robbat2: airlied, I tried an X700 w/x86-firmware on my G5 originally, which didn't work at all, fbconsole or X
robbat2: airlied, trying my hand and digging in xorg-server+libpciaccess atm, the scanpci tool prints the device correctly, so it's corruption somewhere in xorg-server I think
airlied: robbat2: you could try commenting out the goto fails after, the setoperatingstate
airlied: and see what it actually does
airlied: it might just fail somewhere later..
robbat2: hmm, yeah, give me a sec, since the MMIO output later does have the correct address
robbat2: segfault, because the atomBIOS pointer is null
robbat2: #7 atombios_get_command_table_version (atomBIOS=0x0, index=12, major=0xffd19908, minor=0xffd1990c) at radeon_atombios.c:2439
robbat2: 2439 ATOM_MASTER_COMMAND_TABLE *cmd_table = atomBIOS->BIOSBase + atomBIOS->cmd_offset;
robbat2: (WW) RADEON(0): Unrecognized BIOS signature, BIOS data will not be used
robbat2: that's in the log
airlied: robbat2: hmm maybe you don't have an atombios... which would be the wierd..
airlied: robbat2: can you extract a copy of your BIOS?
robbat2: what's the easiest way to do that?
airlied: cd /sys/bus/pci/devices/
airlied: echo 1 > rom
airlied: cat rom > /tmp/rom
airlied: echo 0 > rom
rx__: hey airlied
airlied: rx__: hey
rx__: is src/atombios_crtc.c:243 necessary?
robbat2: airlied, it's empty.
airlied: robbat2: thats not so good..
airlied: robbat2: wonder if your kernel is busted..
robbat2: stock 2.6.23 here
airlied: rx__: I could probably drop sclock.
airlied: rx__: hehe. so I set it earlier.
airlied: oops..
rx__: ;)
robbat2: i'll grab a recent git kernel and give it a spin
airlied: robbat2: yeah sounds like a ppc issue..
airlied: robbat2: do you have another pcie 16x slot?
robbat2: the G5 has x16, x8, and 2 x4
airlied: which slot have you it in?
robbat2: the x16
robbat2: my x8 has an eSATA controller
airlied: hmm I'd expect the rom to work there..
robbat2: i'm going to bump to .24_rc3-git1 and retest
airlied: robbat2: hmm I think it might be busted.. I can't read the bios on the second card I have in my g5 x8 slot.
robbat2: ooh, look at this in the lspci
robbat2: Expansion ROM at 90020000 [disabled] [size=128K]
airlied: robbat2: not so good..
robbat2: every Expansion ROM entry in the entire lspci is marked as disabled
robbat2: so maybe I should give radeonhd a spin, since they don't use AtomBIOS ;-)
airlied: robbat2: I suggest annoying benh ;-)
airlied: robbat2: they still need values from it.. but may set good defaults..
airlied: robbat2: annoy kernel ppl until roms appears then get gpu driver :0)
airlied: robbat2: the days of hacking around the OSes lack of things are over..
robbat2: yeah, i'm talking to benh now
airlied: bbl..
robbat2: airlied, new kernel has no effect. none of the 3 drivers (ati, avivo, radeonhd) succeed, all claiming lack of AtomBIOS
robbat2: benh said it's possible that the card doesn't have any atombios, or that it's relocated
mcgreg: how is the current radeon driver (xorg-driver-ati) installable? git? is it in the current tree already included?
rx__: robbat2; you already tried dumping the bios right?
robbat2: rx__, yup. by multiple methods now
rx__: this? http://tirdc.livejournal.com/7704.html
robbat2: rhd_conntest, avivotool, and sysfs directly (that one)
rx__: hmm.. :(
rx__: can you pastebin your lspci -v?
robbat2: i sent it to the list the other day
robbat2: one sec, let me get you a link
robbat2: http://lists.x.org/archives/xorg-driver-ati/2007-November/003134.html
rx__: oh btw.. from above.. i think you want the line that has 64K...
rx__: not 128k
robbat2: i thought the 64k was the mmap io registers?
rx__: maybe i'm wrong
robbat2: one sec, let me rig up a crazy hack and try to read parts of /dev/mem
rx__: every expansion rom in my lspci comes up disabled too so i don't think that's an issue
robbat2: you're on PPC as well?
rx__: uhm.. no :)
robbat2: in your Xorg.$N.log, grep for 'BIOS @' ?
rx__: BIOS @ 0xfefc0000/17
rx__: yeah that corresponds to the expansion rom
robbat2: does that line up with the Expansion Rom address in your lspci?
