EruditeHermit: airlied: its not acpid, it just happened and I don't hvae acpid running
airlied: EruditeHermit: oh wierd. so any ideas what X is doing?
airlied: like does connecting gdb to it or straceing it help?
EruditeHermit: stracing gives errors
EruditeHermit: but I see errors when running normally too
EruditeHermit: let me paste you a snippet
EruditeHermit: but I can't strace during it, only after
EruditeHermit: what happens is
EruditeHermit: usually its a lid event, when I open the lid after having it closed for a while
EruditeHermit: the CPU usage spikes to 100%
EruditeHermit: I can move the mouse cursor
EruditeHermit: but clicking or typing on anything doesn't cause anything to be painted to the screen
airlied: oh I wonder does it crash the GPU
EruditeHermit: I can switch windows
EruditeHermit: or even by clicking the taskbar
airlied: can you do echo t > /proc/sysrq-trigger
EruditeHermit: I can click the gnome menu at the top and it highlights applications places or system depending on what I click
airlied: and see if it wherre Xorg is?
EruditeHermit: permission denied
EruditeHermit: but the menus don't actually get drawn
airlied: as root
EruditeHermit: yes I did it using sudo
airlied: its wierd its like the GPU is okay then
airlied: you need to sudo to bash first
airlied: sudo echo just does the echo as root not the write to the output file
EruditeHermit: the command doesn't output anything
EruditeHermit: it returns to prompt
airlied: and look for Xorg
EruditeHermit: wow dmesg as root gives something totally different
airlied: search for Xorg and pastebin that bit
airlied: but I suspect something else is going wrong.
airlied: strace -p `pidof Xorg` from ssh session ito the machine might also work well
airlied: as root
airlied: further up
airlied: more Xorg stuff I thihnk
EruditeHermit: I grepped for it but that was the only line
airlied: damn too much running.
airlied: try sudo strace -o xorg.out -p `pidof Xorg`
EruditeHermit: VT switch resets it
airlied: and pastebin some of xorg.out
airlied: I think X closes some fds and reopns them on vt switch
airlied: possibly the acpi one
EruditeHermit: I got confused let me clarify something
EruditeHermit: did you want me to do this from ssh?
EruditeHermit: strace -p `pidof Xorg`
airlied: yes if you do it inside X it'll die
EruditeHermit: ok let me find another machine
EruditeHermit: I don't have access to another machine right now
EruditeHermit: can I do it from a VT?
airlied: nope as X is already switched then
airlied: you could do the strace -o if you can type in X
airlied: that'll output it to a file, so you could leave it running and then kill it aftera few seconds
EruditeHermit: to clarify again
EruditeHermit: do you want me to do this when the machine is "hanging"
airlied: yes pretty much
EruditeHermit: oh I see
EruditeHermit: well I need to wait for that to happen again
EruditeHermit: let me write all these commands down
EruditeHermit: and then try to make it do it again
EruditeHermit: then I can ssh in and debug
EruditeHermit: sorry, for my confusion
EruditeHermit: thanks again for your help
EruditeHermit: I know I must frustrate you =p
airlied: hehe no worries we might have to open a bug so I can keep better track
EruditeHermit: yes let me try these commands now
EruditeHermit: I would have filed one but I want to know how to reproduce it
EruditeHermit: before I do
EruditeHermit: or even what log files to give you
EruditeHermit: since it is not showing up easily
EruditeHermit: it always happens when I don't want it to
EruditeHermit: now that I am trying to make it happen, it isn't
EruditeHermit: Dell Studio XPS looks nice
EruditeHermit: display port/VGA/HDMI
EruditeHermit: all 3
rhodan: hi. does anybody know if the exa-acceldfs-small-pixmap bug will be fixed with qt 4.5?
otaylor: rhodan: how would a qt version fix code in the X server?
rhodan: otaylor: it could do the small pixmaps in other ways?
rhodan: or fix the fallback thing
otaylor: rhodan: well, certainly triggering DFS at all is a bad thing to do
otaylor: so if they fixed that it's better in general, but adding workarounds for X server bugs is usually a big mistake, so I hope they didn't do that.
otaylor: rhodan: Anyways, no clue, doubt anyone else woudl have any clue either
ajax: surely bugs in exa are fixed by fixes in exa.
rhodan: otaylor: since when is it a xserver bug?
rhodan: i thought it was a xf86-video-ati bug...
otaylor: rhodan: from the perspective of Qt it's an X server bug
otaylor: I ddn't say a "bug in xserver-1.5.3.tar.gz" :-)
rhodan: so it's a bug in the driver?
