We have a few of the same nits as before: this list should be static, there should be no space after STAILQ_FOREACH, empty parameter lists should be void. I'll stop pointing them out.

Should we also check for overflow in gva + pointer->off?

To be paranoid, should there be a check that pointer->off + pointer->size <= table->len?

This trick with memcpy() assumes that the host is little-endian, if pointer->size < 8.

I've added an assertion for pointer->off < table->len. So, if gva + pointer->off overflows, vm_map_gpa(ctx, gpa, table->len) should fail. Am I missing something?

"sign" being its own word is slightly less readable (it's not immediately obvious that it is an abbreviation for signature. I would suggest either expanding "sign" to signature in various places here, or perhaps just using "name" instead (the constant is already ACPI_NAMESEG_SIZE)

This is more the style(9) rule about having all variables declared at the start of a block with a blank line before code.

You could use %.4s instead of the four %c and then just use sign as the argument. Just shorter to write.

Again, this isn't really about guest BIOS, but instead about guest kernels, etc.

I would instead do something like:

switch (pointer->size) {
case 4:
            val = le32dec(gva + pointer->off);
            break;
case 8:
            val = le64dec(gva + pointer->off);
            break;
}

This combines the memcpy + le*toh in one step.

Also, I would just assert() in basl_table_add_pointer that the pointer size is either 4 or 8.

This could use the helper routine I suggested previously.

%,4s here as well

Thanks.