Summary: | [Fedora 9 G965] EXA corruption when booted by isolinux vesa menu | ||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Product: | xorg | Reporter: | Bill Nottingham <notting> | ||||||||||||||||||
Component: | Driver/intel | Assignee: | Carl Worth <cworth> | ||||||||||||||||||
Status: | RESOLVED FIXED | QA Contact: | Xorg Project Team <xorg-team> | ||||||||||||||||||
Severity: | normal | ||||||||||||||||||||
Priority: | medium | CC: | brice.goglin, faijeya, jbarnes, michael.fu | ||||||||||||||||||
Version: | unspecified | ||||||||||||||||||||
Hardware: | Other | ||||||||||||||||||||
OS: | All | ||||||||||||||||||||
Whiteboard: | |||||||||||||||||||||
i915 platform: | i915 features: | ||||||||||||||||||||
Attachments: |
|
Description
Bill Nottingham
2008-03-14 09:12:08 UTC
Created attachment 15124 [details]
X config
Created attachment 15125 [details]
screenshot (sorry, crappy digicam)
Sometimes this happens even with DRI (but not always) yeah, pls try my test patch on it, although vt switch is broken.. Patch is in upstream master now, close this. Testing of 2.2.1 + commit 8cdbd55f8075cd18b563badde35815665d7d053e (which I'm assuming is what you're referring to), the problem still persists. Weird, could you try to turn on/off options like "Tiling", "EXANoComposite" to see if there's any difference? EXANoComposite fixes it. Turning off tiling does not. Bill, could you tell me your G965 model? Could you test with 2.2.99.902? thanks. 82G965, rev2. (8086:29a2) Note that this is being reported fairly often in Fedora's bugzilla (6 duplicates already); I suspect it happens on more than one 965 model. Persists in 2.2.99.902. I reproduce this on Q965 and G965 here with fc9 install gui, which is corrupted. But I can't produce this with our Xorg build on two machines, disable DRI both works fine. From your log and recent one from rh bug, it looks intel driver isn't the recent master one, are there any patch list on intel driver in fc9? otherwise we have no clue to what might be wrong. Created attachment 15603 [details] [review] patch #1: intel-2.1.1-fix-xv-reset Patch #1 that's added. Theoretically irrelevant here, as Xv is not involved. Created attachment 15604 [details] [review] Patch #2: intel-2.1.1-efi.patch Patch #2. From the title, it's related to EFI 'fun' on Intel Macs. Created attachment 15605 [details] [review] patch #3: intel-2.1.1-disable-ttm.patch Patch #3: disable ttm Of course, since the issue persists with all of those patches removed, I'm not sure why they'd be at fault. yeah, patch #1 and #2 is already upstream. So is it true that after install fc9 with text mode, X still fail later in normal start? I tried with a Fedora 9 beta LiveCD on x86_64. X started with DRI even caused font disappear etc. rendering corrupt, looking at the log intel_drv.so will load submodule intel_master_drv.so, which I don't know the details. So I pulled and built current upstream driver, and gnome started fine with this driver, with no screen artifacts with/without DRI. Switch back to fedora drivers, corrupt appeared again. So this seems a fedora driver bug, which we can't handle. The 'intel_master' stuff is to allow easy switching between the stock intel driver and the development batchbuffer driver. And without that patch, *the problem still persists*. Have we had enough of the blame game yet? I've tried multiple git versions, I've pulled out irrelevant patches, I've even pointed you to the EXA options that fix the behaviour. Can we get actual work on the problem? Bill, I can't reproduce this with upstream tips (xserver, xf-v-i, etc.) on Q965 here in consider fc9 uses mostly all latest bits. Not sure what more to add. The installer case has DRI not working b/c the kernel modules aren't there. I don't know that that would add any more/different fallbacks than just having it disabled in the config file, but you never know... Tested by remove all drm modules, still works ok here. It also works on my G965 with the upstream code. I've finally tracked down the cause - it has nothing to do with NoDRI; that's just a red herring. When booted via isolinux and its vesa menu, the corruption occurs. When booted via grub (with graphical menu), or isolinux without graphical menu, it renders fine. Created attachment 16437 [details]
register dump when booted from grub
Here's the register dump from a 'good' boot.
Created attachment 16438 [details]
register dump when booted from isolinux
Here's the register dump from when it doesn't work.
yeah, I tried to produce it also on 965 here, it looks isolinux program the device with another mode, maybe like same condition like linux fb. Jesse, any idea? Bill, I tried fc9 install which seems fine on 965G now with gui, this issue is resolved? Still exists for anything booted from isolinux. (In reply to comment #29) > Bill, I tried fc9 install which seems fine on 965G now with gui, this issue is > resolved? > zhenyu, are you using isolinux with its vesa menu? sounds like this has been narrowed down and should be easy to reproduce... *** Bug 15278 has been marked as a duplicate of this bug. *** re-assign to Carl. Carl, could you try it? it sounds should be an easy one to reproduce.. thanks. Hi Bill, Looks like it took me *leaving* the X team at Red Hat before I got pointed to this bug. :-) I'll test with an isolinux image this afternoon to see if I can reproduce this on my 965 machine. -Carl Works with 2.5.0. Not sure when it was fixed. (In reply to comment #35) > Works with 2.5.0. Not sure when it was fixed. Hi Bill, I'm sorry I never got the chance to test this, but I'm quite happy it's fixed now! I do appreciate your reports, and I hope we can be even more responsive to the next one. -Carl |
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.