| Summary: | DMA with mach64 DRI driver crashes X | ||
|---|---|---|---|
| Product: | Mesa | Reporter: | Sven Clasen <sven> |
| Component: | Drivers/DRI/Mach64 | Assignee: | Default DRI bug account <dri-devel> |
| Status: | RESOLVED NOTABUG | QA Contact: | |
| Severity: | major | ||
| Priority: | high | ||
| Version: | git | ||
| Hardware: | x86 (IA32) | ||
| OS: | Linux (All) | ||
| Whiteboard: | |||
| i915 platform: | i915 features: | ||
|
Description
Sven Clasen
2004-06-21 13:37:12 UTC
Can you try again with the next snapshot? A bug was fixed in the DRM that affected mach64 .. I want to rule it out ... i'll gladly try the new snapshot, because i want my rage mobility to work some day ;) new snapshots are announced at the dri-announce mailing list, i guess? *subscribes* (In reply to comment #2) > i'll gladly try the new snapshot, because i want my rage mobility to work some > day ;) > new snapshots are announced at the dri-announce mailing list, i guess? > *subscribes* nope, they are generated every night. grab them here (scroll down to the snapshots section): http://dri.sourceforge.net/cgi-bin/moin.cgi/Download if i only knew when "night" is :) it's 16:04 over here now.. 07:05. the last modified date of the snapshots is about 07:15. so not long anymore. (*sigh* still no time to learn for my maths exam which i'll write tomorrow) downloaded and ./install.sh'd it, works. Xfree86.0.log, output from dmesg and glxinfo -v can be found at http://sven.gotdns.org/~sven/dri/before-reboot/ rebooted, now using opengl screws X in a way which can only be fixed by rebooting and causes the system to freeze until i log in by ssh and kill the opengl application Xfree86.0.log and output from dmesg and glxinfo -v can be found at http://sven. gotdns.org/~sven/dri/after-reboot/ additionally, i found a log of very cryptographic stuff (at least to me), which includes something that might be interesting: Jun 22 16:30:30 amy kernel: [drm:mach64_ring_tick] *ERROR* bad ring head address: 0xfffffff0 and Jun 22 16:31:24 amy kernel: [drm:mach64_do_dma_idle] *ERROR* mach64_do_dma_idle failed BM_GUI_TABLE=0x02850000 a full cat /var/log/kern.log |grep "Jun 22 16:" can be found at http://sven. gotdns.org/~sven/dri/after-reboot/kern.log (fixing product/component and summary. also downgrading to major, since mach64 isn't built by default for security reasons.) mach64 DMA is known to be broken with current CVS. can you try changing the dma_mode line in your XF86Config to Option "dma_mode" "mmio" and see if that's more stable? it will be slower of course, but slower is better than crashing. ( sorry for messing up the bugzilla stuff, i am not familiar with bugzilla :/ ) Option "dma_mode" "mmio" is even worse, it locks the machine up completely, meaning its not even accessible by ssh, and it looks really frightening, at first i thought it burned the display (luckily it didn't) :/ if it's broken with the current CVS - can i get an archived version that will work? (ofcourse i'll try new snapshots also to see if you managed to fix it and provide logfiles if neccessary) (II) ATI(0): [drm] Will request asynchronous DMA mode (WW) ATI(0): [agp] AGP not available (WW) ATI(0): [agp] AGP failed to initialize -- falling back to PCI mode. (WW) ATI(0): [agp] Make sure you have the agpgart kernel module loaded. (II) ATI(0): [drm] register handle = 0xfcfff000 ouch.. . it is a PCI card? you should never see this ... (In reply to comment #10) > (II) ATI(0): [drm] Will request asynchronous DMA mode > (WW) ATI(0): [agp] AGP not available > (WW) ATI(0): [agp] AGP failed to initialize -- falling back to PCI mode. > (WW) ATI(0): [agp] Make sure you have the agpgart kernel module loaded. > (II) ATI(0): [drm] register handle = 0xfcfff000 > > ouch.. . it is a PCI card? you should never see this ... It is the onboard ATI Rage Mobility that shipped with my Dell Inspiron 3700, so i guess it should be pci. is it my fault/can i do anything to change that? Well I've the exact same machine (+ upgraded CPU and 3800 bios!!) that I use to test mach64 stuff on... In the dmesg you posted after-reboot, you never load any agpgart before or after the DRM (it should be loaded before it ... ) I've no idea why XFree86 isn't loading the agpgart module the second time around that is just wierd... you might want to either compile agpgart into your kernel or modprobe it before the mach64 DRM module is loaded... when i manually modprobe -r the mach64 module, then modprobe the agpgart module and then start x, it works! but, it is (with Option "dma_mode" "async") even slower than no using DRI O_O but since this has nothing to do with my crash/freeze and i already saw another bug reporting dri was very slow after upgrading, i guess i should change this to NOTABUG? Thanks for your support, guys :) I think this is solved.. PCI mach64 DRI is probably broken but it isn't this bug.. Mass version move, cvs -> git |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.