Bug#346486: Mouse events before card is initalized?
On Wed, 2006-02-08 at 16:20 +0100, Seiden Tiger wrote:
>
> There are two facts which may make a difference, and help starting x
> successfully
>
> *no mouse events on startup (since the error doesn't occour on every startup
> am am only 90% sure on this on).
> *multihead config (which doesn't work on multihead but starts reliably)
This is interesting, but unfortunately doesn't ring any bells here.
> Also:
> *the backtrace allways stops in the animCursor etc function.
Which in a way is good, because it suggests possible workarounds (SW
cursor, non-animated cursor theme, ...). At the same time, this code
path seems relatively innocuous, with little potential for hangs. Also,
this is definitely not 'before the card is initialized', as it happens
after ScreenInit(), which is the last driver init function called by the
X server. The driver is supposed to cope with anything after ScreenInit.
> Ways to resolve this problem:
> * Is there a way to use pcitweak to look at the registers and look for a
> difference? If yes, at wich memory range are the offset registers?
This approach is impractical because there's a large number of
registers, many of which will differ genuinely. However, the contents of
the RADEON_RBBM_STATUS register when this happens might be interesting
for a start.
> * Is there a way to abort gracefully when the error message:
> "(EE) RADEON(0): Idle timed out, resetting engine..." appeared multiple times?
The driver could detect this situation and call FatalError() or
something I guess... not sure that would be much of an improvement
though.
> *Is there a way rollback the changes on the radeon driver to track down the
> problems by binary searching through the patchsets?
Sure, this can be done in the upstream CVS repository.
At any rate, it would be good if you also brought this up upstream, on
the xorg mailing list and/or http://bugs.freedesktop.org .
--
Earthling Michel Dänzer | Debian (powerpc), X and DRI developer
Libre software enthusiast | http://svcs.affero.net/rm.php?r=daenzer
Reply to: