dtrace: add register bindings for RISC-V

[christos]

Forgot riscv_regs.d

[mhorne]

cddl/lib/libdtrace/regs_riscv.d

25–26

S0 is also the Frame Pointer. Can you add another entry giving it the alias R_FP?

sys/cddl/dev/dtrace/riscv/dtrace_isa.c

324

I think specifying a range in this way is a GNU extension, not standard C syntax, and it should be avoided. Unfortunately that makes this code a lot uglier :(

[christos]

cddl/lib/libdtrace/regs_riscv.d

25–26

Yeah. I had initially added R_FP as well but I rewrote the file and forgot to add it.

sys/cddl/dev/dtrace/riscv/dtrace_isa.c

324

There are other code pieces that use this as well, I don't think that's too much of an issue,
but if you and @markj insist, I can get rid of it.

[christos]

Add FP.

[markj]

cddl/lib/libdtrace/regs_riscv.d

9

Technically this should be 1.13, not 1.0. It comes from DT_VERS_STRING in libdtrace.

sys/cddl/dev/dtrace/riscv/dtrace_isa.c

324

It is a GNU extension, but we do use it a fair bit already[*], especially in bhyve. I'd vote for keeping it since it makes the code easier to follow.

[*]$ git grep 'case.*\.\.\..*:'

[christos]

sys/cddl/dev/dtrace/riscv/dtrace_isa.c

324

I'll leave it as is.

[christos]

cddl/lib/libdtrace/regs_riscv.d

9

Should the version in regs_x86.d be bumped as well?

[christos]

Bump version to 1.13.

[mhorne]

sys/cddl/dev/dtrace/riscv/dtrace_isa.c

324

Sounds great.

336–341

Nit: the final return statement will never execute due to the default case. You can use __assert_unreachable() instead.

[christos]

sys/cddl/dev/dtrace/riscv/dtrace_isa.c

336–341

Won't a /* NOTREACHED */ comment suffice?

[christos]

Add /* NOTREACHED */ at the end of dtrace_getreg().

[markj]

cddl/lib/libdtrace/regs_riscv.d

9

No, the binding reflects the current libdtrace version at the time that the definitions were introduced. The x86 register bindings have been there since the beginning.