EruditeHermit: airlied: how is f10 treating you?
EruditeHermit: tried booting the live snapshot from a flash drive, but it hung at the login screen thing
EruditeHermit: but maybe its a problem with my flash drive
EruditeHermit: is kms looking like it'll be included in f10 for radeon?
spstarr: EruditeHermit:which gpu
EruditeHermit: rv350 I believe
spstarr: EruditeHermit: if AGP and kms and r3xx|| r4xx
spstarr: EruditeHermit: laptop or desktop?
EruditeHermit: laptop
spstarr: GNOME or KDE? :)
spstarr: oooh!
EruditeHermit: AGP
spstarr: EruditeHermit: it hangs for me, you can be another test helper :))
spstarr: EruditeHermit: do you use gnome or KDE?
spstarr: or both
EruditeHermit: I use gnome right now
spstarr: EruditeHermit: if you do use KDE can you try kwin composite?
EruditeHermit: I don't care if you want me to test KDE though
spstarr: does it blow up ?
EruditeHermit: hmm
EruditeHermit: I don't have F10 installed
spstarr: rawhide?
EruditeHermit: I just tried the live CD
spstarr: even live CD is ok
EruditeHermit: hmm
spstarr: if its got KDE on it
EruditeHermit: I can download it and try
spstarr: this goes further back than Fedora 10
spstarr: F9 i had issues
EruditeHermit: thats odd
EruditeHermit: because I've tried it on Ubuntu, and it works ok
spstarr: with gnome or KDE?
EruditeHermit: I did notice that there were artifacts with kwin
spstarr: KDE seems to do something different when seting up composite
EruditeHermit: with gnome it is perfect
spstarr: yes there might me artifacts, but for me if i enable it, it will deadlock gpu
EruditeHermit: with the old radeon 6.9.0 driver
spstarr: or kill X
spstarr: thats the DDX though not the DRI
EruditeHermit: hmm
spstarr: right now there is an issue but its unrelated to this in rawhide
EruditeHermit: hmm
EruditeHermit: well when I was booting the rawhide snapshot livecd, it didn't even get past the login screen
spstarr: hopefully the buffmgr code will get a some cleanups since its hot code
EruditeHermit: it just froze at the livecd login screen
spstarr: EruditeHermit: thats due to kms and no agp support
spstarr: or working agp support yet
spstarr: EruditeHermit: Fedora rawhide F10 live cd?
EruditeHermit: yes
spstarr: if so, then yes this is known
spstarr: EruditeHermit: workaround is boot CD with nomodeset
EruditeHermit: so do you want me to try the kde version?
spstarr: this will turn off kernel modesetting (kms)
EruditeHermit: I don't mind testing Live CDs for you
spstarr: EruditeHermit: you can if you can try out the composite stuff AND disable kms
spstarr: kms does not work for our GPU yet
EruditeHermit: ok
spstarr: the Radeon 9600 M10/M11 Mobile (rv350)
spstarr: if thats what you have
EruditeHermit: yes
spstarr: thinkpad?
EruditeHermit: Dell
EruditeHermit: Inspiron 8600
EruditeHermit: its an oldie
spstarr: 01:00.0 VGA compatible controller: ATI Technologies Inc RV350 [Mobility Radeon 9600 M10
EruditeHermit: over 4 years old now
spstarr: same
spstarr: 2004
spstarr: ok so same gpu great :)
EruditeHermit: I have an r300 GPU as well
EruditeHermit: radeon 9800
EruditeHermit: AGP desktop
spstarr: EruditeHermit: im trying to gather r3xx users so we can provide as much debug on the composite failing, but there's new issues that stop testing at moment
spstarr: me too,Radeon 9800 AIW Pro
EruditeHermit: oh, mine isn't AIW
EruditeHermit: I'll download the KDE spin
EruditeHermit: and test it with nomodeset=1
spstarr: still an r300 i think yours is no difference in GPU
spstarr: you can just do nomodeset also
spstarr: but only on kernel boot line if you do that for module use nomodeset=1
spstarr: that may or may not work
spstarr: i have not tried the 9800 card since agp doesnt work on the rv350 (which is likely not to work on the r300 also)
EruditeHermit: ok
spstarr: im thinking of trying a Ubuntu live CD and seeing if it breaks there too
spstarr: Kubuntu even
EruditeHermit: it works in Ubuntu and Kubuntu
spstarr: if it doesn't then we have some patches in Rawhide that break something
EruditeHermit: in hardy
spstarr: if it breaks in Kubuntu then i will know its a bug in upstream
spstarr: EruditeHermit: KDE 4.x?
EruditeHermit: yes
EruditeHermit: there are rendering artifacts
EruditeHermit: but composite works
EruditeHermit: and the effects work
EruditeHermit: some are slow
spstarr: and if you turn it off/on?
EruditeHermit: its fine if you turn it off
EruditeHermit: no artifacts
spstarr: for me it will work briefly, then the gpu locks up
EruditeHermit: no GPU lockup
spstarr: gets kubuntu hardy today and will test that on live cd
EruditeHermit: kwin isn't as polished as compiz for 3D effects it seems
spstarr: it is in 4.2
EruditeHermit: atleast its interaction with radeon isn't as smooth
spstarr: there's a lot of optimizations being done in the code wait for that to come
EruditeHermit: they work, but they are a little slower it seemed to me
EruditeHermit: ok
EruditeHermit: I tried 4.1
airlied: damn you all with your AGP rv350s..
spstarr: airlied: hehehehehehe
EruditeHermit: so it might be better in 4.2
airlied: tries to buy one on ebayt..
spstarr: :)
EruditeHermit: do you not have one airlied?
hifi: rv350? I haz!
