uv: airlied, may I bother you for a few moments? (programming, kms stuff)
gustaf1: can't say i'm too happy with trying to help out when restarting X fucks up so completely. so seriously, is there ANY way to shutdown X and start it up again without loosing ALL acceleration?
gustaf1: currently I log out, Ctrl+Alt+F1 to a console, /etc/init.d/gdm stop, update the driver and then /etc/init.d/gdm restart. this leaves an X so useless I have to reboot.
gustaf1: and yes, sometimes I've tried to rmmod drm and radeon and it doesn't help at all.
gustaf1: I just realized, it's not just X that runs _extremely slow_ after having restarted it. somehow the radeon driver (I guess) fucks up the entire computer and _everything_ runs slow. a shutdown takes minutes, just to stop all the services is like I had a 20 year old computer. but there's not much i/o.. I wonder what the hell is going on with this driver..
MostAwesomeDude: gustaf1: Honestly, I've never heard of such problems. What HW do you have?
gustaf1: I'm trying to git bisect what one of my monitors (in a dual head setup) is randomly flickering. but for each git operation hence driver update, I have to reboot.
MostAwesomeDude: No, the motherboard, etc.
gustaf1: I get transparency issues too after a driver update
gustaf1: I run metacity composited and the shadows gets pure black
MostAwesomeDude: It's just, well, it sounds like HW issues of some sort.
gustaf1: ASUS P5B or W or so.. pretty standard
gustaf1: SW raid
MostAwesomeDude: Huh. Doesn't sound like anything esoteric.
gustaf1: but it only happens when X has been restarted. I really dare not even Ctrl+Alt+F1 unless I really have to. usually I can't Alt+F7/8 back, because X crashes and when starting it up again, it's the same thing. no accel and everything is _extremely slow_
gustaf1: well, I do have cursor corruption (but I heard some of you developers have seen that before, so I'm not sure it's really broken HW)
MostAwesomeDude: Hm. Yeah, a lot of people have been reportiing cursor corruption.
gustaf1: but all of you have no problem with switching to/from console or restarting X?
gustaf1: I haven't been able to do it for a loong time... I do use debian sid with experimental X (to get 7.4 / 1.5) but I can't see how this would give my results..
MostAwesomeDude: Nope, I do it just fine.
MostAwesomeDude: Debian lenny with experimental X.
gustaf1: I use drm from git too. perhaps I shouldn't
gustaf1: well, time for reboot...
gustaf1: wow. this time it worked.
gustaf1: probably because I rmmod'd radeon drm
gustaf1: is this necessary? seems like insanity to me.
scarabeus: hi, i have tiny problem with drm from git and 2.6.28
scarabeus: i get from kernel: unknown symbol i2c_bit_add_bus
scarabeus: when modprobing
scarabeus: i am missing something in kernel config or is that some mixup somewhere else?
arekm: scarabeus: module i2c_algo_bit
scarabeus: compiling :]
scarabeus: thanks :]
scarabeus: it is working, great and really thanks for kick, i just slipped this option :]
arekm: this means that drm has bug in Kconfig dependencies
MostAwesomeDude: No, it's not a bug. You're using DRM from git.
MostAwesomeDude: Mix'n'match doesn't quite get the right dependencies because it's out-of-tree. An in-tree DRM works fine.
arekm: what's the problem with adding required dep to drm git? i2c_algo_bit exists like forever
MostAwesomeDude: arekm: IIRC you should be building git DRM with OS_HAS_GEM=1, which will set everything for you.
arekm: btw. which drm/ati-driver to test gem and modesetting stuff for r300? (if gem for r300 exists at all)
MostAwesomeDude: modesetting-gem is the branch. There's corresponding branches for other repos; check airlied's blog.
arekm: ok, googling (my gprs is becoming "hot" ;-)
arekm: got it
arekm: and checked out 8-) gprs is damn slow. It looks like I need to enable more stuff since many unresolved symbols found when trying to load radeon
arekm: looks like radeon-gem-cs branch won't fly with upcoming xserver 1.6
arekm: drmmode_display.c:249: error: too many arguments to function ‘xf86CrtcRotate’
scarabeus: well modesetting-gem and live xorg server works like charm for me now
scarabeus: X.Org X Server 220.127.116.11
scarabeus: with this xorg server :]
mlankhorst: When will specs for my hd 4850 card be available?
