I have to bring this Fedora 9 bug here: https://bugzilla.redhat.com/show_bug.cgi?id=451349 The reporter has an X1200 which worked fine with 45fdec79, but 26ccf117 and ad59f09e do something strange with the memory setup ("Splitting WC range", "Failed to set up write-combining range") and he ends up with a scrambled screen. Screenshot: https://bugzilla.redhat.com/attachment.cgi?id=309325 Xorg.0.log diffs: https://bugzilla.redhat.com/attachment.cgi?id=309324 https://bugzilla.redhat.com/attachment.cgi?id=309294 Software versions in question: 45fdec79: xorg-x11-drv-radeonhd-1.2.1-1.1.20080429git 26ccf117: xorg-x11-drv-radeonhd-1.2.1-2.20080605git ad59f09e: xorg-x11-drv-radeonhd-1.2.1-2.20080613git I could ask the reporter to git-bisect, but perhaps the experts here know at once. No hurry though, reporter says he is going to be away for 1.5 weeks.
New report, similar problem, different hardware (amd780g/HD 3200): https://bugzilla.redhat.com/show_bug.cgi?id=452050
There have been some changes to the mapping code recently. It could be that this problem has been taken care of. If not please provide a new log (verbose 7). You also may want to try the following option: Option "UnverifiedFeatures" "No"
Well, https://bugzilla.redhat.com/show_bug.cgi?id=452050 suggests more that the very changes you are talking of broke it for 780? 1f65f354 is reported to be broken, and ad59f09e to work for the 780.
The reporter with the 780 ( https://bugzilla.redhat.com/show_bug.cgi?id=452050 ) reports that Option "UnverifiedFeatures" "No" fixes his issues.
OK, I see the problem. Fix will be in shortly.
A sanity check to avoid this situation has been added to git. Ndim, maybe you can create new Fedora packages for testing. Thanks! Reassigning to you, please assign back to me or close depending on the test results.
Mostly, but not completely solved, I guess. Tested with 1eff3e78 aka xorg-x11-drv-radeonhd-1.2.1-3.2.20080622git.fc9, both with and without UnverifiedFeatures: https://bugzilla.redhat.com/show_bug.cgi?id=452050#c13
The X1200 guy reports a completely white screen with 1eff3e78: https://bugzilla.redhat.com/show_bug.cgi?id=451349#c12 His Xorg.0.log diff has two insteresting parts: https://bugzilla.redhat.com/attachment.cgi?id=310135&action=view -(II) RADEONHD(0): Mapped IO at 0xb7f20000 (size 0x00010000) +(II) RADEONHD(0): Mapped IO at 0xb7ffe000 (size 0x00010000) Uhm... a non-aligned address? That does look fishy. -(II) RADEONHD(0): Mapped IO at 0xb7f20000 (size 0x00010000) -(II) RADEONHD(0): Using 3883 scanlines of offscreen memory +(II) RADEONHD(0): Mapped IO at 0xb7ffe000 (size 0x00010000) +(II) RADEONHD(0): IPG sideport memory not present. +(WW) RADEONHD(0): IGP memory base 0x08c69c90 seems to be bogus. +(==) RADEONHD(0): Not Mapping IGP memory +(II) RADEONHD(0): Mapped FB at 0xa7fed000 (size 0x10000000) +(II) RADEONHD(0): Using 6511 scanlines of offscreen memory Hmm. Is that IGP stuff relevant?
The X1200 guy reports that the Fedora package built from commit c2139d8c works for hime, so I'll close this bug.
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.