As discussed in bug 3911, disabling bus mastering in the RADEONSetFBLocation() function can have undesired side effects. The purpose of this bug is to derive a minimal fix for those side effects from Benjamin Herrenschmidt's more extensive cleanup patches, that can be backported to X.Org 6.8 with minimal risk.
Created attachment 4643 [details] [review] WIP patch against 6.9 This patch disables all memory requests to the framebuffer instead of disabling bus mastering. When applying to 6.8, just ignore the missing BUS_CNTL stuff.
Created attachment 4649 [details] [review] WIP patch for xf86-video-ati HEAD This patch is a merge of the 6.9 patch on top of Ben's memmap fix.
The current patch has been reported to hang on some systems, see https://bugzilla.novell.com/show_bug.cgi?id=151631 . This is being investigated.
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=182471 is the Red Hat master bug for this issue.
Can somebody reproduce and track down the hang pointed to in comment #3? Thanks.
Created attachment 4864 [details] [review] Even more minimal patch against 6.8 While we're waiting for the issues with the previous patches being tracked down, here's an even more minimal patch that's basically a backport of the relevant parts of Ben's memmap fixes to 6.9. This has been confirmed to fix the so-called 'diagonal line corruption' and there have been no reports so far of it re-introducing any SERR conditions. Again, when applying to 6.8, just ignore the missing BUS_CNTL context.
Can we get this checked into the stable branch, so that it becomes part of the 1.0.2 X server release?
(In reply to comment #7) > Can we get this checked into the stable branch, so that it becomes part of > the 1.0.2 X server release? Ok um.. I'm not sure what exactly I was thinking when I said that. Duh. What I really meant, was can we get this checked into the stable branch of the radeon driver, and a new stable driver release made.
Comment on attachment 4864 [details] [review] Even more minimal patch against 6.8 I incorrectly described this as being against 6.9, it's really against 6.8. Nominating for the 6.8 branch, as dead as that is, because doing so should increase the chances of it being picked up by Red Hat according to Mike A. Harris. :) FWIW, this patch has been tested extensively at ATI and Dell, and no regressions have been reported so far.
Is this patch merged into mainline X?
Added patch keyword
Sorry about the phenomenal bug spam, guys. Adding xorg-team@ to the QA contact so bugs don't get lost in future.
closing. most recent versions of the driver have this.
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.