MrCooper: sannes: https://bugs.freedesktop.org/show_bug.cgi?id=27767
sannes: MrCooper: Thanks, I'll look at it when I get back home :)
anv: I can't "Xorg -configure"
adamk_: Certainly you can give us more information than just that, though.
anv: _XSERVTransSocketUNIXCreateListener: ...SocketCreateListener() failed
anv: _XSERVTransMakeAllCOTSServerListeners: server already running
anv: Fatal server error:
anv: Cannot establish any listening sockets - Make sure an X server isn't already running
anv: Please consult the The X.Org Foundation support
anv: at http://wiki.x.org
anv: for help.
anv: Please also check the log file at "/var/log/Xorg.0.log" for additional information.
anv: (WW) xf86CloseConsole: KDSETMODE failed: Bad file descriptor
anv: (WW) xf86CloseConsole: VT_GETMODE failed: Bad file descriptor
anv: I suppose it relates on selfcompiled radeon-initrd, which earlier stopped to load KMS (under latest Arch kernel)
adamk_: Use a pastebin service from now on.
adamk_: And pastebin /var/log/Xorg.0.log and the output of 'dmesg | grep drm'
anv: drm: http://pastebin.com/5x5XkANF
adamk_: Nothing odd in your dmesg output.
adamk_: That first pastebin is actually your /var/log/Xorg.0.log file and not just the output you get on your console?
anv: Xorg.0.log.old : http://pastebin.com/vaqFh4nb
adamk_: Well your log file shows this as the fatal error:
adamk_: Cannot establish any listening sockets - Make sure an X server isn't already running
adamk_: Is X already running?
anv: now it is, If I do Xorg -configure without X running then it hangs leaving cursor on top left corner
anv: and I have to reset
anv: but as told earlier I suppose it relates on that selfcompiled module "radeon-initrd" which I installed couple of kernels ago and now it is not needed but messed the system, after I did update in radeon-initrd after last kernel update it let me to run KMS but even so I can't do Xorg -configure, can I just remove radeon-initrd ?
adamk_: I have no idea what radeon-initrd is.
adamk_: But you can't run 'Xorg -configure' while X is running.
adamk_: Does X start up properly with an xorg.conf file?
anv: same here, it somehow helps loading radeon module in bootup
anv: yes it does
adamk_: Is there some issue you are running 'Xorg -configure' then?
anv: Xorg -configure without X running then it hangs leaving cursor on top
adamk_: Right, I get that.
adamk_: But if X starts up with an xorg.conf file, is there some reason you are running 'Xorg -configure' ?
anv: system just totally hangs
anv: some tools need it
anv: and because it has worked earlier it is no use just to quit using xorg.conf because something is broken but instead to repair system
adamk_: Well to see why it's hanging, we'd need to see the /var/log/Xorg.0.log file from when you run 'Xorg -configure'
adamk_: I didn't say it was. But, quite frankly, I don't know what is screwed up with your system, so I'm trying to ask questions to find out what's going on.
adamk_: I never suggested running without an xorg.conf file was a long term solution.
adamk_: When you run 'Xorg -configure' when X isn't already running, and the sysytem appears to hang, are you able to remotely access that machine via ssh?
anv: not needed knowledge
anv: I could try to get that log with ssh if removal of radeon-initrd won't help
adamk_: "not needed knowledge" makes no sense to me. Knowing if you can ssh into the box is certainly something most people would need to know to help you.
anv: ok give some link how to do it
adamk_: Setup an ssh server. I do not know the proper way to do that in arch.
anv: ok thanks anyway
adamk_: Just wow.
