Neo_The_User: how do you build r600_blit? or does it build automatically?
Neo_The_User: nevermind. found it in makefile
airlied: hmm drm-next is an oops fest
airlied: not sure what to blame yet
airlied: glisse: I'm blaming the dummy page stuff so far but haven't had time to bisect yet
airlied: I merged your + alex page changes into a big patch along with my oritinal clear page stuff
airlied: agd5f: not sure about those tlb fluhs in set page
airlied: we should be calling that in the tlb flush callback
airlied: sine we can get a lot pf page sets per object
airlied: appears drm-next is okay on my r500
airlied: so its r600 issue
dileX: airlied: drm-next has latest drm-linus and latest drm-radeon-testing?
airlied: dileX: its got drm-core-next which is hte main drm changes + drm-radeon-testing + nouveau + intel
airlied: its based on drm-linus
airlied: I just spent the day moving patches around
dileX: airlied: so it has the pm stuff
airlied: its also seems to have horrible broken on my r600 desktop
airlied: but since I won't back there for a few days it'll probsably remain screwed unless some else bisects
dileX: hmm, OK I see same merge conflicts in -next as I tried drm-linus and then pull drm-radeon-testing on top
dougmencken: airlied, ^^
airlied: dougmencken: hey, so benh is taking a look for me , not sure when he'll get a chance
dougmencken: cool anyway
airlied: dougmencken: so when it boots before X starts you get a console?
dougmencken: btw, I built a kernel with two patches applied...
airlied: dougmencken: oh cool dmesg from that owuld be nice
dougmencken: airlied, runlevel 3 in /etc/inittab
dougmencken: airlied, I can't; maybe infinite loop (no panic, nothing, just silence)
airlied: can you ping it on the network?
dougmencken: or how to get some log from not-fully-booted-kernel
airlied: dougmencken: if you boot with radeon.modeset=0 option
dougmencken: rebooted with old one
airlied: and then when its booted without X running
airlied: rmmod radeon ; modprobe radeon modeset=1
airlied: and see if you get any dmesg
dougmencken: hmm, worth a try :)
glisse: airlied: my patch was working fine on top of radeon-testing
glisse: i tested it with r6xx/r7xx too
airlied: glisse: cool, it might have been my combinging it with Alex's
airlied: I'll reset and just apply mine + yours for now
glisse: i think Alex patch should do the flush only in the gart flush callback
glisse: that should be enough
airlied: yea me too so I''ll drop it for now
glisse: airlied: what kind of oops you got ?
airlied: glisse: random memory corruption ones
airlied: like all over the place randomly once X start
glisse: oh sweet
benh: airlied: monday
benh: airlied: I'm off to syd tomorrow
airlied: benh: cool no worries
benh: airlied: might help to remind me :-)
airlied: hehe... I will
airlied: glisse: only possible issue is allocating the dummy page
airlied: we might need to use same page flags as TTM allocator
airlied: glisse: if you get a chance can you boot the drm-radeon-tsting I t log
airlied: glisse: 82568565683b4991964a5fc89a9ca0c7122818e8 is the HEAD
airlied: it'll take a while to mirror out
twnqx: if i want to pin down the exact command that stalls my gfx card...
twnqx: do i have to break the packet down to one command/packet?
airlied: I'm not in the office until Monday so my r600 is too far away
airlied: glisse: also for AGP we probably should se the GTT page to uncached
twnqx: or is it even possible that the gfx card stalls while the driver does continue to submit packets?
airlied: twnqx: yes its possible
glisse: iirc correctly we allocate the dummy in 32bits so it should be enough everywhere
airlied: the CPU doesn't wait for the GPU the whole time
glisse: the caching shouldn't matter as it's dummy :)
airlied: glisse: its still a cache aliasing issue in theory
airlied: but it shouldn't happen
glisse: well only from the gpu pov
airlied: glisse: another idea might be to periodically confirm that the page is still zeros
airlied: with a debug flag maybe
glisse: yeah we could check that on lockup or with a debug flag
twnqx: hm i seem to remember that there was a version back somewhere in november that workes with KMS on rv635...