mlankhorst: has bought one half a year ago because specs were going to be open soon :/
spstarr: airlied: welcome back, happy holidays :)
spstarr: airlied: baby delivered? not yet?
airlied: no baby yet
uv: airlied, hi, do you have a few moments?
airlied: uv: ask away, if I don't answer my gf woke up :)
uv: Regarding the mail I sent you about an uninterruptible sleep, please ignore it, I think it was a compiler bug.
uv: About KMS on R635, I have it basically working, and showing a jpeg.
airlied: uv: cool
uv: however, there is a bug with the screen width. I'm setting to a 1680x1050 mode, and the right 16 pixels and the left 16 pixels are identical
uv: I.e. the buffer only has (1680-16)*4 bytes for each row
airlied: sounds like a rounding bug
uv: Am I to expect stuff to work on r635? (should I report bugs?)
uv: I don't need 2D/3D acceleration or Xv
airlied: well I haven't really tested it on r6xx in a while
airlied: I ported the code but it probably does need testing
airlied: uv: did you allocate the buffer large enough btw?
uv: I can do the testing, and user space programming. I also did some kernel programming in the past, but that would require some mentoring
mlankhorst: feels like talking to a brick wall
uv: mode->hdisplay * mode->vdisplay * 4
airlied: uv: you may need to align the width up
uv: to what function? gem_create?
airlied: uv: see radeon_fb.c in the kms code, where it calcs the fb size
airlied: yes the BO size
airlied: also the pitch for addfb
airlied: mlankhorst: nobody knows.
airlied: or at least anyone that knows can't say.
mlankhorst: airlied: figures :)
uv: all rounding up in radeon_fb.c is commented with "#if 0"
uv: or do you mean "radeon_align_pitch"?
EightEleven: My RV250 Radeon card+LCD screen are working okay on 1400x1050, but the screen stays black on lower resolutions. How can I fix this?
airlied: uv: the align pitch one
uv: ok, I'll try it early tomorrow.
uv: In a different matter, why are intel and radeon drm ioctls duplicated (with exact same prototype, but different ioctl number)? Can they be unified?
airlied: uv: which ioctls? GEM ones?
airlied: they don't take the same arguments
airlied: some of them might but not all of them will
uv: The pwrite does (although not implemented on radeon) and IIRC mmap does too
airlied: pread/pwrite may not happen on radeon at all
airlied: the GEM interfaces are driver specific for the most part,
airlied: attempts to make generic interfaces lead to very ugly interfaces
uv: Why not to implement pwrite?
uv: Currently, I mmap a buffer, memcpy to it, and munmap it.
airlied: uv: not sure pwrite will provide any perfomance increase on radeon
uv: aligning up the hdisplay to 16 did the trick
uv: although I'm not sure how (as 1680 is already aligned to 16)
revx: a quick aside here: is there a specific set of messages on the lists I could read to help introduce myself to GEM?
revx: I found Keith's document from May 13 but I don't know how current that is
airlied: revx: thats it really.
airlied: radeon GEM doesn't do as much with domains.
uv: aside: aligning hdisplay up to 16 did not do the trick. Its just that for some reason, the code decided to use 800x600 mode, which works fine.
uv: aligning the pitch in drmModeAddFB to a multiple of 32 (1696) really solved it.
uv: airlied: should I report bugs about r6xx kms/gem? if so, where?
mlankhorst: There is such a thing?
uv: I can send you a "hello, world" program if you like.
airlied: uv: report them against the xorg driver for now in bugs.freedesktop.org
airlied: I need to setup somewherre for bugs against kms.
airlied: make sure you mention its a kms issue in the subject
uv: even though its a kernel thing?
uv: I don't test on X. (the goal is to use wayland)
airlied: uv: yes as its not in the upsteram kernel yet
airlied: and won't be until we finish off some of the TTM bits.
uv: ok, thanks a lot.
mo3sizlak: so airlied have you had any luck w/ your rv350?