EruditeHermit: lol
spstarr: hifi: laptop or desktop?
airlied: EruditeHermit: I've worked out whatever chip I don't have is the one that is most broken.
hifi: wasn't 9600 Pro RV350?
EruditeHermit: lol
spstarr: is lucky today all the rv350 people are coming out
airlied: I don't have an rs480, but I got one for the afternoon to make it work.
spstarr: hifi: Pro is desktop, but its also fine for this :))
spstarr: hifi: do you use GNOME or KDE?
spstarr: or both
airlied: I bought an X600 last week now it works..
hifi: what are you guys doing exactly? :p
hifi: I don't run GNOME or KDE but for testing purposes, sure
EruditeHermit: airlied: can't you get ATI send or RH to pay for the computers you need?
spstarr: hifi: trying to get composite work on r3xx so we can play with eyecandy and such in KDE (it works in GNOME)
spstarr: hifi: which distro
airlied: EruditeHermit: its hard to get old crap.
airlied: EruditeHermit: no matter who pays for it.
EruditeHermit: lol
hifi: debian on that box with 9600 pro
spstarr: hifi: which Xorg, mesa versions , DDX?
spstarr: kernel
airlied: like I have all the shiny new r600/700 cards here..
hifi: uhh, I think the kernel is 2.6.26 and Xorg 1.4/1.5, not the current one for sure
airlied: and I'm scrounging around on ebay for a 9600..
EruditeHermit: airlied: but I imagine ATI has some of every chip lying around; surely RH can ask them to ship it to you for a week every 3 months or so
spstarr: whats the r7xx give you vs r6xx?
EruditeHermit: airlied: let you borrow it
spstarr: more shaders?
airlied: EruditeHermit: not really..
spstarr: hifi: hmm
hifi: spstarr: though if I upgrade to experimental there's the new Xorg package
spstarr: X.Org X Server 1.5.2
spstarr: im using fedora rawhide (same as debian sid basically, except with more rawness)
airlied: EruditeHermit: AMD people scrounge around on ebay also :)
hifi: Debian sid has 1.4.2 and experimental 1.5.2
EruditeHermit: airlied: lol what?
EruditeHermit: airlied: they don't even keep theirr own chips?
hifi: I suppose I need 1.5.2 to help?
spstarr: hifi: think of rawhide as experimental repo
spstarr: hifi: no
airlied: EruditeHermit: they would have a stash of chips and devel cards, but production cards aren't made by AMD
spstarr: 1.4 im not sure.. fedora has had 1.4.99.xxx which was almost 1.5 for a while
EruditeHermit: airlied: I'll ask my friend who has an old desktop AGP 9600XT that he may not be using
airlied: EruditeHermit: I found one locally on ebay so hopefully I can get it for 10 bucks :)
EruditeHermit: airlied: cool
spstarr: yum
hifi: spstarr: the current ati driver can't be compiled against 1.4 so I'm stuck with 9.6.0
spstarr: airlied: then you can see all the pains with the r3xx brings :)
hifi: unless I upgrade to experimental
EruditeHermit: airlied: if you could send a replacement AGP card to me, i'd let you borrow my AGP 9800 desktop board
spstarr: hifi: thats just the DDX driver (2D)
EruditeHermit: airlied: then send it back when you are done
airlied: EruditeHermit: hehe.. well the turn around would probably miss my window, my other AGP cards are crappy :)
spstarr: airlied: im keeping X now on a gdb debug each time i start it, as exa keeps locking gpu up randomly
airlied: spstarr: my other r3xx is fine.. its PCIE
spstarr: airlied: i wouldn't expect the PCIE vs AGP cards to differ that much(?)
spstarr: or they do
airlied: spstarr: PCIE cards don't lockup as often.
spstarr: doh
EruditeHermit: lol
spstarr: agp is the suck
spstarr_desk: hahaha
spstarr_desk: SEE
spstarr_desk: it just LOCKED UP
EruditeHermit: what the hell is up with ATI anyway. They gave ubuntu a working Xorg 7.4 driver and didn't release the same version to the public today
spstarr_desk: my ssh session is dead too
spstarr_desk: a real nice deadlock
spstarr_desk: ok so well that didnt help :/
EruditeHermit: hmm I started getting some of those with git snapshots of the xserver-xorg-driver-ati
EruditeHermit: right after all the bicubic merges
spstarr_desk: airlied: does exa still have lockup issues known?
spstarr_desk: as with no xorg.conf it is using exa
terracon: why doesn't fglrx have all these lockup? The open source driver has always been lockup city
airlied: spstarr_desk: well XAA still has unknown lockup issues.
spstarr_desk: airlied: hmm
EruditeHermit: terracon: fglrx has its own issues and lockups =p
airlied: most lockiup issues are unknown in that we've no idea why they die.
EruditeHermit: lol
spstarr_desk: airlied: if we do get GPU reset functionality would those provide incites as to why?
