On Tue, 2009-09-29 at 12:47 +1000, Ben Skeggs wrote:
On Mon, 2009-09-28 at 19:37 -0700, Adam Williamson wrote:
On Tue, 2009-09-29 at 02:52 +0200, shmuel siegel wrote:
because we don't want to spend our time maintaining a large database of what cards to use kms for and what cards not to?
why aren't you optimistic, anyway? nouveau bugs are generally fixed quite rapidly.
This particular chipset is not well supported. In fact, the last time I looked, nobody, including nvidia, seemed to be supporting it. Way back when we did the nouveau test day, it was accepted that modeset would not work with this chip. I was pleasantly surprised to see that the driver in kernel 48 can recover from whatever problem it has but messages like "expect strangeness" and "GPU stuck", don't sit well with me. see https://bugzilla.redhat.com/show_bug.cgi?id=525905 for current status. P.S. I do want to thank you and Ben for taking this problem seriously.
ah, I see. If it's a known issue that it will be hard to make modesetting work on this chip, Ben may want to implement an exception in the driver to always disable modesetting for this chip, I suppose. However, I believe the long-term plan for nouveau, as for other drivers, is to _only_ use kernel modesetting, no more X driver modesetting code available, so that won't be practical in the long term. Ben, do you have a plan here?
The modesetting isn't an issue it seems, as the report says that a mode actually does get set. There's several bug reports of NV4E (the chipset the card reports itself as having) chips not functioning correctly for whatever reason.
Unfortunately I have no access to an effected chipset to investigate what's going wrong exactly, it's very difficult to fix such bugs.
shmuel seems to be suggesting that it works with nomodeset, though.
james, is there a way we could check if any system at Red Hat has the affected hardware? maybe we could ship it to Ben or something. it seems to be quite common:
http://www.smolts.org/reports/view_device/C51G%20%5BGeForce%206100%5D
so this is probably a bug we'd like to fix...
-
Adam Williamson