Hello,
On 04.03.2016 20:21, Peter Lindener wrote:
Genode on ARMv8_64 sound like it will be a good thing, when it ends up happening...
... [Would Rump kernels help here?] ...
porting Genode to a new CPU architecture requires work on the following areas, which are unrelated to our use of Rump kernels:
- Tool-chain support (enabling the new architecture in Genode's GCC- based tool chain) - Architecture-specific assembly startup code - Port of the base-hw kernel (or enabling another kernel's support for for the architecure), which implies among other things the platform bootstrapping, MMU driver, page-table definitions, cache and TLB handling, timer driver, IRQ controller driver, system-call bindings, and the code paths for entering and exiting the kernel mode. - Extending the dynamic linker with the support for the architecture- specific parts of the ELF binary (e.g., architecture-specific relocations, constructors and destructors) - Extending our (FreeBSD-based) C runtime with architecture-specific back-end code
This is just a very high-level view. To get a more concrete idea about the construction sites, it is instructive to look at the architecture-specific code of an already supported architecture, e.g., x86_64:
cd genode/repos find base base-hw os libports | grep x86_64
Regards Norman