On Wed, Nov 28, 2007 at 07:49:12PM +0100, svn@openbios.org wrote:
a patch against LinuxBIOS to allow the Serengeti-Cheetah platform to come up on its emulated self in the SimNow simulator.
Should this go into LB proper?
//Peter
On 28/11/07 21:19 +0100, Peter Stuge wrote:
On Wed, Nov 28, 2007 at 07:49:12PM +0100, svn@openbios.org wrote:
a patch against LinuxBIOS to allow the Serengeti-Cheetah platform to come up on its emulated self in the SimNow simulator.
Should this go into LB proper?
That is a good question. The defining factor is how many of the differences are because of bugs in SimNow, and how many are because of fundamental differences in how SimNow works. In this case, we have two problems. The first is the non-coherent hypertransport problem, which one is definitely a SimNow issue that has been fixed in the NDA version, and will hopefully dribble out to the public version. But until it does, we need to work around it. In this case, I would want to stick with an external patch , because its not LinuxBIOS's fault that SimNow is buggy.
The second problem has to do with the behavior of the other nodes after CAR is turned off on the BSP. Its hard to tell if this is a SimNow bug or limitation, we're still investigating. If it does turn out to be a SimNow limitation, then it would be proper to add a SimNow configuration option to LinuxBIOS and build it as a separate target. Obviously, thats a non-optimal solution, and I would like to avoid that at all costs.
Short story long, I think the patch should stay as a patch for now, while we're optimistic that SimNow bugs could account for the failures. If in the future, we figure out that this is just part of SimNow's nature, we can add it to be a permanent part of LinuxBIOS.
Jordan