dougmencken: nope; looks like it's not video problem at all; it just stops booting after "eth0: Link at up; ..."
airlied: dougmencken: wierd
dougmencken: v2.6.33-rc7-117-ge28cab4 << this version; don't build it :)
glisse: airlied: so you just removed the dummy page stuff ?
glisse: you want me to test it on top of that ?
airlied: glisse: no I merged it wiht my patch
airlied: since your patch just removed most of mine ;-)
airlied: so just test that drm-radeon-testing boots on an r600
airlied: I'll remake drm-next tomorrow, its too much effort to do now
glisse: you are on which r600 ?
twnqx: should it fix the hangs? :>
twnqx: if yes i'd try
airlied: glisse: it was rv635 of some sort
twnqx: airlied: is drm-radeon-testing based on 2.6.32.y?
airlied: twnqx: yes
twnqx: or which base kernel would i need?
airlied: well 2.6.32
airlied: it should pull cleanly into most things
twnqx: 'cause i'm using 2.6.33-rc7 at the moment
airlied: it should pull cleanly in there as well
airlied: since that's what Linus gets
twnqx: checking out the .y branch takes... long
twnqx: especially with 80kbyte/s on
glisse: here lastest drm-radeon-testing lead to massive system ram corruption
twnqx: thanks, NOT testing it
glisse: ok it seems that my cs checker is to blame .... sight
taiu: glisse: case CB_COLOR0_BASE: isnt't tmp used before it's initialized?
glisse: taiu: yeah good catch
glisse: i fixed it at one point
glisse: i need to rm 000*
twnqx: woah, drm-next pulled into 2.6.32.y has merge errors all over Oo
airlied: glisse: cool gimme patches and I'll push em ;-
glisse: it's not enough
glisse: i still have massive corruption
twnqx: so how would i try to build a current drm-next if i get conflicts even in ext4? :X
twnqx: in the pci code... scheduler... wireless lan...
twnqx: i thought drm-next should be like patch to update the drm parts only?
airlied: glisse: thus didn't happen with older version?
glisse: airlied: well my patch was working on another computer
glisse: this is very weird
suokko: airlied: Flags: PMEClk- DSI- D1+ D2- AuxCurrent=0mA PME(D0-,D1-,D2-,D3hot-,D3cold-)
suokko: Status: D0 PME-Enable- DSel=0 DScale=0 PME-
Nightwulf|work: hi all
suokko: airlied: In the whacking code cpi mode never stays enabled. PLL_PWRMGT is all the time 100c0
suokko: or it is all the time acpi mode :) I read that wrong way that 1 would be acpi
glisse: airlied: btw with the cs fix, the dummy gart thingy seems ok
pihhan: what is needed for support of HD5770 card?
pihhan: is there support for at least basic 2D support of card, maybe without xv acceleration?
suokko: pihhan: you need drm-radeon-testing branch
BioTube: pihhan: git master of ddx also supports it for UMS
suokko: There was stil lyesterday some bugs in UMS modesetting
pihhan: suokko, drm-radeon-testing is branch where? it is not name of branch in drm or ati git repository i have here
suokko: It is kernel
pihhan: how big is that drm-2.6.git? i downloaded 60MB and i am at 9%
pihhan: can i run that driver with fedora12 without reuilding all Xserver, mesa, drm, ... etc?
adamk: Well simply math would suggest that drm-2.6.git is going to be roughly 660 megs.
adamk: And, yes, a new kernel shoudl work just fine on F12, I believe.
pihhan: so i need only kernel changed, no other X drivers?
pihhan: are supported two cards, one primary pcie, and one secondary integrated?
pihhan: i have now HD 3200 integrated and HD5770 pcie, and noone does work with free drivers
pihhan: only with vesa driver, but not to native resolution 1650x1050, only 1440xY
pihhan: should i be able to use integrated card, when that card is not primary? log reported something about atombios calls failed
suokko: pihhan: Does integrated work if you don't have discreet pluged in?