EruditeHermit: well if you knew why they happened it wouldn't be hard to fix =p
hifi: anyone have 1.5 Xorg debian pins for experimental at hand? :p
terracon: EruditeHermit: well I can play (name your fav game here) for hours on end, Mesa. You're lucky to get an hour in before you have hard lock
terracon: and yes agp does suck
spstarr_desk: airlied: it happened when i just loaded firefox then it locked
EruditeHermit: terracon: lol, I don't play too many games
spstarr_desk: airlied: window creation it seems even loading a movie (window creation)
terracon: EruditeHermit: well blender or name your fav 3d app
EruditeHermit: blender is ok for me
EruditeHermit: but each chipset is different
EruditeHermit: fglrx video playback for me is really slow
EruditeHermit: textured video on radeon is really good. It works better than my friend who has an intel chip
EruditeHermit: and intel actively helps their OS driver
EruditeHermit: or has more than 1 guy on it
terracon: heh
EruditeHermit: the OS radeon driver is really good for the number of people that work on it
EruditeHermit: which seems to be 2
hifi: but they have a big cheerleader group backing up!
EruditeHermit: and its not even airlied's job to work on it really
EruditeHermit: so what they have accomplished is amazing
hifi: \o/ R-A-D-E-O-N RADEON yaay \o/
hifi: hides
terracon: haha
EruditeHermit: most of it in the last year too
spstarr_desk: watches spstarr die
spstarr_desk: port 22: No route to host
spstarr_desk: resets laptop :/
terracon: It would be nice if amd/ati actually funded some dev work. But they like to use. Well Linux's market share suck donkey di** so we're only going to put 1/2 dev's on it, so there
EruditeHermit: hmm
EruditeHermit: I think they have more people than that working on it
spstarr_desk: my next experiment, use GNOME and see if i can wedge it
EruditeHermit: its just its very fragmented
spstarr_desk: if KDE can crash with EXA gnome should also
spstarr_desk: terracon: they do I think
EruditeHermit: there is their own proprietary driver, radeonhd and radeon
spstarr_desk: terracon: Red Hat people also work on drivers
EruditeHermit: ati funds all 3
EruditeHermit: or atleast helps
terracon: and let's not forget Hui Yu
airlied: spstarr_desk: we tried radeon.dynclks=0 didn't we :)
airlied: spstarr_desk: can you pastebin a normal Xorg log file from your machine as well.
terracon: waves by zombie spstarr
spstarr_home: ugh, 4 of me
spstarr_home: its bad enough having one of me ;)
terracon: I mean bye
airlied: spstarr_desk: can you pastebin a normal Xorg log file from your machine as well.
airlied: or spstarr_home
spstarr_home: from now, sure this is a normal Xorg log with no xorg.conf used.
spstarr_home: no kms
airlied: cool.
spstarr_home: http://www.sh0n.net/spstarr/Xorg.0.log
terracon: to kms or not to kms. That is the question
terracon: ./topic
spstarr_home: airlied: interestingly when i had kms enabled i noticed one line alignment stuff
spstarr_home: Oct 15 20:51:41 segfault kernel: [ 76.272035] mtrr: base(0xe1144000) is not aligned on a size(0x5a4000) boundary
spstarr_home: lemme check if that was with kms on
spstarr_home: no w/o kms
spstarr_home: i dont know if radeon uses mtrrs or not
spstarr_home: common sense is yes
rx__: glisse; p.s. atomtools doesn't segfault in identifying, it segfaults setting the engine clock
spstarr_desk: weee
spstarr_desk: i just wedged the rv350 in gnome ;(
spstarr_desk: I will put it to xaa for now, that was more stable
spstarr: (**) RADEON(0): Using XAA acceleration architecture :(
spstarr: perhaps exa is doing two rop ops at the same time with same memory in some situations
airlied: spstarr_desk: can you try Option "ExaNoComposite" "true" with EXA enabled to see if you get crash still
stefanb: Hmm, radeonhd has R6xx EXA/DRI already?
SirEmo: I don't think so. See topic.
stefanb: SirEmo: Not 3D. radeonhd 1.2.2 release note http://lists.freedesktop.org/archives/xorg/2008-October/039353.html
stefanb: it specifically says EXA and DRI, i.e. 2D acceleration for r6xx too
SirEmo: ah right, I hadn't checked since monday
SirEmo: my humblest apologies
libvde: stefanb: r_S_6xx
libvde: stefanb: important little letter there
libvde: stefanb: and thank ATI marketing for that confusion
stefanb: ahh
stefanb: libvde: thanks for the correction. I guess I need new glasses :-)
stefanb: But at least radeonhd 1.2.3 should no be as fast as -ati already is.
libvde: s/no// yes
stefanb: s/no/now/
libvde: right
stefanb: libvde: OK, I'll give it a spin...
stefanb: DRI still fails on my T60 (RV515/M52)
stefanb: yes, I use Option "DRI" "on"
kifffkiff: hi
kifffkiff: Is there anybody in there
kifffkiff: ?
kifffkiff: clear
adamk: kifffkiff, No, we are all dead.
oga: ARGH. Xpress 200M still doesn't work on openbsd, even with wbinvd() in the gart setup.
oga: could the bus mastering changes recently help?
kifffkiff: hello
spstarr_work: hullo
spstarr_work: airlied: I will add this tonight, and do any other tests
ibrahim: I am using opensource radeon driver on ubuntu 8.04 , I would like to make TV-out and VGA out. How can I do that?
orkid: ibrahim: xrandr can probably do it
ibrahim: orkid, thanks for your reply, do you have an example for that?
chithead: ibrahim: note that tv-out will not work properly up to r300
chithead: ibrahim: note that tv-out will *only* work properly up to r300 (sry)
ibrahim: chithead , I am using mobility radeon X1400 , what is that card series?
ibrahim: Is that compatible or not?
