Bugzilla – Bug 543927
radeonhd [RV530/X1650] No outputs detected
Last modified: 2009-10-12 13:12:35 UTC
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9.1.3) Gecko/20090909 SUSE/3.5.3-1.2 Firefox/3.5.3 FirePHP/0.3 After the first step of installation of opensuse M8 on 64bits with hardware radeon X1650 Pci-e 16x card inside, the second part run in text mode ( not a trouble for me ). Now after a reboot only have console. No X Neither sax2 or fglrx compilation give a way to have X running. Reproducible: Always Steps to Reproduce: 1. found a RV530 graphic card chipset 2. try to install opensuse 11.2 3. fail and this would affect many people around Actual Results: No way to obtain a X running Expected Results: Capability to have a working module for this card. flgrx or radeon I will put a maximum of log in attachement.
Created attachment 320866 [details] Sax log trying to adjust after first X crash
Created attachment 320868 [details] Xorg start log with backtrace
Created attachment 320869 [details] Full hwinfo of pc
Created attachment 320870 [details] /var/log/messages Zipped due to pollution of quota bactrace polluting logs
Created attachment 320871 [details] Full Yast2 installation log
Created attachment 320872 [details] Lastest boot messages
Smolt profile is here http://www.smolts.org/client/show/pub_f3dbc38a-37fd-4b46-b3de-d0293a0d94ab/ Also I've found this remarks on http://wiki.cchtml.com/index.php/Hardware RADEON Supported with the legacy ATI 9-3 Catalyst release. Currently there is NO support for 2.6.29 or newer kernels. (long, very long list of hardware) In comparaison of http://wiki.cchtml.com/index.php/Open_source_drivers (short, very short list) My first remark would be : how can we avoid freeze users ( especially newbies ) to a simple console if they have such hardware. Can we find a way to have a cool fallback to vesa or other. ---- I've try to add the ATI repository but only pointing to suse/11.1 Lastest ati-driver-installer-9-3-x86.x86_64.run refuse to compile or do anything.
Created attachment 320956 [details] output of modinfo (kernel-vanilla) of radeon.ko I've just try to compare radeon.ko between opensuse11.2 and the lastest ubuntu 9.10 ... Just one big difference, it's works out of the box in ubuntu.
Created attachment 320957 [details] Output of modinfo radeon in ubuntu 9.10 Here's the output of radeon.ko which works. Can we have the same in opensuse 11.2 or will we see all of our ati/amd users going to another distro ?
You're comparing "radeon driver + Kernel mode setting" (Ubuntu 9.10) with radeonhd (openSUSE 11.2). BTW, we had KMS enabled (unintentionally; it's enabled by default in upstream kernel although considered experimental!) for radeon kernel module for one Milestone. We received a lot of feedback during this time. A bunch of radeon card users were complaining about black screens. We decided to disable it. Also having it enabled would have required a complete new user stack, i.e. libdrm, Mesa and xf86-video-radeon from some special git branch, also considered experimental. I bet this new user stack wouldn't have worked at all without KMS. I believe we should concentrate on figuring out why no output is detected with your graphics card when using radeonhd. Thanks.
Thanks Stefan for your comments. I pretty agree with you about the "experimental" things. but we really need to find a issue, for this. This chipset is also present (with the mobility version ) in quite a number of MacBook pro. Don't hesitate to ask something. The machine is ready to try anything you need :-)
Well, Matthias/Egbert are our radeonhd driver developers. :-)
First thing to try would be to update radeonhd driver to the latest available version, i.e. release 1.2.5_20091001_be7216f http://download.opensuse.org/repositories/X11:/Drivers:/Video/openSUSE_Factory/
ok I've added the repository and made a zypper dup. so now inside system I've this one installed. xorg-x11-driver-video-radeonhd-1.2.5_20091001_be7216f-2.1.x86_64 But when I give it a startx I've always received a fatal error. Also I try the two connecteurs (normal is DVI) and the analog one with same non success. The most crazy thing is if I try a sax2 -r on the pc it doesn't work. If I do it from ssh, sax2 finally present a decent config. But the build xorg.conf doesn't work. the Xorg.0.log talk about : found an RV530 on unknown card 0x71c6:0x17AF:0x206A blabla ... contact radeonhd@opensuse.org Card informations : PCIE-16x Manuf : HIS Excalibur Radeon XT1650 Pro IceQ 256Mo 1xDVI + 1xVGA, Tvout +lspci coming in attachement.
Created attachment 320970 [details] The full lspci -vvvvv full detailled listing
Can you please run rhd_conntest as root with one output connected only (once per output)? I assume the AtomBIOS connector table is broken.
Bruno, could you please describe what you have connected to this card? You said your card has 1 DVI, one VGA connector and TVout. This doesn't quite match what the BIOS tells us: (II) RADEONHD(0): Connector[0] {RHD_CONNECTOR_DVI, "DVI-I DFP3 CRT1", RHD_DDC_0, RHD_HPD_1, { RHD_OUTPUT_LVTMA, RHD_OUTPUT_DACA } } (II) RADEONHD(0): Connector[1] {RHD_CONNECTOR_TV, "COMPOSITE TV1", RHD_DDC_NONE, RHD_HPD_NONE, { RHD_OUTPUT_DACB, RHD_OUTPUT_NONE } } (II) RADEONHD(0): Connector[2] {RHD_CONNECTOR_DVI, "DVI-I DFP1 CRT2", RHD_DDC_1, RHD_HPD_0, { RHD_OUTPUT_TMDSA, RHD_OUTPUT_DACB } } This suggests that we have two DVI-I outputs. Also please provide what Matthias asked for in comment #16. (BTW: this looks like a very card specific issue as we have working RV530 cards here)
Created attachment 321284 [details] rhdcontest with DVI monitor attached here's the -d
Created attachment 321286 [details] rhdcontest with VGA monitor attached the -d vga result
Created attachment 321287 [details] rhd_contest -s ( first dvi, second vga )
Created attachment 321289 [details] Picture of the card connexion As you can see there's a DVI -> a TV -> a VGA connexion here.