glisse: when it's not the primary it likely won't work
pihhan: yes, it did work as primary
pihhan: it did not work with radeonhd driver, but with radeon, it did
agd5f: airlied: I think we should still flush the HDP cache when we update the gart pages since we are writing to vram
agd5f: even if we don't do the full TLB/L2 flush
glisse: agd5f: but isn't it enough to do it once we update all the pages for the current bo ?
glisse: rather than doing it for each page ?
papillon81: do you also get x lockups after startx with drm-radeon-testing?
agd5f: glisse: hmmm, yeah. I'm just wondering if there would be a case where it would be missed
glisse: agd5f: no, the gart flush will happen each time we bind/unbind iirc
agd5f: glisse: I also wonder if we should flush the L2 as well in the gart flush
glisse: btw we have an atombios regression on rs480 :(
glisse: agd5f: would it hurt to do so ? :)
agd5f: glisse: also, r6xx+ requires 8 face alignement for cubemaps
agd5f: glisse: rs480 isn't atom
glisse: oh yeah it's combios
glisse: well we have a regression their ...
agd5f: what's the regression?
glisse: my rs480 was suspend/resume properly before
glisse: now it lockup hard in combios post
agd5f: glisse: weird. haven't touched that code in ages
glisse: yeah i was looking at history and it seems pretty untouched
suokko: glisse: Is there somewhere code in suspend path that forces clocks in KMS?
agd5f: suokko: there's a pll table in combios that sets up the clocks
suokko: I was jsut wondering because my card doesn't got D3 and I was looking what radeonfb has for suspendin
agd5f: radeonfb uses d2 IIRC
glisse: yeah d2 + shutting down most of the things
agd5f: hardcoded paths sorted out by benh
suokko: ok. I was just wondering if our suspend would have to set something too
agd5f: suokko: the tables should take care of it as that's what the bios uses during post
suokko: I also noticed in thinkwiki that many thinkpads with radeons fail to suspend the mobility m6 or m7 cards
suokko: But that was quite old information mostly
suokko: Maybe even from time that there was still fglrx ...
agd5f: suokko: s2ram is probably the biggest PITA of just about anything to support
agd5f: too many weird oem quirks
mjg59: agd5f: pm-utils disables quirks in KMSland
mjg59: agd5f: If s2ram isn't doing that, it's b0rked
agd5f: mjg59: I meant suspend to ram in general, but yeah, lots of people still use old pm-utils with quirks enabled
mjg59: Many of them are basically guaranteed to break KMS
glisse: agd5f: ok it seems to die in dynamic table
eichi: someone has any idea, what i can try, if my system freeze on 3d applications - game. but not on compiz
soreau: eichi: Tried with and without kms?
eichi: soreau: no, only without
eichi: soreau: can this maybe fix this problem?
soreau: Possibly, yes
soreau: eichi: Here without kms, compiz freezes my machine, but 3D apps work ok. With kms, everything works without lockup
eichi: soreau: thanks, i will give it a try
pihhan: i have working framebuffer on juniper
pihhan: but radeon X driver does not work with that kernel
pihhan: can i find somewhere devel branch with X driver also, for HD5xxx?
suokko: pihhan: see the topic
papillon81: i can't start X anymore when using the testing branch of drm
papillon81: this happens since an update and worked until yesterday
suokko: papillon81: dmesg&Xorg.log?
papillon81: suokko: yes. I'll try to debug it
pihhan: i have git repository with xf86-driver-ati
pihhan: but it seems there is no support for PCIID of my card
suokko: pihhan: master branch has evergreen
pihhan: suokko, do you mean git://anongit.freedesktop.org/xorg/driver/xf86-video-ati
pihhan: or is there something else?