[Enrico]: really :/
adamk_: I may not be the friendliest person in the channel, and perhaps give out the wrong advise now and then, but I didn't think I was particularly off-base with this :-)
TGEN: you were not at fault here
anv: KMS set off I can "Xorg -configure"
anv: so I do it with another card sameway and set KMS on on next boot.
anv: it worked
agd5f: mjg59: is there a way to get ac/dc and battery events in the kernel?
mjg59: agd5f: Yes
agd5f: mjg59: any pointers?
mjg59: Huh. Now, where did that patch go...
mjg59: agd5f: There's an acpi notifier call chain that was added for fglrx
adamk_: anv: Please explain exactly what you did.
mjg59: Then you can call power_system_is_supplied() or something like that
agd5f: mjg59: 68b92b567c0c5e6f4d0b264d438f97
agd5f: I just wasn't sure how to use it
mjg59: I think you need to register for the acpi notifier
mjg59: Let me look
mjg59: mjg59: Ok. Get a struct notifier_block, then call register_acpi_notifier on it
mjg59: agd5f: That'll give you events. When you get one, check that the device_class on the event matches a power supply
agd5f: mjg59: ok
anv: adamk_: i turned of KMS, did Xorg -configure, enabled KMS, boot X on
edwin: glisse: GPU lockup after resume
edwin: glisse: and it doesn't recover
anv: evereything works perfectly, only thing is if I need to do Xorg -configure, I can't have KMS turned on, fine for me.
edwin: ^and it keeps looping there, i.e. GPU reset+lockup+reset...
glisse: edwin: something keep locking up the GPU
glisse: the reset is working
edwin: glisse: I see briefly the terminal
edwin: then it all goes black again
edwin: then I see the terminal again
glisse: edwin: ssh in and do init 3
glisse: you will have working terminal
glisse: reset is definitly working
glisse: the log says so
edwin: 3 you mean multiuser + nox?
edwin: I think thats 1 on Debian
glisse: 3 is no X
glisse: with multiuser
edwin: ok I'll try to reproduce the lockup (already rebooted)
edwin: glisse: ok that worked (stopping X)
edwin: glisse: its easy to reproduce: boot, login to KDE, sudo s2ram -f
edwin: press power to resume -> GPU lockup
edwin: its drm-radeon-testing before the pm stuff was committed
edwin: Author: Christian Koenig
edwin: Date: Sat Apr 10 03:13:16 2010 +0200
edwin: ^this commit
edwin: glisse: starting X worked
edwin: glisse: is there any information I could collect to help you see what causes the lockup during the resume?
agd5f: mjg59: how do I get from an acpi_bus_event to a device?
glisse: edwin: try suspending a stock Xorg ie from init 3 do Xorg -retro &
glisse: then suspend
glisse: and resume
agd5f: mjg59: this is what I've got so far: http://pastebin.com/K7v1rvyu
edwin: glisse: what is -retro?
glisse: just to have something on the screen
glisse: it's the old Xorg default patern
edwin: [TTM] Buffer eviction failed
edwin: Memory corruption detected in low memory
edwin: and the X pattern is completely corrupted
edwin: glisse: it locks up during suspend!
edwin: glisse: copy+pasting the dmesg now
edwin: glisse: though it doesn't lockup during resume now
edwin: glisse: http://paste.debian.net/72160/
edwin: glisse: it doesn't lockup after resume with X -retro (only during suspend). It does lockup with KDE though during resume too
glisse: so KDE is triggering a lockup
edwin: well with X -retro it locks up during suspend
edwin: is that normal?
glisse: no the trace is interesting thought
glisse: i need to go over the code
edwin: I'm using mesa and xf86-driver-ati master if that matters
edwin: the card is a rv730 pro
askhader: Hey, I have a Radeon 7000 and I am attempting to do dualscreens. X11 detects both monitors and even gets the dual screen working (as in I can move my mouse across both screens) but the desktop becomes entirely garbled, any ideas on how to fix this>?
agd5f: mjg59: nevermind, I think I figured it out
mjg59: agd5f: Oh, sorry, managed to miss your earlier paste
agd5f: mjg59: something like this should work right? http://pastebin.com/vqbyzJZy
mjg59: Yeah, then either toggle off the event state or (if you want to be more paranoid) call power_supply_is_system_supplied()
agd5f: mjg59: I suppose there's not generic method to support non-acpi systems as well
mjg59: agd5f: power_supply_is_system_supplied() is generic
mjg59: But will probably never give you a useful answer unless you're embedded or ACPI
agd5f: yeah, and you need an event or your stuck with polling
mjg59: The notifier could actually have been implemented at the power supply level
mjg59: With hindsight I should have encouraged that
agd5f: mjg59: so power_supply_is_system_supplied returns 0 for battery and >0 for non-battery?
agd5f: mjg59: thanks! works
DanaG: hmm, what is a "Mobility Radeon HD 530v"?
evil_core: hi all
TheBrayn: howdy folks
RCX_Slinger: Any developer here that happen to know if VBO support will improve? (am I right in thinking VBO calls are currently just translated to VA calls?)