rx__: you're right
rx__: sorry about that
robbat2: so the disabled doesn't mean much, but it's more critical that mine seems to be empty
rx__: is it possible to stick in a x86 box? ;)
robbat2: for two reasons no
robbat2: 1. no x86 box around with free PCIe
robbat2: 2. it's the mac version with openfirmware on it
rx__: great
robbat2: i've got an x86 box, but in it's sole PCIe slot, i've got a SCSI controller
rx__: i was thinking maybe you could isolate the problem.. use a livecd if you have to
rx__: but i guess not :)
robbat2: trying to use pread against /dev/mem with the addresses from lspci, I can't read any of them at all
robbat2: BAR0/2/ROM
robbat2: i'm going to try something else
robbat2: I have an x86 X700 around, that I didn't use before because you can't boot the G5 with it. however if I leave the X1900 in the x16 slot, and put the X700 in the x8
robbat2: i'm wondering how it will go
rx__: hmm.. good luck
robbat2: that in the morning I think
robbat2: bedtime now
OipOS: What's the current status on supporting the Xpress 200M series?
egore: airlied: interesting stuff there in your xf86-video-ati git repository. I was wondering when you get bored telling people that it's so easy to get r300 with composite working and start it yourself ;-)
mcgreg: robbat2: as far as I read, you have instaleld latets version of radeon driver, now how did you do that? via git? I would very much like to try my x1300
robbat2: mcgreg, grab the atombios-support branch of the git tree, and build
robbat2: however that only gets you the driver. most (~40%) of my X is built from the git sources
robbat2: if you're using Gentoo, start with the x11 overlay via layman, and modify the ebuilds to pick the branches you want
mcgreg: no, using debian sid
mcgreg: hmm I guess it wouldnt be an easy task for me, since I am not very experiences in git/compiling X-modules ;)
robbat2: having a fair knowledge of C and build systems is a help
mcgreg: and besides that, there might be compatiblity problems between current git and my X version
robbat2: yeah, I didn't do it on my active work machine
robbat2: (just the box that I'd like to migrate to once it's fully operational)
mcgreg: well, I guess I'll wait .. but I would be very interested if it works and how good it is since I am planning to use it once it is (more) stable .. and probably once it has 3d support for r500, I'll never touch anyhting else :)
egore: I'm running the atombios-support branch against a normal 1.4 x server and it works great
ordchar: egore, what card?
robbat2: is there anybody else besides me on PPC?
mcgreg: I was , 7 months ago ;)
egore: ordchar: x1650
mcgreg: hmm I did : git-clone git://anongit.freedesktop.org/git/xorg/driver/xf86-video-ati ; cd xf86-video-ati/ ; git-checkout origin/atombios-support .... do I need to do something else for the r500 support?
ordchar: mcgreg, not that i know of
mcgreg: ok, thx :)
mcgreg: ok, lets see, ibnstalled and configured xorg.conf ..
mcgreg: ok, X in general works at leats so far ;)
mcgreg: radeon driver seems to work fine , I looked into the x-log and it said atom bios detected, loaded xaa and stuff.. no special error (ecxept no dri of course)
mcgreg: looks great for now
mcgreg: ahh great. changing resolutions on the fly works fine, just tried wine with diablo2 and it worked great :)
dli: mcgreg, desktop or mobile?
mcgreg: desktop
mcgreg: x1300pro
mcgreg: amd64
mcgreg: pci-e
dli: mcgreg, looks like it only works for desktop up to now
mcgreg: ok, starcraft works fine too, great, my 2 favourite games works with no problems :)
dli: mcgreg, what about speed? like video playback?
mcgreg: hmm I dont have any speed issues, is there any test in particular I could do to tets it?
dli: mcgreg, like fullscreen dvd video?
dli: mcgreg, I couldn't play anything fullscreen right now with radeonhd, too slow
mcgreg: hmm well, mplayer doesnt allow me to do fullscreen ;) but I will play a dvd now, and see what I can do :)
dli: mcgreg, try "mplayer -vo xv", or "mplayer -zoom -vo x11"
dli: mcgreg, you can always type "f" to get fullscreen
mcgreg: dli: hmm well, I though "radeonhd " has no 2d acceleration, like the current "radeon" driver (no xaa?)