suokko: that one yes
papillon81: i get an instant and complete crash, no ssh working after that
twnqx: there was something about major system memory corruption with drm testing earlier today
agd5f: revert the cs checker or apply glisse' patch on the ML IIRC
papillon81: agd5f: i also suspect the cs checker being the problem, however I highly appreciate it, since I got GPU crashes before
papillon81: agd5f: do you have a URL?
agd5f: papillon81: check dri-devel archives from earlier this morning
agd5f: or revert 961fb597c17e2e4f55407d56b7211c188ab41eff
pihhan: suokko, thank you, yes, there it was, now i have fullscreen desktop :-)
papillon81: agd5f: that fixes it. thx :)
papillon81: WTF?!: "You have old and broken userspace. Please consider updating MESA". I updated it some minutes ago....
kdekorte: I just got the latest drm-radeon-testing kernel and it crashes on me with a list_add corruption. I've rebuilt the kernel twice, once with make clean and git clean -df first
kdekorte: Here is the crash: http://pastebin.com/m308c5732
xming: kdekorte: there were some talks about latest git and memory corruption, probably the CS, you need to revert that
eichi: someone knows, how to set virtual screen size in hal without xorg.conf?
ajax: you don't.
ajax: X doesn't read virtual size from hal, period.
kdekorte: xming thanks
legume: papillon81: Maybe you need the latest libdrm and then re build mesa and ati against that.
papillon81: legume: that's what I normally do
Obscene_CNN: xming, I have it benchmarking good now :D
legume: papillon81: OK, that's me out of ideas then, apart from checking you don't have old stuff in /usr and new in /usr/local.
eichi: soreau: tried it, but to hard so get all packages for my distribution
soreau: eichi: Which kernel version and which card do you have?
eichi: ati radeon mobility x1300 and Name : kernel26 Version : 18.104.22.168-1
soreau: eichi: Which distro then?
soreau: I believe arch has all the packages you need
soreau: libdrm, mesa and xf86-video-ati
soreau: If you are able to install libdrm with kms support, and the other two compiled against it, you should have everything you need
soreau: Have you tried booting with radeon.modeset=1 ?
eichi: got an error with mesa-git because of file conflicts
eichi: with freeglut, libgl-git and so on
Obscene_CNN: xming: patches headed your way :)
airlied: pushed glisses fix to d-r-t
twnqx: push him to fix rv635 KMS :X
twnqx: as he admitted to have one himself :P
airlied: my rv635 works fine
airlied: uses it everyday
twnqx: with KMS?
twnqx: wihtout KMS it is fine indeed
twnqx: seems my suspend/resume fail yesterday was related to booting in the docking station
airlied: in fact both the rv635 I have seem to work fine, latop and desktop
twnqx: if i start thunderbird or filezilla with KMS the stall with fade to white happens... at 100% accuracy
airlied: using latest kernel?
kdekorte: twnqx: I use a rv635 in a desktop as well
kdekorte: I'm building the latest git kernel now with the cs fixes
kdekorte: git drm-radeon-testing kernel
twnqx: i tried half of the 2.6.32s and 2.6.33rcs, though
Ganang: can i get help to set up dual monitor with radeon xpress 200m on ubuntu here, or this is not the right place?
kdekorte: Ganang: what driver are you using?
Ganang: kdekorte: thanks... i am using open-source ati...because, fglrx simply crashed my comp....
Ganang: I am using ubunt 9.04...and i user xrandr and it recognizes my output devices...i followed pretty much all the steps on their page, but still something is missing
kdekorte: have you tried gnome-display-properties ?