TheBrayn: Enable modesetting on radeon by default - NEW DRIVER <- is it wise to enable this in the kernel?
TheBrayn: I'm using bleeding edge drivers from the gentoo x11 overlay
TGEN: twnqx: unfortunately updating pixman didn't help
TGEN: twnqx: somehow dbus_malloc() calls _exit() which kills xfce4-session, even though there's no call to _exit() in the code for dbus_malloc()... :)
TGEN: DanaG: iirc, some rv7xx
DanaG: Ah. Then that's misleading... 5xxx implies r8xx.
TGEN: that's prolly the effect they're going for
TGEN: it's the nature of the beast, to rebrand
TGEN: lists DirectX 10.1 support, 55nm, and 80 shaders
TGEN: so it's rv710
RCX_Slinger: ...by VBO I meant "Vertex Buffer Object", which should store data on the graphics memory. But right now, radeon drivers completely lacks graphics memory management, right?
DanaG: ah: " Two independent display controllers "
TGEN: one LVDS, the other for external DVI/HDMI/VGA I reckon
mgottschlag: RCX_Slinger: didn't that come with dri2?
RCX_Slinger: Perhaps, but testing a demo (http://nehe.gamedev.net/data/lessons/linuxsdl/lesson45.tar.gz) which can render both VBO and VA, the fps is exactly the same, which seems strange.
RCX_Slinger: btw, my card is AGP (hd4670) and even though I get 3d acceleration, I've never been able to get kms to work. So perhaps its related to my specific hardware?
odin_: does this channel related to drivers for RV100 hardware, such as ATI Technologies Inc Radeon RV100 QY [Radeon 7000/VE] ? does this channel relate to drivers for ATI FireMV 2250 PCIe drivers ? (Im not sure of they are "HD" or not, they are not marked by ATI as HD)
RCX_Slinger: ok, looking at xorg log reveals dri2 not working. And if dri2 is needed for vbo, then this gives a start. I will try to get kms+dri2 working and see if that might solve it. Thanks!
adamk_: odin_: As long as you are using the open source drivers for those GPUs under Xorg, then this is your channel.
adamk_: rnoland: FYI, I'm almost done updating -CURRENT again. Not sure if you need me to test something.
Obscene_CNN: I just posted a patch to mesa on the dri-devel mailing list and I was wondering I could get some people to test it? http://pastebin.com/2C1rL76S
Obscene_CNN: it should work on all cards
DanaG: wow, that resize-bench thing from the mailing list is amusing.
ajax: you're welcome
AndrewR: Obscene_CNN, guess my rv280 below "all" mark .. it has some vertex shaders, and ati_fragment_shader, but it is all, and i'm not sure if i have anything to test exactly those ...
Obscene_CNN: do you have anything 3d? I don't think glxgears will show it
AndrewR: Obscene_CNN, usual array of mesa demos/tes progs/ 3dmark2001 under wine (not sure if it still working)
Obscene_CNN: okay thanks. Also if you see any more bug than normal that would be helpful too.
AndrewR: for some reason spectex started to work after some time. something really weird happens here, sometimes spectex demo need additional kick from UMS, but for now it worked even without ....I'll try in few mins.
TheBrayn: http://codepad.org/JWrdHJ9Z how can I fix those errors
agd5f: TheBrayn: nothing to worry about
TheBrayn: well I downloaded hedgewars (a worms clone)
TheBrayn: and the fps are low like hell
TheBrayn: I didn't test it with before the kernel update though
TheBrayn: radeon's opengl support isn't that good eh?
AndrewR: Obscene_CNN, git apply dislike this patch ... (as downloaded from pastebin). it complains about @error: while searching for: READ, WRITE, FLOW, END " (at newlines )
AndrewR: ", not @
Obscene_CNN: hmmm.... strange.
