1
0
0
The Talos platform already has an open source bootloader, called Hostboot, so why port coreboot to it? Well, there are many reasons: licence, programming language, hard to understand codebase.
Talos is a POWER9 CPU platform, that looks like a regular motherboard.
2
0
0
The POWER9 SoC has many Power9 and MCU (usually Power5) cores. Starting it is complex and has many steps.
1
0
0
The reset and initialization vectors are complex and cumbersome with many moving parts.
Debugging is essential, so QEMU was used, although it's nowhere closed to real hardware. There is also a BMC pdbg interface.
POWER9 is an ISA that has standard stuff, except the part where it can use both big and little endian, with no CPU reset between mode switching, and different piece of software might use different modes, which is a source of bugs.
1
0
0
While porting coreboot and reading the POWER9 documentation, Krystian took notes and published them at https://github.com/3mdeb/openpower-coreboot-docs to help future developers.
1
0
1
Some shortcuts had to be taken to only work in Krystian's usecase: workstation (not servers), less RAS support, support only DIMM memory, etc.
(Which is pretty standard when porting code to a new platform IMHO)
1
0
0
Currently, the port is mostly done with Linux already booting, and is being slowly split in review-able pieces for upstreaming in coreboot.
0
0
0
Jarkko Sakkinen
jarkkoHaving worked with RISC-V for over past year or so (and continuing to do so albeit as byte code in PolkaVM), at this point of time I don't have a lot of faith in it when it comes to "with MMU" use cases.
This has made me hope that OpenMIPS would get some air under its wings (commercially) in such use cases both because RISC-V does not really deliver in that area, and secondly because competition would enforce it to deliver. Maybe there is also just some plain love for its legacy (SGI, Playstation 2, Nintendo 64 for etc) ;-)
0
0
0