RISC-V: add minidump support

[mhorne063_gmail.com]

Summary

Add the required functions for minidump support, based on the ARM64 port.

[mhorne063_gmail.com]

Did some of this work before I realized the obvious blocker that we don't have kgdb for RISC-V yet. It dumps successfully but I can't inspect it yet until we have that. Regardless, I figured I'd put this up now to get some initial looks at the changes. I will look into adding the required bits to kgdb to get it working.

sys/riscv/riscv/minidump_machdep.c

235

First major differences to the ARM64 version are in this loop.

This is built on my understanding that we support 3 levels of page tables (L1 through L3), and that L2 and L3 level entries can be leaf entries.

327

Second major differences are in this loop.

sys/riscv/riscv/pmap.c

663

Setting up dump_avail is required so that pmap_page_dump_size will be non-zero (see vm_page.c).

sys/vm/vm_page.c

661–662

Do we need a || defined(__riscv) here?

[mhorne063_gmail.com]

Small update fixing some extra #ifdefs.

Ripped out the initialization of dump_avail[] to a separate review (D19170).

[mhorne063_gmail.com]

Fix a small bug while dumping pagetable entries.

Reading dumps seems to now be working without issue. Take a look at P247 for a sample from crashinfo(8).

[markj]

Looks ok, just a couple of nits.

Am I right that we only write 4KB at a time?

sys/riscv/riscv/uma_machdep.c

57–58

No need to keep the extra newlines.

72–73

Ditto.

sys/vm/vm_page.c

661–662

Yes, since riscv uses a direct map.

[mhorne063_gmail.com]

In D18320#415484, @markj wrote:

Am I right that we only write 4KB at a time?

Yes. I tried writing out more than one page at a time and it failed quite spectacularly, so this could be a good spot to revisit and introduce pmap_kenter_temporary() as we discussed on IRC.

[mhorne063_gmail.com]

Fix whitespace.

[markj]

In D18320#415796, @mhorne063_gmail.com wrote:

In D18320#415484, @markj wrote:

Am I right that we only write 4KB at a time?

Yes. I tried writing out more than one page at a time and it failed quite spectacularly, so this could be a good spot to revisit and introduce pmap_kenter_temporary() as we discussed on IRC.

I'm happy for that to be handled in a separate diff/commit; just wanted to make sure I wasn't missing something. It might be worth adding an /* XXX */ comment for now noting the 4KB write limitation.

[mhorne063_gmail.com]

Add advisory comment to blk_write().