rhodan: will it be fixed before gnu hurd 1.0? ;(
otaylor: rhodan: Yeah, or lack of a workaround in the driver for a bug in the hardware perhaps
otaylor: (which is more or less the same thing, but harder to figure out)
rhodan: ok, but this bug is known for some time now.
rhodan: when can i expect it to be fixed? with dri2 perhaps?
ajax: you have a link to a bug in bugzilla? perhaps a small testcase?
rhodan: hm, i wonder why it isn't in bugzilla. but it's in the manpage
rhodan: but it was mentioned here, too
rhodan: < onestone> MrCooper: software fallback on radeon with accelDFS = ugly icons in qt4 apps
MrCooper: yeah, we already discussed that recently
MrCooper: oh, guess you didn't want to include me in that line ;)
rhodan: MrCooper: is there a bugzilla entry for this?
MrCooper: as I said last time, there's several about AccelDFS issues
rhodan: MrCooper: ok, where can i find the bugzilla page in general? freedesktop.org or where?
rhodan: sry, i don't remember you mentioning bugzilla
MrCooper: bugs.freedesktop.org, xorg, Driver/Radeon
rhodan: has this anything to do with dri so there are chances for it being fixed automagically with dri2?
rhodan: sry if this is a dumb question
MrCooper: it has nothing to do with DRI2
rhodan: are there any alternatives to acceldfs? or will it be replaced with something different which provides speedups=
MrCooper: there may be other ways to implement the same thing, e.g. with GEM/TTM
MrCooper: btw, has someone tried to characterize what 'small' means exactly?
MrCooper: one simple workaround might be to use direct CPU reads for 'small' DownloadFromScreen operations
ajax: it's still not clear to me whether we're discussing a performance or a correctness issue.
MrCooper: AccelDFS corrupts in some cases on some setups
ajax: ah, lovely.
ajax: would be pleasant to avoid software though
agd5f: rhodan: you could always disable acceldfs
rhodan: agd5f: slow ;(
MrCooper: not sure, but for small operations the direct read might actually be faster...
ajax: i suspect it is, if that's the only thing you're measuring.
ajax: but, in some magical world where we could pipeline getimage requests and just sleep the clients until the dma completes, i'd rather not stall the gpu even if it would make -getimage10 performance better
rhodan_: sry could you please repeat what you said? i was having connection problems
agd5f: rhodan_: does this patch help? http://www.botchco.com/alex/xorg/fixdfs.diff
MrCooper: agd5f: did you pull those numbers out of a magic hat or somewhere else? :)
agd5f: MrCooper: magic hat :)
Erektium: does r430 support tv-out
ajax: hmm. now that i mention it, i wonder how hard it would be to change the GetImage calling convention to allow async completion.
ajax: that would be hot.
agd5f: Erektium: not yet
Erektium: ok thanks
spstarr: agd5f: UFS/DFS seems to be an expensive operation
ajax: of course it's expensive. it's moving pixels across a bus that's both contended and lower bandwidth than the memory pools on either end.
ajax: and in the download direction, you have to flush rendering and then interact with the memory controller in funny ways
spstarr: we can't just upload damaged areas ? or does it do the full screen?
rhodan: fixdfs.diff didn't fix dfs
rhodan: in fact, if i click something, my graphic systems hangs and the cursor is drawn with ~5fps
ajax: spstarr: we do upload just the damaged areas.
spstarr: ok, i thought that would be the case
rhodan: i don't remember who posted that patch in here...
MrCooper: agd5f did
agd5f: rhodan: try this patch (http://www.botchco.com/alex/xorg/write_sc_after_idle.diff) along with the the fixdfs.diff
rhodan: agd5f: compiling
rhodan: agd5f: didn't fix it. still locks up xorg pretty much
agd5f: rhodan: anything in dmesg when it locks?
rhodan: agd5f: agd5f it doesn't "hang". the cursor still moves, but is drawn at a low frame rate.
rhodan: agd5f: nothing in dmesg
agd5f: rhodan: the GPU has hung
rhodan: hmm k
agd5f: rhodan: what chip are you using?
rhodan: agd5f: r500
rhodan: has this dmesg anything to do with it? it occurs on x startup
agd5f: rhodan: are you on fedora 10?
rhodan: agd5f: gentoo
rhodan: no kms
agd5f: rhodan: does it still hang with if you disable renderaccel? Option "RenderAccel" "FALSE"
rhodan: agd5f: i'll try
rhodan: but what's the purpose of disabling it?
rhodan: doesn't this slow down everything?
agd5f: rhodan: yeah, but I want to know what's causing the lock up
rhodan: agd5f: k...