One work about those specific card, The motherboard is a Asrock aliveNf6G-DVI http://www.asrock.com/MB/overview.asp?Model=ALiveNF6G-DVI It has already a vga card (nvidia geforce 61xx ) inside but is disable, and a have the ati/amd http://ati.amd.com/products/certified/his.html line Radeon X1650 PRO HIS Radeon X1650 PRO 256MB DDR3 PCIE ICEQ Turbo 256MB PCI-E DVI-I, DVI-I, & TV-Out Hoo .. did you see they talk about 2 DVI ( like the bios ) but only one is a real DVI ? This card was working nicely with openSUSE 11.0 and the flgrx drivers ( If you need other information, I've just the old disk at a hand )
Please don't reassign, unless you want to fix the issue yourself ;-)
Oups sorry, I've not reassign, perharps when I've made the attachement I've check the box which say "ok this give needed info" ...
Ok, I think this can be fixed. With a config file you could simply try the option: Option "HPD" "off" but we don't have one any more :(
According to comment #14 Bruno knows how to create one. :-) Bruno, please give that option a try (Section "Device").
Pushed fix upstream, will be included in next driver rpm.
Bruno, if you can create a config file please do so and add the option line. This will help to verify that the assumption was right.
Created attachment 321326 [details] Screen shot under X krandr A screenshot with just DVI connected and option HPD off
Created attachment 321327 [details] krandr screenshot with DVI & VGA active Just to see the mapping of different port.
Ok this option is Good :-)) Now I'm having the X running, but I'm stick to a 1024x768-24 resolution maximun and I should goes to at a minimal of my monitor ( HP L2045w ) which is 1650 x 1050 @ 60Hz (any other wonderfull trick ? ) Thanks guy for your help ... (really really appreciate here ... )
Sounds like another incarnation of fdo bug #24330. Good news is we approximately know what to do. So stay tuned.
Thanks Mathias, good news to now you're on a good way. Arf the #24330 is not accessible with my opensuse contributor's profile :-) If you can I really love to be cc of it. So I'm just waiting new release of packages. Ping here when you think it's ready to new tests.
I've pushed a fix for the connector table upstream today. The resolution issue still needs some investigation.
This one should be fixed with the latest update which you can find in the repository: http://download.opensuse.org/repositories/X11:/XOrg/openSUSE_Factory/ Could you check this please?
Yeap : with the 20091009 version of drivers, the correct resolution is applied with or without a xorg.conf. I've remove the xorg.conf and play a bit with kde. I've encounter a "big" trouble as it's a crash freeze if you enable desktop effect and chosse the woobly windows. Push a apply, if you move the first windows which is systemsettings you see a woobly effect, but if you click somewhere else you loose X, mouse and keyboard. (I can't even get the machine over ssh) But the power button permit to shutdown it smoothly. Without desktop effect it work quietly well. Did you think desktop effect are regression on the drivers, or nothing to do, and we close this bug, and open a new one ???
Thanks. The issue when enabling desktop effects is likely a duplicate of Bug #542453. Check your /var/log/Xorg.0.log for "(WW) RADEONHD(0): DRMCPIdle: DRM CP IDLE returned BUSY!" lines.
Thanks for the pointer Stefan, I would add my log there (even if I've not the line you mention) There's nothing in Xorg.0.log just this is /var/log/messages Oct 12 11:36:07 x-wing kdm_config[2199]: Multiple occurrences of key 'UseTheme' in section [X-*-Greeter] of /usr/share/kde4/config/kdm/kdmrc Oct 12 11:36:07 x-wing kernel: [ 2517.051342] [drm] Setting GART location based on new memory map Oct 12 11:36:07 x-wing kernel: [ 2517.051581] [drm] Loading R500 Microcode Oct 12 11:36:07 x-wing kernel: [ 2517.051607] [drm] Num pipes: 1 Oct 12 11:36:07 x-wing kernel: [ 2517.051614] [drm] writeback test succeeded in 1 usecs Oct 12 11:36:50 x-wing python: hp-systray[2462]: warning: No hp: or hpfax: devices found in any installed CUPS queue. Exiting. Oct 12 11:36:56 x-wing kernel: [ 2565.442442] [drm] Num pipes: 1 Oct 12 11:36:56 x-wing kdm[2198]: X server for display :0 terminated unexpectedly If you believe that we have check all possibility for the subject of this bug, you could close it. The job is done. well done. Thanks for your help & work.
Hmm. Then it might be a different issue. Egbert will close the bug, since he's the assignee. :-) I believe it won't hurt to attach the new Xorg.0.log.
Created attachment 322041 [details] At least a finally working Xorg log With correct output detected and correct screen resolution.
The original issue is fixed. What remains is the lockup which many people still see with R5xx.
Thanks for closing. Yeap the hang is due to compositing ( under kde, I've not tested gnome ). If composite is enable you can say good-bye to your desktop. :-) I've really appreciate your help (all guys) to fixe this one.