mo3sizlak: cus mine still locks my system when i use compiz
mo3sizlak: sometimes it will lock in a weird way where i cant change VT's or ctrl-alt-bksp or do anything, but the screen will remain drawn, and i can move the mouse, but the mouse moves VERY choppily
Erektium: I have the same issue
Erektium: on r430
mo3sizlak: yeah its not a kms bug either cus ive also tried kernel.org 2.6.28...same issue
mo3sizlak: runs VERY well for about 10 minutes or so, then locks
Erektium: well I have this issue with or without compiz
Erektium: and the lock ups are random
mo3sizlak: try using metacity, i bet the lpockups disappear
mo3sizlak: cus it happens w/ kwin for me as well
x1250: Hi guys, I'm getting: OpenGL vendor string: Mesa Project, with a X1400 using the radeon driver. Shouldn't this be working out the box?
x1250: here's xorg log: http://paste.ubuntu.com/93935/
Erektium: using metacity doesn't solve anything
x1250: xorg.conf: http://paste.ubuntu.com/93936/
x1250: glxinfo output: http://paste.ubuntu.com/93938/
airlied: mo3sizlak: on hols so not really looking, you could try a 2.6.26 upstream, I've heard of a regression in there.
airlied: x1250: LIBGL_DEBUG=verbose glxinfo
x1250: airlied, da1l6 fixed it for me on #radeonhd. Thanks :)
Urchlay: so, anyone using tv-out with an older (r280) card?
arekm: tried modesetting-gem and radeon-gem-cs. Just loading radeon with modeset=1 failed with huge oops and black screen
Urchlay: tv-out on r280 card works fine, *except* the right-hand 16 pixels are cut off
Urchlay: anyone have any ideas?
z3ro: Urchlay: I've never used TV out before, but I guess it might have something to do with a timing value.
z3ro: you could try looking to see if there are any adjustible timing values
z3ro: eg in xorg.conf
Urchlay: was trying to use xvidtune to change the timings, but it can't do anything
Urchlay: xrandr can't add modelines either
Urchlay: I can move the picture around a bit with "xrandr --output S-video --set tv_horizontal_position 4"
Urchlay: and there's a tv_horizontal_size as well
Urchlay: tv_horizontal_size can shrink the whole picture down some, so I get a black bar on the right... but the right-most 16 pixels are *not* displayed
z3ro: well, I don't know much about tv out... but you could try filing a bug report about it.
z3ro: I'm just not sure how many developers work with r200 anymore.
Urchlay: actually, where would I file a bug report?
Urchlay: guessing someone who knows what theyre doing, would be able to either fix it easily or say "it can't be fixed"
z3ro: yeah, hopefully someone will be able to help you out (might take some time though, holidays etc)
Urchlay: yah. Other than the missing strip of pixels, the radeon driver works *much* better than the ati proprietary driver I used to use
MostAwesomeDude: airlied: Evening. What's cookin'?
MostAwesomeDude: Almost got a compiling LLVM setup.
MostAwesomeDude: Sooo close.
x1250: I'm getting this message after playing for some time. When this happens, the game becomes very very slow
x1250: Please report at bugzilla.freedesktop.org
x1250: Mesa 7.3-devel implementation error: radeon_program_pair.c::allocate_input_registers(): Don't knowhow to handle inputs 0x8
ssieb: airlied: are you around?
spstarr: he's partying up the last of his weekend :)
z3ro: shit, why didn't I think of that. ;)
spstarr: my sunday begun 52 mins ago
spstarr: which means the end of my weekend is here his monday has begun :(
spstarr: i believe its Monday morning in .au now
z3ro: no, Sunday afternoon according to the all knowing Google. :)
spstarr: is it?!
spstarr: i suppose that could be right given it just turned Sunday here
spstarr: it gets even more screwed up with you have GMT+25
spstarr: z3ro: technically there can be '3' days going on at the SAME time
spstarr: for about 1 hour or so
z3ro: hmm interesting
spstarr: that really confused me
spstarr: its all this stupid daylight saving crap
spstarr: having '3' days going on at the same time is fscked up if you ask me
spstarr: so it means I could be going from Saturday to Monday if I flew by plane from those Islands east of New Zealand
z3ro: yeah, that would be a bit weird.