rhodan: still lockup
rnoland: agd5f: any news on the release yet?
rhodan: sry, lockups again
rhodan: could you please repeat what has been said since the tesint of the patches withoud RenderAccel?
agd5f: rhodan: you didn't miss anything
agd5f: rhodan: try this patch (http://www.botchco.com/alex/xorg/always_flush_IB_before_sw.diff) along with fixdfs.diff
agd5f: rnoland: not yet
rhodan: i'm tired of lockups ;(
rhodan: agd5f: did you fail-tab or did you actually mean rnoland?
rhodan: RenderAccel disabled again?
agd5f: rhodan: I meant rnoland
rhodan: agd5f: OK
agd5f: rhodan: you can leave renderaccel on
rnoland: agd5f: *sigh*, ok thanks...
rhodan: if i don't report back in 90s it's still locking up
rhodan: still lock
rhodan: agd5f: how many patches do you have for me? should i go grab some coffee?
agd5f: rhodan: tha's it for now, I'll have to dig into it more later
rhodan: agd5f: ok :')
rhodan: thanks for what you did until now, whatever you did ;)
rhodan: i don't really understand xorg, as you may already have noticed
agd5f: rhodan: no worries. thanks for testing
xnguard: With xorg.conf being ignored now, what's the recommended way to: a) increase the number of mwheel scroll lines and b) disable DPMS?
airlied: use xorg.conf
airlied: its not ignored.
xnguard: I was under the impression that at least some sections of it were, esp. input devices, in favour of autodetection.
airlied: its not ignored, its just not required
xnguard: Aha. Thanks.
airlied: garnted I'm not sure about mwheel scroll lines where that goes
Dephisto: hi, need help, i have xubuntu, notebook imb thinkpad A30, and I 'd like to run 3D acceleration, i found this
Dephisto: do not tell me that you are all sleeping
airlied: what GPU is in an A30?
Dephisto: how can i find it out?
Dephisto: i am beginner.. in linux :-(
Dephisto: you meant cpu?
airlied: lspci |gerp VGA
airlied: lspci | grep VGA
Dephisto: could i copy it here, or to pastebin
airlied: that should have 3D support out of the box.
Dephisto: out of the box ?
Dephisto: what does mean?
airlied: yes the distro should install 3D.
airlied: no need to install any other driver
Dephisto: yes, but my distro - xubuntu
Dephisto: install in the way, that it works only through cpu,
Dephisto: games, and others, are not playable
airlied: can you pastebin glxinfo?
Dephisto: now i have low-graphic mode, because i tried the https://help.ubuntu.com/community/RadeonDriver and did not succeed
enouf: Dephisto: you got very lucky ... you're dealing with a developer .. count your blessings
enouf: (s/he was in debian asking for help about xubuntu)
Dephisto: yes and they sent me here
Dephisto: because on ubuntu is no one to help me
Dephisto: or i did not find them...
airlied: okay pastebin /var/log/Xorg.0.log
Dephisto: i think that distro like xubuntu, ubuntu... is for beginners, few clicks and it works, so people on it are not skill
airlied: sounds like you haven't enough video RAM
Dephisto: maybe it is because of my big resolution - 1400x1050
airlied: paste the log and I'll read it,
airlied: or pastebin rather :)
Dephisto: i used cat /var/log/Xorg.0.log
Dephisto: but i can not copy all information
Dephisto: i do not have it in the terminal
Dephisto: is it because of the low-graphic mode?
Dephisto: thats all i could get
airlied: oh wierd you are using the vesa driver
airlied: sounds like UBuntus failsafe mode
airlied: change xorg.conf to use radeon again.
Dephisto: i do not know what happened, but i could not send you any message
Dephisto: the last think i got was about using versa driver, or something like that
airlied: use radeon instead of vesa in /etc/X11/xorg.conf
airlied: ubuntu has a failsafe mode that make things really hard to debug.
Dephisto: could i copy jou the xorg.conf?
airlied: I don't have one
Dephisto: because i do not know where to change the vesa
Dephisto: there is nothing about that
airlied: damn its some ubunut specific stuff then, /etc/X11/xorg.conf should have a driver section,
airlied: I gotta run, but iits likely the xorg.conf is broken somehow and failsafe is kicking .
Dephisto: this is how it looks
Dephisto: so i try to change ati for radeon
Dephisto: in device
airlied: gmm ati should be fine
Dephisto: oh so not
airlied: so its ending up in the failsafe mode anyways.
airlied: really has to go.
Dephisto: if you need go
Dephisto: i could come later
Dephisto: or i could write on forum, but do not know whichone
Dephisto: good night