chithead: ibrahim: those are r500, tv-out is not yet supported
orkid: u could try the fglrx driver ibrahim
ibrahim: but I am using xorg1.5 and latest radeon maybe that is included
chithead: ibrahim: if you want to try tv-out, you have to build from source and revert the commit from bug 16186
ibrahim: I am using version number "6.9.0+git20081015.435cf7da-0ubuntu0tormod" How is that ? Is that enough to do that?
chithead: ibrahim: this doesn't look like you built it from source
ibrahim: chithead, yes you are right from repo of xorgsudoers , but that driver is released yesterday if you attention the number? Is that still necessary to compile from source of latest version?
chithead: ibrahim: I will state it more clearly: you need to make changes to the source code before it is compiled
ibrahim: ok that's more clearly now :) I will wait a few more times I think they will publish corrected version without bug then I install it. Thanks man have a nice day
endschranz: Hi, can anyone tell me howto to force X to run at 75 hz?
chithead: endschranz: pass --rate 75 to xrandr
endschranz: I am using ingnoreedid in my xorg.conf so i tell me taht 75 hz is not aviable
adamk: endschranz, You should add a modeline to the monitor section of your xorg.conf file that uses 75 hz.
adamk: Then that mode will be available via xrandr (or can even be selected directly in your xorg.conf file with the PreferredMode option).
endschranz: adamk: I added one but it does seam to work
endschranz: adamk: ok I will try
endschranz: adamk: still at 75 hz
adamk: What does that mean? You wanted it at 75 hz, right? And now it's 'still at 75 hz'?
endschranz: adamk: opps still at 60 hz
adamk: Well pastebin your /etc/X11/xorg.conf and /var/log/Xorg.0.log file, as well as the output of 'xrandr'
endschranz: adamk: one moment
endschranz: adamk: xorg.conf
endschranz: adamK: xorg.log
endschranz: http://pastebin.com/m272d0144
endschranz: http://pastebin.com/m137fa7a4
endschranz: randr: http://pastebin.com/m4b78cdde
adamk: And it isn't showing up in xrandr? I would think that /var/log/Xorg.0.log would explain why.
endschranz: adamk: xorg.conf http://pastebin.com/m272d0144; xorg.log http://pastebin.com/m137fa7a4; xrandr http://pastebin.com/m4b78cdde
endschranz: adamk: x restart
spstarr_work: downloads Kubuntu to test r3xx
adamk: I wonder if you need to add an monitor option to your device section.
adamk: Option "monitor-DVI-0" "Monitor0"
adamk: So that it binds that monitor to that device.
adamk: I know it's necessary with dualhead setups.
endschranz: x restart
endschranz: adamk: thank you very much, now it works perfectly
adamk: Yay.
adamk: I'm not completely useless :-)
endschranz: adamk: :-)
otaylor: spstarr_desk: Have you tried rawhide recently on your rv350?
spstarr_work: otaylor: yes
spstarr_work: daily, we have issues, lots of 'em
spstarr_work: scattered all over :(
spstarr_work: otaylor: airlied is working on getting AGP support for kms, still in progress (hes might have found something on ebay for an r3xx)
spstarr_work: otaylor: random crashes with DDX (EXA enabled), need to test something airlied mentioned today tonight (in 3 hours)
spstarr_work: otaylor: GLX seems to have a flaw in not handling destroyed windows, but we don't know the cause/trigger, in the DRI
spstarr_work: i just know I cause it to crash repeatedly with KDE and kwin composite enabled
spstarr_work: otaylor: attempting to try Kubuntu today liveCD to confirm the GLX is a problem upstream or some patches in Fedora
otaylor: spstarr_work: hmmm, ok, see random crashes here, looks like it might be drm and not ddx
spstarr_work: otaylor: which kind of crashes?
spstarr_work: right now we have another issue with bufmgr code
spstarr_work: i logged a bug on rh bz
spstarr_work: lemme get #
otaylor: spstarr_work: solid locks of the entire system
spstarr_work: i get that with exa with no xorg.conf
spstarr_work: https://bugzilla.redhat.com/show_bug.cgi?id=467152
spstarr_work: but no composite enabled, no 3D used
spstarr_work: just the DDX + exa
spstarr_work: that bug is composite related however
otaylor: so, airlied thinks that there might be something systematic with kms and agp?
otaylor: seems to work, except for the hangs...
spstarr_work: he had AGP working on the r1xx, r2xx not r3xx yet
spstarr_work: so he went hunting for an r3xx card last night on ebay :)
spstarr_work: if kms is on for me it will lock solid after gdm or kdm loads.
otaylor: ok, that's the lock solid I'm talking about
spstarr_work: ya, i get it too
otaylor: it's a little more random ... it doesn't always lock solid immediately
spstarr_work: nomodeset for now :/
spstarr_work: if i use kdm it sometimes can get further but it will lock pretty soon after
otaylor: I guess I'll ask him if there's anything useful to do for debugging once he's around
spstarr_work: you have which rxxx?
otaylor: (I have an AGP r300 card at home that I'm not using... well, two even. But post to .au is a little slow for the F10 timeline...)
spstarr_work: heh
otaylor: spstarr_work: This is a rv350 in a T42 ... pretty much the same as what you have IIRC
spstarr_work: Radeon 9800 Pro?