Obscene_CNN: I made the patch with git
Obscene_CNN: maybe I screwed it up putting it on pastebin
Obscene_CNN: one moment
AndrewR: patch from ML applied OK
agd5f: AndrewR: patch should work, if you want to test. pastebin tends to mangle
Obscene_CNN: try http://filebin.ca/vmyvxr
Obscene_CNN: okay, thanks
Obscene_CNN: goes to check solar battery charger circuit..... brb
AndrewR: Obscene_CNN, no additional ill effects were observed. same speed, same picture. But i guess my wine just old (1.1.35) and doesn't play well with r200-class hw (r200VertexProgUpdateParams:Params exhausted all the time during Vertex Shader test)
AndrewR: Obscene_CNN, but this is well-known for me message, nothing new here.
Obscene_CNN: okay. Thank you for trying it.
Obscene_CNN: the fact it didn't cause any new problems is good news
AndrewR: arbvp* demos from progs/test works as before ....wait until someone else tested it with more heavy workload, but thank you for working on this stuff ... i can't help here, but it is very nice to see few new faces in developer community
AndrewR: "wined3d: Allow the ARB shader backend to work with GL implementations with low amounts of native parameters." in wine git might be interesting to test .....
AndrewR: for me. but not before i compile it
Obscene_CNN: Well, I have tested Torcs, Nexuiz, Foobillard, Stormbaan Coureur, and Celestia with no ill effects that I can see.
Obscene_CNN: all the ones that I have a way of measuring FPS (that I know about at least) show improvement
AndrewR: agd5f, any interest in debug output from mesa for doom3-compressed-cubemaps-rejected-by-kernel bug 26428?
agd5f: AndrewR: go ahead and attach it to the bug. don't have time to look right now
evil_core: I just found we dont have llvm in repo, it impacts on mesa features/performance?
chithead: presently llvm is only used by llvmpipe
Obscene_CNN: evil_core, what chip do you have?
Dr_Jakob: it is used by the sw vertex shader module
Dr_Jakob: it isn't hooked up to the r300g driver tho.
Obscene_CNN: could I get you to try my patch? http://filebin.ca/vmyvxr
evil_core: so its useless for me?
evil_core: Obscene_CNN: got you any new interesting patches for me?
Obscene_CNN: yes that patch I just posted
evil_core: its reason I am askign ;)
Obscene_CNN: it should work on r300 and r500 too
evil_core: Obscene_CNN: but is it something usable, or next thing giving +1%fps? ;)
Obscene_CNN: it is usable and noticable for me
evil_core: and I dont understand one thing, you dont have shell account?
Obscene_CNN: better than 1%
Obscene_CNN: I get a whole fps in nexuiz.
evil_core: to aggregate all patches in one dir, and everybody should see than whjats new, file dates, etc
Obscene_CNN: and thats a lot when you have only 8fps average
evil_core: in r300c too?
evil_core: now we are forced to go trough phoronic threads or trace #radeon logs to find your patches
Obscene_CNN: I need to get my @ss organized
Obscene_CNN: I'll have to set up an obsessive compulsive git repo :)
evil_core: for me using pastebinit, etc is rude ;)
evil_core: dunno why, I dont like them, especially web interface I must usually dig into
evil_core: I like using nc on both sides, and give links in plain text formats
evil_core: and I never remember, notes my preboius pastebin links ;)
evil_core: Obscene_CNN: so now two of your patches are usable for me?
Obscene_CNN: yep, this last one should give you a much bigger boost than the other one
evil_core: its all I need?
Obscene_CNN: evil core, no try just this patch http://filebin.ca/vmyvxr
evil_core: I know about that one
evil_core: but asked if that are all and only usable patches for r500
Obscene_CNN: there should be some usable stuff in the april 21st patch for r500
Obscene_CNN: not a lot but a little
evil_core: ok, thx
evil_core: Obscene_CNN: are there other hacks except yours?
Obscene_CNN: no just mine
evil_core: radeon_span.c: In function ‘radeon_ptr_4byte’:
evil_core: radeon_span.c:276:3: warning: suggest parentheses around arithmetic in operand of ‘^’
evil_core: radeon_span.c:276:34: error: expected ‘)’ before ‘;’ token
evil_core: radeon_span.c:284:13: error: expected ‘;’ before ‘}’ token
evil_core: gis it your patch, or upstream bug?
