Retire the GENERICSF kernel config.

[jhb]

Summary

Now that hw.machine_arch handles soft-float vs hard-float there is no
longer a reason for this config.

[jhb]

sys/conf/Makefile.riscv

22 ↗

(On Diff #70899)

This means you can 'make tinderbox' on stable/12 again as well (right now it terminates early with an error since the host config is too old when it tries to run 'config -m' on each candidate kernel config)

[imp]

sys/conf/Makefile.riscv

22 ↗

(On Diff #70899)

I should MFC the config changes to there then, but don't take that as any reason to not revert the version bump...

[mhorne]

The following excerpt from kern.mk should also be cleaned up as part of this commit:

.if ${MACHINE_ARCH:Mriscv*sf}
CFLAGS+=        -march=rv64imac
.else
CFLAGS+=        -march=rv64imafdc
.endif

Otherwise, I think compilation will fail when TARGET_ARCH=riscv64sf and the FPE option is enabled.

[jhb]

In D24544#540448, @mhorne wrote:

The following excerpt from kern.mk should also be cleaned up as part of this commit:

.if ${MACHINE_ARCH:Mriscv*sf}
CFLAGS+=        -march=rv64imac
.else
CFLAGS+=        -march=rv64imafdc
.endif

Otherwise, I think compilation will fail when TARGET_ARCH=riscv64sf and the FPE option is enabled.

FWIW, I think the right answer is that kernels should always be riscv64 and only the userland ABI varies (so riscv64sf is really a userland-only ABI if that makes sense), but yes, we can probably remove that bit as well.

[jhb]

• Drop riscv64sf special case from kern.mk.

[jhb]

Ping (especially @kp since you added this)