spstarr_work: otaylor: thats what I have ya
otaylor: the randomness of the hang makes it unclear how to start debugging it ... it doesn't seem that there is a particular sequence of isntrtuctions that trigger the hang reliably
otaylor: I tried stracing X and got it to hang that way, but that doesn't give you much useful data
spstarr_work: for the kms lockup, i can't attach X even
otaylor: as far as I can tell, my system is completely dead at that point ... if I'm ssh'ed in, thats frozen, etc
otaylor: So maybe the kernel is in an uninterruptable loop somewhere, or maybe the pci bus is wedged
spstarr_work: it might be
airlied: spstarr_desk: can you try Option "ExaNoComposite" "true" with EXA enabled to see if you get crash still
airlied: oops..
airlied: spstarr_desk: ignore that one :)
airlied: otaylor: kms hanging is one of two things, missing command submission things that r300 needs, or bad AGP card init sequence.
logari81: some days ago a user with a X1600 card had no xv with XAA (vlc crashed). After switching to EXA xv worked but his system became very slow, here is his Xorg.0.log
logari81: http://pastebin.ubuntu.com/58518/
logari81: I would be thankful for any advice
agd5f: logari81: if you disable composite, Xv with XAA will work fine. As to EXA, it seems like there may be some issues with EXA on xserver 1.5.x
burberryjam: hello there
otaylor: airlied: wouldn't a bad AGP card init sequence cause reliable hangs when initing the card?
burberryjam: im running fedora9 on an ati igp 330m for a laptop, and Im trying to a) get compiz fusion/desktop effects to work properly and b) run a dual monitor spanning desktop setup. neither works, could it possibly be a driver issue?
burberryjam: I get cloned output at grub loadup, but once i boot, i lose cloned screen
otaylor: burberryjam: well, have you tried adjusting things from the screen resolution preferences dialog?
spstarr_work: airlied: dont try that option? :)
burberryjam: otaylor yes
otaylor: burberryjam: do all the outputs show up there?
airlied: spstarr_work: no try ExaNoComposite.
spstarr_work: I will
spstarr_work: airlied: thats for the composite stuff or just regular 2D use?
burberryjam: no
otaylor: burberryjam: then, yes, sounds like a driver problem
spstarr_work: since i didnt use composite with EXA (it crashes with buffmgr stuff right now)
burberryjam: got any recommendations on where to update from?
spstarr_work: airlied: you picked up a r3xx ? :)
logari81: agd5f: is there any probable workaround to avoid such performance issues with EXA?
burberryjam: on ati site my gfx card doesnt show up on drivers list, its too old :D
otaylor: burberryjam: well, you can try F10 beta .. though that will likely give you bigger issues at the moment
burberryjam: no doubt
otaylor: burberryjam: the F9 drivers are really pretty much the latest isn't very bleeding edge. I'm not sure updating is worthwhile unless you have particular knowledge of a fix
burberryjam: thx anyway
airlied: spstarr_work: haven't won yet..
agd5f: logari81: we need to figure out why it's slower with 1.5.x
spstarr_work: airlied: is not like people are gonna create a bidding war for that thing :)
otaylor: airlied: anything useful I can do to debug the kms hangs?
burberryjam: i gtg make dinner, bbl thanks for advice otaylor
otaylor: spstarr_work: I managed to lose 2 or 3 auctions trying to complete a r100/r200/r300 colllection, then again, that probably had something to do with being unwilling to pay more than $10 or so for such a card
spstarr_work: otaylor: heh
airlied: otaylor: nothing springs to mind, try ExaNoComposite maybe it might help narrow down which engine is busted.
otaylor: airlied: I've seen some some hangs when X is not running, though they could conceivably not be the *same* hang
otaylor: airlied: it also is a little odd to me that command submission bugs would cause the entire system to hard lock
airlied: otaylor: not really, the PCI bus is a fragile thing..
airlied: otaylor: once the card FIFO fills up the bus gets wedged..
otaylor: So, that's "Option" "ExaNocomposite" in the device section?
airlied: yup.. it might need "true"
otaylor: tries
otaylor: airlied: what changes about the command stream with kms?
airlied: otaylor: we use glisses new submission method, it shouldn't be wildly different from the X server point of view though.
airlied: its a lot more like what fglrx does.
otaylor:
airlied: so the hw flushing may have different ordering etc.
otaylor: airlied: Assuming I got the option setting right, no difference from ExaNoComposite
airlied: otaylor: xorg log will tell you..
airlied: interesting so it dies just using the 2D engine..
otaylor: (just got another (unrelated or not) hang in plymouth without starting X)
airlied: damn r300s...
otaylor: EXA: Disabling Composite operation (RENDER acceleration)
otaylor: so it looks like that option stuck
spstarr_work: hehe airlied :)
airlied: cool.. narrows it down at least.. not the 3D engine.
spstarr_work: otaylor: does kms lockup on reboot? for me it does..
airlied: my r300 laptop won't even stay on for longer than 2 seconds..
airlied: need to find the short inside it...
otaylor: spstarr_work: Not sure ... usually I've been locked up before then :-)
otaylor: can try
spstarr_work: otaylor: when i reboot it hangs, on shutdown sequence, note run with runlevel 3.. not with 5
otaylor: spstarr_work: no, that works fine for me
spstarr_work: !
airlied: spstarr_work: oh I might be able to debug that one :)
spstarr_work: airlied :)
otaylor: (well, this time it worked... can't say if it reliably works)
spstarr_work: airlied: with no kms, kernel shows disabling irq 9 on reboot
spstarr_work: with kms it just hangs heh
spstarr_work: same spot after network manager shuts off
spstarr_work: irq 9 is where the radeon is living on
spstarr_work: otaylor: im surprised this is not crashing for you.
