On Wed, 8 Dec 2010 11:22:59 -0800 Dan Malek <ppc6...@digitaldans.com> wrote:
> > On Dec 8, 2010, at 11:11 AM, Scott Wood wrote: > > > Probably want to complain to the user if reg is invalid and not > > zero/missing. > > I think you guys are making this too complicated. > There are many ways to pass stupid mistakes via > a device tree, don't get carried away trying to single > out this one for error checking where the user is likely > to really know what they are doing. This isn't a required > specification to get correct, without anything u-boot > will provide the proper information. I don't think that's what this is for. I think it's for AMP scenarios where you want to carve up memory between guests -- in which case you want U-Boot to not overwrite the device tree with its own idea of how much memory is present. This patch was trying to guess whether that's the case based on whether the existing value looks reasonable, but I think it has to be an explicit configuration (board config, environment variable, etc). Once that is done, then the test should either go away, or complain -- don't just silently do something different if it's been told that the memory node is supposed to be complete and correct. As for the merits of error checking in general, obviously it's not going to be complete. But it can be useful to check for some common errors, such as when a board has multiple DTSes and the user has to pick the right one for how U-Boot has been configured. Maybe granting an AMP partition a memory region beyond the bounds of detected memory is another such case. -Scott _______________________________________________ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
