Fixing R300 hangs

[Mike A. Harris]

John Thacker wrote:
> On Mon, Feb 27, 2006 at 07:11:34AM -0500, Mike A. Harris wrote:
> 
>>My current plan for the r300 dri driver, is to leave it out of FC5, and
>>once the various issues are resolved upstream we will re-enable it in
>>rawhide post-FC5.  This will allow the driver to get the level of
>>testing that is needed, and hopefully to stabilize sooner.  Then, once
>>the major problems are resolved upstream by X.Org/DRI developers, and
>>new upstream stable packages have been released with the fixes, we will
>>probably include them in rawhide.
> 
> 
> Sounds decent.  I'd be extremely reluctant to update FC5 to something
> that's only in CVS.  The only reason for suggesting it is that it seemed
> like you were saying that some people were seeing lockups even with
> the r300 dri driver from Mesa not even present on their system.  Maybe
> I misread it, but if it turns out that there's no other way to prevent
> hangs on r300-based cards then the CVS patches should perhaps be considered.

There are multiple different problems, rather than a single problem.

Some users experience lockups or otherwise non-working X without the
r300_dri driver present.  That issue is still there, and hopefully
can be diagnosed and fixed for FC5.

With the r300_dri driver installed, users who do not experience the
above problem (or problems - there are many reports, and it may be
multiple bugs), may still have hangs wether or not the "Load "dri""
is present or not, and wether or not "Option "nodri"" is specified.

If none of these problems happen, and no other non-3D related hang
kills a person's system, then it is just a crapshoot wether DRI
enabled operation will work on a given R300 or higher card with
the driver present, and what magic incantations might be needed
to get it to work on a given motherboard.

All and all, R300+ support is quite shaky right now, both 3D and 2D,
so it is a process of reducing the instability by removing variables,
and then over time resolving issues one at a time.  The priority
for FC5 is to get as many people's systems in "X starts and I can
surf the web" state.  Anything on top of that is icing on the cake,
and bonus-plan material.  If we can get to that state, there will
be far fewer incoming bug reports when FC5 goes out the door, giving
us much more time to spend fixing bugs rather than triaging mass
bug duplicates.



> Note that some of the bug reports claim that even without DRI lockups
> occur from using the framebuffer device, and that a lot of the patches
> apply to the main radeon X driver rather than dri.  I don't know if
> removing the dri driver is enough to fix it.  Hopefully so.

Exactly.  There is no "it".  There are numerous bugs occuring, some
of which might even be kernel related.  Removing the dri driver,
removes incoming bug reports related to people who are trying to
use the DRI driver, when it is known to be unstable and experimental
by definition, allowing us to concentrate on bugs that are occuring
with the non-experimental code.

Anyway, it's a done deal.


-- 
Mike A. Harris  *  Open Source Advocate  *  http://mharris.ca
                       Proud Canadian.