| Summary: | [2.6.29 regression] graphic corruption or xorg infinite loop | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Robin Bankhead <gentoo> |
| Component: | [OLD] Core system | Assignee: | Gentoo Kernel Bug Wranglers and Kernel Maintainers <kernel> |
| Status: | RESOLVED OBSOLETE | ||
| Severity: | normal | CC: | x11 |
| Priority: | High | ||
| Version: | 2008.0 | ||
| Hardware: | x86 | ||
| OS: | Linux | ||
| URL: | https://bugs.freedesktop.org/show_bug.cgi?id=22597 | ||
| Whiteboard: | linux-2.6.29-regression | ||
| Package list: | Runtime testing required: | --- | |
| Attachments: |
xorg.conf
kernel 2.6.29-gentoo .config with KMS enabled Xorg.0.log with KMS enabled 2.6.29-gentoo .config with KMS disabled Xorg.0.log with KMS disabled Xorg.0.log with 2.6.29-gentoo-r4 and xorg-server-1.6.1.901, KMS + UXA + DRI2 |
||
|
Description
Robin Bankhead
2009-03-29 01:17:38 UTC
Created attachment 186602 [details]
xorg.conf
Created attachment 186603 [details]
kernel 2.6.29-gentoo .config with KMS enabled
Created attachment 186604 [details]
Xorg.0.log with KMS enabled
Created attachment 186610 [details]
2.6.29-gentoo .config with KMS disabled
Created attachment 186612 [details]
Xorg.0.log with KMS disabled
After updating mesa, then reinstalling xorg-server 1.6 with appropriate drivers and running X -configure, the issue no longer occurs for me. (In reply to comment #6) > After updating mesa, then reinstalling xorg-server 1.6 with appropriate drivers > and running X -configure, the issue no longer occurs for me. > xorg-server-1.6 is not in portage yet, though. Sorry but we cant fix this problem as described in bug 262315 Please follow those directions Closing as duplicate Re-open if needed *** This bug has been marked as a duplicate of bug 262315 *** This bug is different, or at least half of it. There is graphics corruption in X even when KMS is not being used. I think it may be fixed in 2.6.29.3 but we have yet to get that into gentoo-sources. Allright Could you please try to change this line on your xorg.conf Option "AccelMethod" "exa" Use "UXA" instead of "exa" and let us know how this ll go (In reply to comment #10) > Allright > > Could you please try to change this line on your xorg.conf > > Option "AccelMethod" "exa" > > Use "UXA" instead of "exa" and let us know how this ll go > Can you give me further info? I've never heard of this UXA and it seems from what info I can find that this is only supported by xorg 1.6. Is this correct, because I'm only using portage tree here. If not, are there other changes I need to make in xorg.conf to go along with this change? (In reply to comment #11) > (In reply to comment #10) > > Allright > > > > Could you please try to change this line on your xorg.conf > > > > Option "AccelMethod" "exa" > > > > Use "UXA" instead of "exa" and let us know how this ll go > > > > Can you give me further info? I've never heard of this UXA and it seems from > what info I can find that this is only supported by xorg 1.6. Is this correct, > because I'm only using portage tree here. If not, are there other changes I > need to make in xorg.conf to go along with this change? > I ve been using it with xorg-1.5 . EXA was supported until xf86-video-intel-2.5 but for later versions of this driver you could use UXA which is recommended. So if you are using >xf86-video-intel driver please use UXA. You dont need to change anything else on your xorg.conf No improvement. Still get the stippled kdm screen without KMS, and with it, the kdm wallpaper appears but is sort of low-depth, as though in VGA mode. The login box doesn't appear, and acpi-button shutdown doesn't work either now, have to use SysRq and that still doesn't kill the screen. BTW am now on 2.6.29-gentoo-r3. Please test with gentoo-sources-2.6.29-r4, it includes a patch which is likely to solve this Not much change without KMS (have an emerge -uND world to complete before I can test with KMS now). With exa = stippling. With uxa I can enter my login but the process times out on "Loading the desktop" (desktop icon half-focused on the spash-screen). ACPI-shutdown still works though. I've overwritten my Xorg.0.log from that session, but I did notice something in it along the lines of "Output VGA attached to monitor LFP0" (you can see this Monitor section in my xorg.conf attached above). The screen quality didn't look like VGA on this run as it did with the old kernel, though. Sorry this is a bit vague. I'll supply logs for both KMS and non-KMS when I have the chance. Did 2.6.28 work any better? 