How do you know offset 12 on the stack is a good place to store this? I'd expect there to be something like:
cprestore will store gp to <offset>(sp) and then load it from there after every jal operation to restore gp.

I'm also at a loss for how this code could possibly work: there's no delay slot instruction after the JAL and there's no jr ra. But I'm willing to suspend my disbelief on this one: All hail mips assembler magic.

At the very least we need to add a addiu sp,sp,-16 after the .cpload and before the .cprestore. But then we'd need a addiu sp,sp,16 after the above missing jr ra....

At least according to https://techpubs.jurassic.nl/manuals/0530/developer/Cplr_PTG/sgi_html/apa.html which appears to be the reproduction of the sgi assembler manual from whence all this stuff came.

If prefer to just spell out the assembly rather than just the magic cp* macros.
Apparently they place a $gp reload after every call: Additionally, it causes the assembler to emit:

lw     gp,offset(sp)

after every jump-and-link (jal) operation (but not after a branch-and-link (bal) operation), thereby restoring the gp register after function calls. The programmer is responsible for allocating the stack space for the gp.

And it does seem like there is a nop missing after the JAL since we do a .set noreorder.

Do we actually use this code for MIPS? I think cheribsd still uses the old CSU bits.

In D21697#475800, @arichardson wrote:

If prefer to just spell out the assembly rather than just the magic cp* macros.

Except that's not really an option. These are extra special directives, in addition to being a macro and there has to be matched pairs.

Apparently they place a $gp reload after every call: Additionally, it causes the assembler to emit:

lw     gp,offset(sp)

after every jump-and-link (jal) operation (but not after a branch-and-link (bal) operation), thereby restoring the gp register after function calls. The programmer is responsible for allocating the stack space for the gp.

Correct, which we've not done here, it seems from my reading of their use, but with so many levels of indirections, it's hard to know for sure.

And it does seem like there is a nop missing after the JAL since we do a .set noreorder.

Do we actually use this code for MIPS? I think cheribsd still uses the old CSU bits.

I think that we do. Otherwise we wouldn't have hit this error we need to fix...