spstarr_work: i have a kubuntu 8.10 CD, im gonna try composite and see if it locks up or not
spstarr_work: if it doesn't im gonna wonder what patch(s) is causing it if it does, its upstream
otaylor: airlied: I think I'm going to try and isolate the hang during plymouth to see if I can get any traction there ... could well be the same thing that is hanging x
spstarr_work: otaylor: your card is 64MB or T42p?
airlied: ponders bringing r300 craptop into office, oh my back..
spstarr_work: the p had i think 128MB
otaylor: spstarr: 64MB (non-p)
spstarr_work: identical one
spstarr_work: otaylor: rawhide?
otaylor: spstarr_work: yeah
airlied: otaylor: how far into plymouth does it hang?
spstarr_work: 2.6.27-13?
otaylor: airlied: random ... usually about 2/3's of the way through, sometimes right near the end
otaylor: somtimes earler
spstarr_work: otaylor: with kms and plymouth it goes fully though for me
spstarr_work: (rhgb grub option)
airlied: otaylor: hmm so not udev hang..
spstarr_work: i do get the udev hang but its rarer now hasn't happened in weeks
otaylor: spstarr_work: yeah 2.6.27-13 yeah
spstarr_work: otaylor: I did a yum update last night then shut down for the evening
otaylor: spstarr_work: it was actually going through fine earlier today but now it is semi-consistently hanging ... if I didn't know it was reliably correlated with modesetting and the f10 upgrade, I'd think hardware problems
spstarr_work: so unless something changed in plymouth you should have it go though fully to gdm at least (then lockup)
spstarr_work: i notice it does not flip screen from plymouth -> gdm, the mouse cursor just appears
airlied: spstarr_work: yup thats most of the reason for doing modesetting :)
spstarr_work: there's a smooth transition from plymouth to gdm
spstarr_work: airlied: thats fscking awsome :)
spstarr_work: at first, i thought there was a bug, maybe the video didnt reset :)
spstarr_work: or at least clear plymouth from screen
spstarr_work: welcome to the world of kms
spstarr_work: airlied: even more funny was when i had pressed escape, and saw the kernel bootup initscripts going.. then the X mouse cursor appears ;)
spstarr_work: that really looked freaky
otaylor: spstarr_work: halfline wants to add a clear when you are in text mode before starting X, since cursor over the text is pretty weird
spstarr_work: yeah
airlied: otaylor: btw I spent a while trying to get that corruption on the t60p yesterday to no avail.
airlied: I wonder did my new DFS fix it.
agd5f: spstarr_work: rv350 AGP works fine here with kms, etc.
otaylor: airlied: once I stop playing with my laptop, I'll reboot this machien and check
airlied: office &
otaylor: agd5f: it's intermittent and flaky enough here that I can easily believe that it mighit not happen at all on other machines
airlied: otaylor: you on laptop or desktop?
otaylor: airlied: laptop
agd5f: otaylor: I believe it. AGP sucks
airlied: hopes agd5f isn't on the same laptop :)
agd5f: I have pretty much the same laptop as spstarr_work
otaylor: agd5f: OK, it's pretty much the same as mine then too
agd5f: thinkpad t41p
otaylor: (from transitivity :-)
otaylor: agd5f: t42 here
airlied: office &
mike-bur1s: I'm looking at the ThinkPad T500s which offer "ATI Mobility Radeon 3650 with 256MB", which seems to have both VGA and DisplayPort connections. Does the radeon driver support all this at WSXGA+ resolution?
agd5f: mike-bur1s: I've heard it works
chithead: mike-bur1s: no, displayport is not supported
agd5f: it should, but I don't have one myself
agd5f: displayport is not yet supported
mike-bur1s: The VGA should work just fine, though?
chithead: t500 is the one with switchable graphics, yes? I think this is still untested
mike-bur1s: Yes, switchable graphics.
mike-bur1s: So if I want to configure a laptop that I'm sure will work today, the T500 with the ATI card is not the way to go?
chithead: http://ajaxxx.livejournal.com/60080.html you will either be able to use the integrated x4500 or the ati but which one is hard to predict
otaylor: mike-bur1s: I think it depends on the amount of adventure in you :-)
otaylor: mike-bur1s: it's going to be *close* to working :-)
mike-bur1s: Hm, *close* to working sounds promising.
xnguard: mike-bur1s: I believe you'll end up with a laptop that, at least right now, won't be able to switch graphics cores until/unless it's booted into Windows.
mike-bur1s: I don't mind not being able to switch, so long as whichever I'm using can drive both the laptop display and an external monitor at a high resolution.
chithead: mike-bur1s: according to the link, you can set the active graphics core in the bios. so yes, everything should work except 3d and displayport with the radeon driver
mike-bur1s: Fantastic.
mike-bur1s: Is there ongoing work on DisplayPort?
chithead: it's still on the todo list afaik
mike-bur1s: Awesome. Thanks a bunch, guys.
chithead: maybe work on it will pick up as displayport is now part of apple's and lenovo's current lineup
mike-bur1s: Nice. This all sounds very promising.
bridgman: re: displayport, what is everyone seeing out there in terms of monitors ?
bridgman: my understanding right now is that there are lots of displayport plugs on PCs and boards but very few displayport plugs on displays... any sign of that changing in "real world user land" yet ?
spstarr: airlied:
spstarr: airlied: I tried Kubuntu live CD, it did *not* crash kwin or composite on/off, changing settings...