Obscene_CNN: it might be the patch. It applied cleanly when I made it.
Obscene_CNN: of course my machine may not have compiled that section
Obscene_CNN: let me take a look
evil_core: its easy to fix by adding radnom parentheses ;)
Obscene_CNN: could you post line 276 from your radeon_span.c
evil_core: offset += ((y ^ x & 0x10) << 5;
evil_core: theres one too much
Obscene_CNN: yep, my mistake
evil_core: so remove the forst?
Obscene_CNN: or add one after the x
Obscene_CNN: that would be better
evil_core: Obscene_CNN: will they optomize UMS/Classic too?
Obscene_CNN: <-- uses only UMS
evil_core: me too :)
evil_core: UMS like Elvis and Punks is not dead!
evil_core: so maybe you are interested in fixing this ;)
evil_core: it causes hardlock when xorg.conf is started with pageflip
Obscene_CNN: unfortunately all I have is an r6xx
evil_core: I wonder if AVIVO_D1GRPH_SECONDARY_SURFACE_ADDRESS shouldnt be exchanged with AVIVO_D2GRPH_PRIMARY_SURFACE_ADDRESS, etc, becaus it would be more like old code logic
Obscene_CNN: try it and see ;)
evil_core: anyway agd5f wrote that part, and he dont want to support UMS
evil_core: +if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RS600)
evil_core: dunno if r600 code is separatred, but maybe its valid also for R600
evil_core: its generally AVIVI code/regs
evil_core: and even evergreen driver lacks half of #DEFINES or simply also share most of r500 registers
evil_core: Obscene_CNN: what res are you using?
evil_core: do you know maybe UXGA is way slower than 1024x768 for me, why 1024x768 usually is perfectly quick in UMS?
Obscene_CNN: sorry, had to step out and make a measurement .
Obscene_CNN: 1440x900 is the res I run at
evil_core: you got better card and 1440 is probably like 1024 for me ;)
Obscene_CNN: could be
evil_core: I know its probaly not lack of pageflip, because disabling buffer copying didnt helped
Obscene_CNN: its been a long time since I tried page flipping
evil_core: lol, I am dumb probably
evil_core: i changed it, but ddx lost patch during upgrades
evil_core: but wads wondering why q3a is smooth, but got tearing
evil_core: I wonder if its really much more smotoh like I think
evil_core: its hard to tell difference if you believe/hope in something ;)
odin_: adamk, thanks for your reply
Obscene_CNN: if you have the patch that only modifies the file prog_optimize.c it should be smoother
odin_: what is the heratige with X11 server from 2005 version 6.8.2 and todays 1.7.6 ?
evil_core: I applied all your 3 patches
odin_: I am trying to understand why the same hardware is performing worse today than it was a few years ago
odin_: while I have reinstalled the entire OS, I don't believe I am using or have enabled any features that should slow it down, basic screen redraws are visible (i.e. I can see the sync line move over the redraw area), is the 6.8.2 version from X.org?
AndrewR: odin_, may be it was XAA?EXA move ....
odin_: does anyone know of a website where x11perf tests are put up, against a variety of hardware and versions of X11 ? is x11perf considered a good metric to compare with ?
Obscene_CNN: x11perf is about as good as your going to get for a 2d benchmark
odin_: so it would be worth while me booting back up in FedoraCore3 from 2005, using 6.8.2 and running x11perf -all and then doing the same with Fedora12 from today using 1.7.6 for a compare ?
odin_: sorry I do know know about XAA and EXE, I have heard of these things but the last time I played with X11 protocol was in 1998
evil_core: it still hangs GPU :/
evil_core: but no the system
odin_: also I am aware my RV100 hardware is an integrated graphics suite, because my workstation is a "enterprise server class" system, thats its target market
evil_core: [ 14169.314] [mi] EQ overflowing. The server is probably stuck in an infinite loop.
evil_core: [ 14169.314]
evil_core: [what it means?
