I am on KMS now
Ronis_BR: MostAwesomeDude: very good to know
evil_core: anyway movie works now
Ronis_BR: MostAwesomeDude: because performance (for me) is already acceptable
evil_core: but probably will start lagging after some reboot or suspend
Ronis_BR: MostAwesomeDude: because I just use composite of 3D and eventually play some games
airlied: xvattr -a XV_BICUBIC -v 0
airlied: that might do it
evil_core: ok, thanks
evil_core: something is fucked up in KMS
evil_core: it works way bette than few days ago
evil_core: but especially in gallium in q3a the room is msaller the game is more laggy
evil_core: the less I see(near wall) more lags
evil_core: anyway Ignition works
evil_core: but Glide stil wont work :/
evil_core: airlied: pm2 branch is synced to d-r-t?
airlied: evil_core: it should be based on d-r-t
evil_core: I know its based, but if its updated frequently
evil_core: synced to d-r-t
evil_core: it owuld be nice to get pm2 and fresh d-r-t
airlied: you can just use git pull to do it
airlied: I doubt I'll keep them synced, its too much effort
evil_core: KMS glxgears makes noise on mine T60p
evil_core: but doom3 works 1000x quicker than in UMS
evil_core: uknown problem initializing graphics
evil_core: nfs6: hot pursuit 2
evil_core: similar problem to glide wrappers
evil_core: http://pld.pastebin.com/spam.php?i=FMAGkP1c
evil_core: its KMS
evil_core: [drm:radeon_cs_ioctl] *ERROR* Invalid command stream !
evil_core: [drm:r100_cs_track_texture_check] *ERROR* Texture of unit 0 needs 8320 bytes but is 4096
evil_core: [drm:r100_cs_track_texture_print] *ERROR* pitch 1
GoGi: If you use Alt+SysRq+K you can't switch to a text console anymore
GoGi: however you can restart X from an ssh session and then everything works again