spstarr: your option to Xorg isn't listed in log using "true"
spstarr: they are using 2.6.27-4-generic (likely their own patches?)
spstarr: Mesa Mesa 7.2
spstarr: Libdrm 2.3.1
spstarr: hmmmm
spstarr: airlied: unless drm 2.4.0 is regression causing some bug?
airlied: spstarr: what X server?
spstarr: looking now
spstarr: xorg (1:7.4~2ubuntu4)
spstarr: the server is..
spstarr: xorg-server (2:1.5.1-1ubuntu1) intrepid; urgency=low
spstarr: airlied: so its a different libdrm they're using and radeon drm version is..
spstarr: strings radeon.ko
spstarr: ATI Radeon
spstarr: 20080528
spstarr: drm.ko:
spstarr: 20060810?!
spstarr: that can't be right
airlied: spstarr: ignore that..
airlied: its only going to be server/mesa interaction..
spstarr: ok
spstarr: so then they only big difference seems to be libdrm being 2.3.1 vs ours 2.4.x
spstarr: im looking at their patch set..
airlied: shouldn't matter.. it might be harder to hit the race.. ...
spstarr: hmm, well i even vt switched with composite on no lock
spstarr: even in packages.ubuntu.com they have 1.5.2-1 now but libdrm2 (2.3.x)
spstarr: libdrm2_2.3.1-0build1_i386.deb
spstarr: for mesa
airlied: spstarr: libdrm version doesn't matter..
airlied: spstarr: its not involved in the issue you see.
spstarr: is it patches in Xorg server or mesa ?
airlied: I'm not sure what might be causing it.. the race might just be harder to hit..
airlied: theats the problem with race conditions.
spstarr: hmm
spstarr: well, i wont worry about that for now, one step at a time :) i await your kms changes for testing
otaylor: I feel like a witch doctor trying to find out a reproducible dance for the udev crash
airlied: otaylor: chicken sacrifice..
spstarr: that one is terrible to debug
airlied: otaylor: I really wish I coukld make udev leave the kernel msgs on .
Terman: spstarr: talk to the people who finally found the "e1000e"-bug ;)
otaylor: "boot once in runlevel 5, let X hang, pull the battery, make sure that the network is plugged in, reboot"
otaylor: (err, the plymouth hang, which may or may not be related to the udev hang)
spstarr: when udev hangs for me my hd disk is solid
spstarr: if i power it off hard, it will boot up next time ok (though i've had it hang twice in a row once)
spstarr: now, the disk is from 2004... so i know i have to be careful too :)
otaylor: spstarr: the trouble is that when something is really intermittent, it's hard to tell between superstition and something useful
spstarr: heh
spstarr: well at first i thought it was my hard disk beginning the sign of death.. to which it could go
otaylor: (I can't tel if the disk is solid on this machine, since I replaced the hard disk once and got the led connectiosn wrong when I put the new one in...)
spstarr: doh
spstarr: airlied: doesn't mesa call drmFunctions ? is it possible 2.4.0 is passing something bad back into the DRI assuming libdrm 2.3.1 did something different
spstarr: this is like a wild goose chase ;/
airlied: spstarr: libdrm isn't the problem.
airlied: I won't say it again.
spstarr: takes a nap
otaylor: grrr, 20 or 30 reboots later, this hang during boot seems to be basically whether someone has their microwave on in a house a mile away or something
otaylor: no pattern, other than clearly not being coin flip uncorrelated random
otaylor: (I did get it to hang a couple of times booting into runlevel 1, no plymouth, so my tentative conclusion is that the hang during plymouth is the udev hang, and is distinct from the (100% reproducible) hang after starting X)
airlied: hmm my r300 is busted as well.. guess I get to fix that.
EruditeHermit: wow, Xorg 7.4 is noticeably faster
EruditeHermit: it seems all text related stuff is rendered faster
EruditeHermit: firefox scrolling is smoother too
EruditeHermit: on 4 year old hardware, my compiz effects look faster than new Macs and Windows machines
airlied: otaylor: so its pixmap cache only that corruption with kms
airlied: oh I lied..
airlied: I think I just seen it without glyph cache.
airlied: goes and resizes windows lots more.
airlied: hmm it happens without glyph caching.. so its something in the driver pixmap code most likely then.
otaylor: without the glyph cache it might be individual glyphs rather than glyph strings
airlied: no both strings and glyphs..
airlied: its a bit harder to trigger though.
airlied: I've just got xchat window with text in it and resize it a lot
airlied: or pg-up/pg-dn a lot
EruditeHermit: airlied: I see you pushed patches for GEM
airlied: EruditeHermit: yup hopefully they are all wonderful :)
EruditeHermit: thanks for the hard work
airlied: well all that is mostly Intel... when I get radeon going ti'll be more fun.
EruditeHermit: Xorg needed a lot of work and you guys are moving it forward quickly now
EruditeHermit: now if only Sun were more open with openoffice development
airlied: just need to have a stablisiation period sometime soon :)
EruditeHermit: do you know if Gallium has been merged into Mesa yet?