2.6.28 is fine, and with uxa enabled it's even better (with exa there's a lot of periodic screen corruption in KDE, especially when using compositing) x11 people, any comments here? (specifically the non-KMS issues, not too bothered about KMS failures, especially so on xorg-server-1.5) Which bugzilla is the right upstream for this kind of issue, freedesktop or kernel? My first thought is that you should definitely try disabling uvesafb _completely_ (even without KMS). Framebuffer drivers have been a constant source of problems. You might also want to try xorg-server 1.6 which is p.masked at the moment (it still needs a bunch of patches but it could be interesting to try it out). If you do, don't forget to enable UXA like Markos suggested so that you get both UXA and DRI2. In a nutshell, this will get you _much_ closer to what upstream is currently testing and hacking on. Thanks I didn't realize you could have *no* framebuffer, I'll give that a try. Am I right in thinking that xorg-1.6 requires a ton of packages to be rebuilt simply because of a renamed library? I seem to recall a problem like this when I looked at the x11 overlay recently. No, just unmask xorg-server 1.6, you should be fine. The libxcb/libX11 madness is completely unrelated (fortunately!). Thanks Created attachment 191603 [details]
Xorg.0.log with 2.6.29-gentoo-r4 and xorg-server-1.6.1.901, KMS + UXA + DRI2
Here's the log for xorg-1.6 with KMS enabled under UXA/DRI2. It represents a successful login to KDE followed by shutdown, which was only possible after disabling compositing in kwin. Enabling compositing seems to have been what previously caused the login to halt at the "loading desktop" stage.
Even now, with KMS I still get the VGA-like blocky graphics, and this persists onto the desktop once loaded. It doesn't affect all graphic elements: on the KDM screen only the wallpaper is affected whereas the central box, both at the login prompt and during loading, is perfect quality. Once loaded, the desktop wallpaper and kickoff are OK, but the inner areas of plasmoids (e.g. System Monitor components) are blocky.
Note: I built new versions of libdrm and mesa prior to xorg-server, and afterwards rebuilt all xf86-* drivers. I also commented-out the ServerLayout, Monitor and Screen sections in xorg.conf leaving only the Device sections which are as they were in my first attachment, apart from uxa. This seems to work fine with 2.6.28 as well.
Could you disable KMS by putting "i915.modeset=0" on your kernel command line in grub.conf? If that doesn't change anything, please file a bug in FreeDesktop's bugzilla [1] and paste the url here. Thanks [1] http://intellinuxgraphics.org/how_to_report_bug.html Disabling KMS gives similar results. Without compositing I can log in, but there is still a slight amount of blockiness in the same affected areas as before (not nearly as noticeable though). There's also some corruption on the chrome of kwin windows, which may have been true before but I hadn't checked this. When attempting to enable compositing in System Settings, the window blanks-out and the screen freezes, needing ACPI-shutdown again. I also see that KDE seems to run a bit slower. It's barely noticeable with KMS on, but definitely feels slower with it off. In sum, there's no permutation of settings that doesn't have some negative impact in comparison to 2.6.28, so upstream it is. I'll backlink the report when I've done it. Sorry I haven't been near this for some time - the patient is now my secondary machine, and all my free time has been occupied dealing with some showstoppers on the primary. In the meantime, I've progressed on this machine to 2.6.30-gentoo-r1, where the symptoms are worse still. To wit: with KMS, X crashes on launch (only get as far as blank screen with frozen text-cursor) and hammers CPU, a freeze only rescuable by SysRq (ACPI buttons have no effect). Without KMS and with KDE compositing disabled the desktop will load as before, but window-deco corruption and blocky gradients as before too. I think fdo already have this bug [1], though the report is for AMD64 - can anyone advise before I latch onto it? [1]https://bugs.freedesktop.org/show_bug.cgi?id=20598 (In reply to comment #25) > I think fdo already have this bug [1], though the report is for AMD64 - can > anyone advise before I latch onto it? > > [1]https://bugs.freedesktop.org/show_bug.cgi?id=20598 Just to be safe, file your own bug with the versions _you_ are using and keep testing new kernels (especially the -rc ones) as they keep rolling out. Please paste the url here once you've opened a bug report. Thanks Right, finally done. Sorry for the delay. https://bugs.freedesktop.org/show_bug.cgi?id=22597 Let's track this upstream then. Thanks |