mcgreg: dli: if I press f, the screen get filled, but the video stay as it is ;) senseless, but it is like that ;)
dli: mcgreg, avivo allows me to "mplayer -zoom -vo x11" and fullscreen
dli: mcgreg, try "mplayer -zoom -vo x11" then, fullscreen
mcgreg: yeah, one minute :)
dli: if it works, you can put "zoom = yes " and "vo = x11" in ~/.mplayer/config
mcgreg: ok, so far : first, mplayer (normal mode) dvd player matrix3(VO: [x11] 720x576 => 1024x576 Planar YV12 [zoom]) , cpu usage: mplayer 15% , xorg 6%
mcgreg: that was not fullscreen though
dli: mcgreg, works?
mcgreg: and full screen , it works just normal
mcgreg: no speed issues at all
dli: mcgreg, I have to wait airlied to fix fuzzy display for mobile first :(
mcgreg: honestly I havent expected it to work as good as that :)
mcgreg: oh , and my fullscree is 1280x1024x32
dli: mcgreg, avivo works like that for me, but avivo has lid-closing issue, display messed up after lid closing/opening
mcgreg: btw, shozuld suspend/hibernate to ram/disk work?
dli: mcgreg, I have no idea, radeon doesn't work for me yet
mcgreg: testing hibernate-ram now :)
mcgreg: worked just fine :)
dli: mcgreg, I hoped avivo would survive, so facing competition, radeonhd would not turn out to be another fglrx. Now, we have radeon, should be great for users
mcgreg: yeah, I am very happy with that
mcgreg: dli: thx for the tipp with "zoom = yes " and "vo = x11" in ~/.mplayer/config"
dli: mcgreg, no problem
mcgreg: egore: a few hours ago we talked about the new driver .. I got i installed and working and it works fine :) better than I have expected
egore: mcgreg: the new driver does not depend on anything new outside the driver so I expected it to be that way :-)
mcgreg: egore: well, yes, but I havent expcetd to get full speed with dvd-fullscreen without xv or gl ;) but well, actually with that hardware it should be without problems :)
mcgreg: egore: yes best parts are, diablo2 and starcraft works without problem, and as I said, fullscreen video is working fine .. so except 3d everything is very very fine
mcgreg: even suspend2ram works fine
egore: ok, haven't tested that far
egore: I really like what airlied and agd5f did that far
mcgreg: yeah me too and I tahnk them very very much for their work
mcgreg: thank
robbat2: airlied, I came up with a test, and I think it proves that the 'rom' sysfs interface/kernel on PPC is broken
robbat2: rx__, of interest to you as well ^^
groo: hi/2 all.. is airlied alive?
groo: i have some questions regarding xf86-video-ati and the rs485 chipset
airlied: groo: kinda here.
[_Groo_]: i have some questions regarding xf86-video-ati and the rs485 chipset, anyone alive?
airlied: [_Groo_]: ask q than hang around to see if anyone knows :0)
[_Groo_]: well i backported xserver 1.3 and dependencies (lime mesa 7) to my feisty system (don't ask why i don't upgrade :D) and after i compiled the git xf86-video-ati, i now have 3d support
[_Groo_]: i wrote about it in your blog comments
[_Groo_]: BUT aiglx doesn't work.. X crashes, and the vt switch is broken.
[_Groo_]: sooo, do you need any help from me ? (bug report etc,) to add suport for my card faster? :D
[_Groo_]: and why does aiglx works in all the xf86 ati card expect this one?
airlied: when you say aiglx do you mean compiz?
[_Groo_]: airlied: yep :)
airlied: or does LIBGL_ALWAYS_INDIRECT=y glxgears work?
[_Groo_]: aiglx works... the xorg.log shows that.. as well dri ... since im using kernel 2.6.22
airlied: [_Groo_]: for me compiz runs but renders all wrong
[_Groo_]: i tried glxgears directly and it worked just fine.. when i try to run compiz with or without LIBGL_ALWAYS_INDIRECT=y the X dies
[_Groo_]: airlied: same chipset?
airlied: [_Groo_]: I don't have it anymore I used to have a lend..
airlied: [_Groo_]: but yes rs485
[_Groo_]: hmmm, how can i help you then?
airlied: [_Groo_]: does the X server print something about evictig pixmaps?
airlied: before crashing?
airlied: [_Groo_]: in the Xorg.0.log
[_Groo_]: airlied: can't tell right now, since i went back to fglrx
airlied: [_Groo_]: well compiz won't work in any cases as there is somewierd bug in vertex shading..
airlied: [_Groo_]: as rs4xx don't have vertex shader hw..
[_Groo_]: yes, i read in your blog..
[_Groo_]: bug is a mesa bug or a driver bug?
