Dear Rasmus,
In message 86e1c45a-fc23-6794-7ca9-e96910af4743@prevas.dk you wrote:
Testing on a single platform (which apparently has aother problems, or you would not need any such change) is not convincing.
I don't, actually, need this change, but suggested it as a way towards making the primitives available a little more consistent across architectures since I stumbled on this implementation detail while single-stepping my board to figure out why it would reset in the SPL. As I'm unable to convince you that the benefits of that outweigh the risk of introducing a regression, feel free to drop it.
OK.
I do, however, need the other ppc patch I sent yesterday: https://patchwork.ozlabs.org/patch/1255844/ . That one should satisfy the requirement that it cannot possibly break anything for existing boards.
This other patch lacks documentation - just from reading the code nobody will be able to understand what this piece of code is intended for, so there is a risk that it might be removed by some later "cleanup".
Please add some comment in the code to explain the intentions.
Best regards,
Wolfgang Denk