moondrake: hmm...is this bad: [drm:drm_unlocked_ioctl] *ERROR* ret = 5b -22?
airlied: moondrake: I don't think its good..
airlied: but not sure, depends on what ioctl 5b is.
moondrake_: airlied: it was bad:) machine hung on modeswitch
moondrake_: but still i'm much more happy with modesetting-gem now than master..lets find out which ioctl does this
moondrake_: probably its DRM_RADEON_SURF_FREE
moondrake_: oh, i get 57 as well
airlied: nothing should be calling that if you have kms userspace
airlied: if you are running an old userspace on a kms kernel with modesetting on it won't work
moondrake_: airlied: i do not, but i didn't turn modesetting on
airlied: ah without modesetting it should all work.
airlied: but probably needs more testing.
moondrake_: i'm not so interested in the modesetting, but the gem branch contains some fixes i wanted
moondrake_: i'll see if i can figure out wat calls it
moondrake_: 57 is DRM_RADEON_IRQ_WAIT
airlied: wierd there should be nothing majorly different in that branch that isn't uptream.
airlied: except maybe the ring alignment fixes.
moondrake: airlied: i got some garbled pixmaps i talked about the spstarr today, they seem gone with the modesetting branch
moondrake: airlied: also, xrandr crashes on modeswitching with master
moondrake: at least, with AIGLX enabled
moondrake: there's a bug reported on that
airlied: wierd, i wonder what that is.
moondrake: actually, the garbled pixmaps are not fully gone, maybe it's just less now
moondrake: at least i'm not looking at 2 seconds of rubish when starting a presentation
airlied: thats a bug kwin worked around already in GLX context destruction
airlied: I assume u have desktop effects on
moondrake: otherwise it works fine
moondrake: "in GLX context destruction" means where, mesa?
airlied: yup mesa/server interaction
airlied: but the KDE guys worked around it, at least on create/destroy
airlied: not sure about whats happening on mode change
moondrake: well, this is 4.1.85 now, and it still seems to have it with master
moondrake: although...actually I got the crash before on an older version
moondrake: the newer version just hung up X
moondrake: maybe it is a different problem (or a bad fix)
airlied: yeah it might be an issue still, really need to track down that context destruction issue.
moondrake: the new problem didn't create a log file though
spstarr: airlied: still crashes on VT switching but i logged a bug for them
spstarr: airlied: i wonder if kwin destroys window on VT switch
moondrake: well, the DRM_RADEON_SURF_FREE happens in RADEONChangeSurfaces, without modesetting
damentz: gn8... maybe
moondrake: *ERROR* ret = 57 -4 is a bit of a puzzle, can't find where the ddx waits for irqs
moondrake: according to the debug log, it only happens sometimes though, not always
rhodan_: hi, what do i need to test gem and dri2?
rhodan_: i already have xorg-server, mesa, xf86-video-ati and all other xorg-packages from git-head
rhodan_: oh, and the newest kernel-rc
glisse: rhodan_: testing dri2 is an adventure !
glisse: if you want to test it you need drm+radeon from modesetting-gem branch of drm repo, and my xf86-video-ddx along my mesa branch
mcgreg: hmm perhaps you guys who that ... is there a tool or a possibility how to measure the speed of writing/reading to/from vram?
rhodan_: glisse: ok, link? ;)
rhodan_: glisse: k, i'll wait until it's in git head
rhodan_: glisse: when can i expect that?
glisse: rhodan_: no idea i will be busy with others things for next few weeks
rhodan_: glisse: ok
rhodan_: rhodan_: what things? ;)
rhodan_: oh, i forgot christmas
spstarr: airlied: any stuff for me to test?
airlied: spstarr: nope its all working now :P
spstarr: i dont have agp speeds :)
airlied: you do, you justhave slower fallbacks.
airlied: now I have to go figure out that side of things.
airlied: also make PCI(E) use cached memory
spstarr: airlied: i will test when you start on that
spstarr: airlied: but yes, its fixed
spstarr: no lockups only 3D lockups
spstarr: but im not using compiz/kwin composite right now
HaC|away: hi there
HaC|away: i need some help using free radeon drivers xrandr 1.2 and dualhead configuration
HaC|away: how can i enable the tv output ?
HaC|away: i'm using a radeon x800 pro agp card
spstarr: deletes all these png and oggs for kms corruption :)
airlied: HaC|away: xrandr --addmode S-video 800x600
airlied: xrandr --output S-video --mode 800x600
HaC|away: but xrandr -q doesnt show mit the tv-output...does that matter?
HaC|away: i think so (:
HaC|away: i just do see VGA-0 and DVI-0
airlied: HaC|away: oh thats not right.
HaC|away: hm ?
airlied: hmm I don't think we support it on that card yet.
HaC|away: its a R420 card
airlied: I wonder if you comment out a few lines in the driver would it work.
HaC|away: but i'm really happy about the new xrandr1.2 think...using free radeon drivers and having opengl support
HaC|away: thumps up (:
airlied: HaC|away: if you feel like compiling from git, you could try the atom-tvout branch
HaC|away: yeah i'll try that, thanks so far (:
m03sizlak: airlied do you expect the agp patch in kernel-...-163 to be commited to kernel.org kernel? and if so, what version?
airlied: m03sizlak: its not an upstream bug
m03sizlak: i see. well i tried -163 and found it to make my system incredibly unstable
m03sizlak: comiz may have run a lil better tho, couldnt really tell cus my mouse was skipping like a broken record
airlied: I suspect it slowed things down
airlied: I think I can speed them back up for PCIE/PCI cards.
m03sizlak: yeah i have a agp rv350
airlied: are you running with kms though?
m03sizlak: no, i tried nomodeset first, compiz still locks up pc
m03sizlak: so im using kernel.org kernel
airlied: glisse: ping