[_Groo_]: i mean but
airlied: mesa bug
[_Groo_]: airlied: do you know if amd will release any docs for the rs4xx in the next batch?
airlied: but a mesa driver bug..
[_Groo_]: drm?
airlied: [_Groo_]: I doubt it.. we mostly know how it works..
airlied: [_Groo_]: just no developers own them ..
[_Groo_]: is it hopeless then? stuck with fglrx?
airlied: [_Groo_]: until someone fixes it compiz isn't going to work..
[_Groo_]: well if you have the time to tell me what you need i can send whatever info i can about this damn card
airlied: [_Groo_]: no it isn't possible to dump anything.. it just needs developer + machine in one room..
airlied: and time..
airlied: I can reproduce it on real r300 hardware..
[_Groo_]: well that i can't do.. i just have this acer :)
[_Groo_]: i see..
airlied: but haven't had time/inclination to chase it down ...
airlied: really I need a testcase that shows the same problem that isn't compiz
airlied: as debugging it with compiz is major pain..
[_Groo_]: ah, just for the info, the xrandr rotate worked, kinda... it really rotate my desktop, but didn't changed the vertical/horizontal size (croped the desktop), and then the laptop freezed.. but.. its the only driver except the windows one that i saw being able to roate the desktop
airlied: [_Groo_]: yeah it goes very slow.
airlied: [_Groo_]: hoping to fix that as well.
[_Groo_]: wine implements their own vertex shader code, can't ou hack something around it?
airlied: [_Groo_]: no its a hw interface problem..
airlied: [_Groo_]: we are sending something to the card it misinterprets and renders compiz wrong
[_Groo_]: ahh another good news, glxgears gave me the same FPS the new fglrx driver does.. i don't know if its a real improvement but...
mosgreg: [_Groo_]: I dont think glxgears is a good ground to test 3d speed ... what about a quake3-demo? doest it work well?
[_Groo_]: well you probably know this bt if you are running fglrx , rmmod it and modprobe the radeon module, X goes crazy 100% of the time, and the laptop freezes... my understanding is that fglrx and radeon activate diferent interrupts for the same function
[_Groo_]: gonna try those next week if i can find the time...
[_Groo_]: do you use the fglrx driver?
mosgreg: well, I used it till today
mosgreg: I thunk I will use the radeon driver now, mostly
mosgreg: think
mosgreg: since the last relesses of fglrx are just a big mess
mosgreg: opengl is not usable with it
mosgreg: since it leacks memory like "whatever"
[_Groo_]: i know why aiglx is broken , my question is: when amd finally add the remaining texture_from_pixmap to the server xetensions and fixes the composite code, does the true opengl apps still will get stuck to the screen (aka no root window)
mosgreg: well, there are some more important problms with fglrx that just stupid aiglx ...
[_Groo_]: mosgreg: well in my tests, yes opengl apps leak memory but after a while they stop!.. example, glxgears and fgl_gears start leaking but both stop about 4% of the memory each (2G)
[_Groo_]: mosgreg: yeah i know, but that particular bug is lack of extensions in fglrx right?
mosgreg: here they did .. I ran a tets and after 1 minute I lost >7% of my 2GB ... and I know of some games (warcraft3) that after a few minutes they totally crashed the computer (out of mem)
mosgreg: yeah, I think so
[_Groo_]: 8.42 or 8.43 (aka 7.11)
mosgreg: both do that
[_Groo_]: strange.. here they go up a little and then stop
mcgreg: uhh sorry ..
mcgreg: I just started "google earth" to tets if it would work with indirect rendering ...
mcgreg: but it just restarted my xserver
[_Groo_]: well gotta go... thanks airlied and mosgreg.. airlied i'll keep following your blog and gitweb, and just hope that eventually the rs485 will be 100% supported :d getting close though
mcgreg: now thats not nice
mcgreg: [_Groo_]: nothing to thank for :)
[_Groo_]: mcgreg: if you are using compiz, you can force it to use indirect rendering by using the WM action plugin
[_Groo_]: i use it to force xv in a window, google earth and flash sometimes..
mcgreg: [_Groo_]: thx, though I have noi interest in 3d desktops
[_Groo_]: well, if you are using plain X, it should work just fine.. 3d apps i mean
[_Groo_]: with latest fglrx..
mcgreg: actually I want to get rid of it once for all
[_Groo_]: mcgreg: good luck.. i also have that dream :D
mcgreg: :)
[_Groo_]: see'ya all later :)
mcgreg: cya
mcgreg: I'll go to bed too, good night