twnqx: drm-radeon-testing is unmergable for me
Ganang: yes, it recognizes my monitor too...then i select it but, after restart, my monitor is blank, and my laptop-screen is all strechted
twnqx: or i'm too stupid to use git, or the wrong source
kdekorte: twnqx: I just use drm-radeon-testing.. it is a full 2.6.32 kernel
Ganang: kdekorte: this is my xorg.conf file
twnqx: how do you get it? i uses 2.6.32.y and pulled drm-radeon-testing in, like in the doc from topic :X
kdekorte: Ganang: I have you tried moving xorg.conf out of the way and just booting.. I have two displays and no xorg.conf
kdekorte: twnqx: I cloned it and then pull periodically.. sometimes I have merge issues, so I fix it with git clean -df or git reset --hard
kdekorte: Ganang: yeah your xorg.conf is messed up
Ganang: kdekorte: do you mean completely erase it? but before trying to change it, it was pretty clean, and didnt work at all too
kdekorte: Your identifiers don;t match the options...
kdekorte: just rename xorg.conf to xorg.conf.broken
twnqx: kdekorte: i hae merge errors all over the place... ext4, scheduler, wlan, ...
kdekorte: and reboot and then use gnome-display-properties
Ganang: kdekorte: which ones? i just noticed from xrandr that the option for my Xpress 200m "enable" true, makes some difference, because it tries to turn my screen on
Ganang: they say something on their website about it
Ganang: otherwise my screen is completely dead
Ganang: ahhh if this makes some difference, in windows, in order to make my fujitsu siemens monitor work, i had to use the option "sync" , or "composite" ( i am not sure now), inside the ati configuration manager
Ganang: here is what i found in randr website: "If for some reason one output is disabled by the driver (for instance VGA-0 on Radeon Xpress 200 because load detection is not reliable in the driver yet), you might want to enable it by force. To do so, add: Option "Enable" "true" "
dhawk: hello since 2.6.33-rc6 and upwards (drm-radeon-testing) i reliably get fence timeouts resulting in softlockup when running anything more exciting than glxgears ( wine, java3d, dolphin-emu i tested and all crash ) log at http://www.pastebin.org/90308
Obscene_CNN: dhawk KMS or UMS?
twnqx: dhawk gpu chip?
dhawk: twnqx KMS RV770
agd5f: twnqx: just use drm-radeon-testing directly. don't try to merge it
Ganang: kdekorte: didn't work. I am pretty sure, that i have to force sync mode...that is how i got it running on windows xp
Ganang: kdekorte: do you know how can i force it?
dhawk: modprobe log at http://www.pastebin.org/90318
dhawk: anyone know how to fix the problem or workaround it?
moeSizlak: my other pc wasnt logged in!
moeSizlak: btw good job my r300 in f12 can do compiz again
moeSizlak: first time since f9
moeSizlak: but iy occasionally hard locks
moeSizlak: when compiz is on
kdekorte: Ganang: might just need to upgrade your driver
Ganang: kdekorte: you know, sometimes i think this is something with my monitor. It is new by the way, but even with the official drivers in windows, it took me days until i could enable this "sync mode"
Ganang: kdekorte: otherwise it wouldn't work
Ganang: kdekorte: i just feel that i have to force configuration....i will give a try on the force mode, with xrandr
airlied: damn no Zajec
evil_core: airlied: will be dynpm w/o flickering?
airlied: evil_core: its on the plan, I suspect the last commit from Zajec made it worse
evil_core: I am not sure wqhy, but for me dynpm(not dynclkcs) casues flickering, is it normal?
xming: .33 won't get pm?
jcristau: xming: for it to make .33 it needed to be ready 2 or 3 months ago
xming: well I was just asking, sometimes I see driver updates/addition in the late RC cycle
xming: will drm-radeon-testing be rebased on .33 once it's released?
agd5f: evil_core: yes at the moment it is
stikonas: airlied: I tried yout newly rebased drm-next branch, and X does not start on RV730
airlied: stikonas: yeah it needs a patch from dri-devel
airlied: I';ll try and rebase it again today
Ganang: kdekorte: do you know any application where i can try to at least enable my monitor, and force to turn it on?
Ganang: twnqx: which comand should i fire to enable it via xrandr?
twnqx: two steps.
twnqx: well, three...
twnqx: xrandr -q will tell you what the name of the external output is, and the available modes
twnqx: then xrandr --output