EruditeHermit: I thought I read somewhere that it had
EruditeHermit: memory manager + 3D goodness +DRI2
EruditeHermit: its going to be a fun 6 months
airlied: gallium doesn't have anything to do with memory manager or dri2
spstarr: :-)
EruditeHermit: I know
EruditeHermit: but I thought I saw somewhere that it had been merged to main branch of Mesa3D for possible release with the next mesa3D stack
spstarr: airlied: lots of badness eh with r3xx ;)
EruditeHermit: so that once that is done, 3D will become the priority
EruditeHermit: mem manager, and dri2 will be great for framework so that the basics are all covered
EruditeHermit: randr1.3 too
EruditeHermit: then cool innovative things like GPU switching and 3D improvements can be worked on
EruditeHermit: right now Xorg is still playing catchup with other OSes
EruditeHermit: feature wise
EruditeHermit: but the pace at which it is going, it can really innovate soon
spstarr: airlied: with that exa option, man is X ever slow redrawing 2D
spstarr: but, it hasn't crashed yet
EruditeHermit: spstarr: really? EXA is way faster for me. Are you using the kms version?
EruditeHermit: I downloaded the KDE live CD
EruditeHermit: i'll try it soon
spstarr: no kms a very small xorg.conf
dmb: so when we get a video lockup, is that from a deadlock usually?
dmb: (i'm in an OS class learning)
dmb: hey MostAwesomeDude :D:D
dmb: i is learning gl!
MostAwesomeDude: dmb: :3
MostAwesomeDude: Have fun.
dmb: MostAwesomeDude, actually, i really hate it
dmb: i have to do a presentation on glsl
dmb: i have no clue what that even is
dmb: we have to have our own example of it also :/
spstarr: hello MAD :)
spstarr: or LAD
spstarr: heh
dmb: there should be a glCircle function in gl!
spstarr: glPie? :)
dmb: yes
spstarr: it exists?
dmb: i wish :(
MostAwesomeDude: There's a GLU function for round primitives. :3
MostAwesomeDude: gluSphere(), etc.
dmb: does glu relate to glut?
dmb: i'm guessing no, as glut is only window management
dmb: OpenGL Utility Library
EruditeHermit: does radeon use EXA or XAA by default now?
airlied: still XAA.. should switch EXA by default on certain setups.
EruditeHermit: isn't EXA faster now?
EruditeHermit: still broken in a few places?
spstarr: airlied: any fixes you're working on for r3xx today?
airlied: depends on the X server..
airlied: spstarr: nothing yet, I've got the RHEL work again..
spstarr: im going back to xaa for the moment, i cant even use X right now properly.. X is using 20-50% cpu
airlied: at least we know its the composite code :)
spstarr: hehe
spstarr: yes it seems to be
spstarr: but i thought you have have composite enabled though?
spstarr: and i dont have it enabled right now im still in just plain ole 2D
spstarr: im going to try composite however and see if that does something
airlied: nope composite means two things.
spstarr: oh?
airlied: composite as in compositing windows.
airlied: and composite as in the hook that gets called in the drivers for lots of functions.
EruditeHermit: spstarr: I thought you were using normal radeon 6.9.0?
airlied: mainly for accelerating render.
spstarr: ah
spstarr: tries driver composite if i lockup brb ;)
spstarr: i am now in driver composite mode...
spstarr: it hasn't crashed yet
spstarr: that's interesting.... very interesting
spstarr: this is a record airlied !!!!!
spstarr: is it possible somehow EXA composite could have been causing composite to fail ?
spstarr: tries a VT switch
airlied: spstarr: its a race, stuff may just happen slightly different.
spstarr: i just switched composite texture filter it did not crash!
spstarr: airlied: yes, but this is interesting that we might have a workaround
spstarr: me tries bilinear texturing
airlied: spstarr: its a race..
airlied: spstarr: you might have to try harder but the bug is still there.
spstarr: i just switched it 3 times still trying to get it to crash
spstarr: switched it again
spstarr_desk: hehehe
spstarr_desk: oh yeah i did alright
spstarr_desk: airlied: seems i can reproduce it less with that option
spstarr_desk: ugh what a mess ;(
spstarr_desk: VT switch will wedge gpu also
spstarr_desk: ok well if i can at least use composite just enough thats a starting point
spstarr: let's see how long composite mode will hold without changing settings
spstarr: the VT switching issue will 'go away' when kms is working :)
spstarr: is so thankful for kms
spstarr: accel is fast with default xorg.conf with only adjusting the GARTSize
spstarr: and MigrationHeuristic smart mode
spstarr_desk: attaches gdb to X
spstarr_desk: that might be the buffmgr bug
spstarr: i have turned off translucent window effects not that i think that makes much a difference
spstarr_desk: Program received signal SIGPIPE, Broken pipe.
spstarr_desk: #4 0x081324cf in _XSERVTransWritev ()
spstarr_desk: wtf
spstarr_desk: wonders what symbol this is from
airlied: spstarr_desk: apps exiting give pipe signals.
spstarr_desk: That was... X
airlied: X gets the signal..
spstarr_desk: main -> dispatch -> FlushAllOutput -> Flushclient ->
spstarr_desk: apparently yes
airlied: spstarr_desk: buy a Unix book.
spstarr_desk: but what issued the SIGPIPE?
spstarr_desk: i have X attached to gdb not the apps
airlied: spstarr_desk: buy a Unix book.
spstarr: :)
airlied: man 7 pipe
spstarr: hmm, I see that
spstarr: its not that, I've never seen X get a SIGPIPE signal when debugging the server before. So i don't know who issued it
spstarr: it's moot now
airlied: spstarr_desk: some app exited.
spstarr: it was plasma that did (well, crashed in some way)
spstarr: notes im still in X with composite enabled..just no transparent windows
spstarr: i guess I can claim a small victory even if EXANoComposite masks the bug...
spstarr: and it certainly is masking it
spstarr: nice, its still holding