Christoph Egger wrote:
Jean-Yves Migeon wrote:Thing is, take if_xennet_xenbus.c for example, it relies on MD functions, like x86 barriers and P2M translations. As a consequence, I am not sure that sys/dev/xen is a good candidate; it stays pretty close to x86, and not suitable for sys/dev/xen yet.For there barriers, there's the MI membar_*() API. The P2M translations would need an MI API with MD specific implementations.
Yes
Without having the P2M translations abstracted in a MI fashion, I think, sys/arch/xen/dev is the proper place.
I agree You are planning it for after the 5.x branch? -- Jean-Yves Migeon jean-yves.migeon%espci.fr@localhost