Support the MangoPi MQ-Pro

[IsaacWoods]

Getting the kernel running

• Get OpenSBI and Seed booting via FEL (Boot on real RISC-V hardware - the MangoPi MQ-Pro #99)
• Support loading files in Seed from a ramdisk loaded into memory with FEL
• Support the Sv39 paging model (as this is all the D1 supports :()
• Generalise seed_riscv's ELF loading to any source of files
• Successfully boot into the kernel!

Boot from persistent media

• Work out how to update SD cards without just writing a whole new disk image to them (or do we care?)
• Get a test boot0 running from SRAM that can output to the UART running from the SD card
• Initialize the DRAM in our boot0
• Load OpenSBI and Seed from the SD card in boot0
• Successfully run OpenSBI + Seed
• Write SDHC driver for Seed
• Load kernel from SD card and boot into it!
• Work out how to fit the weird headers etc. into a nice GPT scheme

[IsaacWoods]

Adding a note here to hopefully remember (we're not going to be near our workstation for a while :( ). I think we need to do a VMA fence after switching to new sets of page tables - I don't think this is going to be what's causing issues on the D1, but is the correct thing to do.

[IsaacWoods]

Also some thoughts about debugging on real hardware - I wonder if we could get some kind of timer running in M-mode (with an altered OpenSBI I guess) that pings back a set of registers / other info over UART. This would be a useful extra tool for dealing with stuff like the core seemingly freezing when entering usermode (although we need to have a look at our trap handling to rule out dumb mistakes first).

Could also look at the ebreak instruction to see if we can do any instrumentation through that via M-mode too.