Clarify textra behavior when XLEN changes.

[rtwfroody]

No description provided.

[pdonahue-ventana]

This seems reasonable.

Writes to this register will clear the high bits of the corresponding fields in {csr-textra64}.

If you do something like csrrsi x0, textra32, 1 then it will set the LSB. Are csrrs, csrrc, csrrsi, csrrci considered to be writes to this register? My inclination is that they are writes because the 32b mode is basically claiming ownership and it might expect that csrrs of value X followed by csrrc of ~X should be equivalent to csrrw of X. It would be good to explicitly say that the set/clear operations are writes.

[rtwfroody]

From the unpriv spec, section 11.1: "All CSR instructions atomically read-modify-write a single CSR"

To me that unambiguously says all those ops, including csrrsi are writes.

Note that we already have similar language in our spec. Section 5.7 says:

In this section XLEN means MXLEN when in M-mode, and DXLEN when in Debug Mode. On systems
where those values of XLEN can differ, this is handled as follows. Fields retain their values regardless
of XLEN, which only affects where in the register these fields appear (e.g. type). Some fields are wider
when XLEN is 64 than when it is 32 (e.g. svalue). The high bits in such fields retain their value but are
not readable when XLEN is 32. A modification of a register when XLEN is 32 clears any inaccessible
bits in that register.

We didn't mention any specific csr instructions there either.

[pdonahue-ventana]

OK, sure. I just like to think about all the weird ways that people can interpret the spec. This just talks about "writes to this register" and not "csrrw to this register" so I guess that it's reasonably clear.