chithead: evil_core: evdev issue
AndrewR: odin_, not sure if all test are needed ... try few of them first. -aa10text -scroll500 -putimage500 -getimage500 and some move/resize/map/create tests if you want ..... usually this is enough
odin_: I believe the graphics uses the system RAM for frame buffer (this was also the case with FedoraCore3 that I was using between 2005 and 2010 using 6.8.2), so I'd like to know what I can expect if I were to purchase an ATI FireMV 2250 PCIe x1 card and put in it ?
evil_core: chithead: wtf?
evil_core: chithead: it happens with r500 pageflip patch
evil_core: mouse still works
evil_core: but GPU doesnt
evil_core: and system generally still works
odin_: the system does not have a PCIe x16 slot, only x4 and x8 both in x8 length slots, so my only option is a PCIe x1 card
odin_: AndrewR, ok thanks Ive noted the command line test info from you
chithead: odin_: there exists a pcie x1 radeon 4350 from club3d, that will probably cost you less. for the price of a firemv 2250 you can buy a mobo with 3 pcie x16 slots
odin_: chithead, thanks but the hardware is a Dell Power Edge 1800 (2x3.2 GHz EM64T CPUs, 6Gb RAM, blah blah) server casis, a new motherboard is not an option
evil_core: kisses PBscene_CNN because q3a now works perfectly smooth in UXGA :*
odin_: is the FireMV 2250 as "good" card ? support ? performance ?
chithead: the firemv should work, it is the lowest end though
chithead: and it costs almost four times as much as the club3d pcie x1
Obscene_CNN: good to hear evil_core. I'm glad thing are running sloothly now
evil_core: but ut2k4 is not ;)
evil_core: will test tc-elite I wanted i for in reality ;)
Obscene_CNN: I guess you can't win them all
odin_: I have no 3d requirement at all, but would like dual-head, I am researching 4350 can find plenty but all pics or the spec indicate x16 slot, which does not fit
Obscene_CNN: oops!! Stormbaan Coureur just hard locked my machine
Obscene_CNN: Now that I think about it it did that to me once before I made patches
chithead: odin_: http://www.club3d.nl/products/products_ending_page_7_with_id.cfm?product_id=192
rogutes: Hello. Is CONFIG_AGP needed by PCIE RV730?
odin_: thanks for the link, is that card well supported for 2D ?
evil_core: its not, but it sound like new OSS/Alsa bug :/
chithead: the 4350 is supported for 2d, 3d and xv acceleration
odin_: what kind of performance loss do I get for having RV100 integrated ? i.e. does the GPU continiously access the RAM ? what must consume a good chunk of the CPUs RAM bandwidth ?
odin_: does X server make use of the extra RAM on the card, for pixmap buffers and backing store and such ?
chithead: the rv100 is a discrete chip with its own memory. if you refer to the es1000, that has no 3d or video overlay part
odin_: ATI Technologies Inc Radeon RV100 QY [Radeon 7000/VE] PCIDS=1028:0183 Dell Device 0183
evil_core: you power button swith cod sucks
evil_core: it breaks OpenGL for me after switching to console
evil_core: hmm..its probably mien patch fault ;)
AndrewR: evil_core, it was with UMS? Or those kms powermanagement patches you tested few days ago?
Obscene_CNN: I didn't think my patches had anything to do with that
evil_core: its UMS
Obscene_CNN: time for me to go have a beer and flirt with waitresses :)
AndrewR: evil_core, i remember very funny bug with nvclock + nouveau: i can speed up card, but in exchange i got 2nd cursor and broken VTs .... so, messing with card from user-space not always work as good as we want .... (but nouveau uses KMS only, and nvclock surely doesn't know all details about nvidia). sorry for offtopic
AndrewR: evil_core, if you have any website/blog, you can put some dri1/dri2/gallium numbers for your machine. in q3a, ut2k4, and anything else ...well, you can post all this on phoronix, too.
evil_core: I dont like phoronix, there are too many AMD fanboys, muvh more dum thab iphone fanboys :/
evil_core: yay, warsow rulez now
evil_core: escept bloom
evil_core: and it looks like offset mappings, etc isnt implemented in UMS ;)
evil_core: !stbr th noupgrade warsow-data