fwiw, this seems to have crept in not too long after 9f324860431ff8199a78d19bbaa74046e1476b89 (before or at 2e1425246ccc75216247b0c2fa6fce2635db472b - can't say more precisely, as i got tired of reboots).

Could you please provide the detailed driver version, according to http://intellinuxgraphics.org/how_to_report_bug.html?

It will be very helpful if you could locate which commit causes this regression. Thanks.

Jesse, is this related to bug#16169?

Yeah, it's probably related, though in this case it may be that we're just programming it wrong for 845G.  I'll come up with a patch for you, Oswald, thanks for reporting this.

Created attachment 17038 [details] [review]
Don't program DSPARB on old, single pipe chips

Can you give this patch a try?

doesn't seem to have helped.

note that pin-pointing this is somewhat hard due to the crash that can be seen at the bottom (how do i get a useful trace without a second terminal?). it doesn't seem to be driver-related (it crashes even with a quite old driver which is verified to kind of work here). after such a crash the chip is in such a state that even the old, kind-of working server+driver show only a black screen until reboot.

Ah, it crashes with an old driver too?  Can you just comment out the call to i830_set_dsparb() in i830_driver.c:I830EnterVT() and see if things work after a fresh boot?

ok, after a few really frustrating hours i think i have halfways isolated the bugs ...

your patch *does* help. and there is no crash related to this, "only" a lockup. heh. more to come in other reports ...

Any news here Oswald?  There's a patch in #16169 that might help too...

Ok, marking this one as fixed now that the patch in 16169 has gone upstream.

just tried master ... nope. :(
specifically, initial activation works (takes basically forever, though), but when i switch away the vt and then re-enter it, it again goes black and floods the log with the underrun messages.

ping Jesse...

Ugg, wonder if we have any 845 machines...

(In reply to comment #11)
> just tried master ... nope. :(
> specifically, initial activation works (takes basically forever, though), but
> when i switch away the vt and then re-enter it, it again goes black and floods
> the log with the underrun messages.
> 

Oswald, would you pls provide a log with modedebug turns on? thanks.

time out for reporter's response. reject..

Oswald, if you get a chance to get some more info, please reopen.  I wouldn't want to hinder any KDE developers from making my desktop better. :)

sorry for the delay. but given what an ordeal compiling and getting the server up again it was after all the build and config changes, the procrastination was justified. :-D

seems to work now. be sure i'll complain when it comes back. :)