On Fri, 10 Nov 2017, James Hogan wrote:
> Building 32-bit MIPS64r2 kernels produces warnings like the following
> on certain toolchains (such as GNU assembler 2.24.90, but not GNU
> assembler 2.28.51) since commit 22b8ba765a72 ("MIPS: Fix MIPS64 FP
> save/restore on 32-bit kernels"), due to the exposure of fpu_save_16odd
> from fpu_save_double and fpu_restore_16odd from fpu_restore_double:
>
> arch/mips/kernel/r4k_fpu.S:47: Warning: float register should be even, was 1
> ...
> arch/mips/kernel/r4k_fpu.S:59: Warning: float register should be even, was 1
> ...
Hmm, versions 2.24.90 and 2.28.51 are otherwise unindentified development
snapshots; I think it would be slightly more appropriate if you referred
to actual release versions, such as 2.25 and 2.29 respectively (if they
indeed expose the same symptoms), especially as people other than
toolchain developers and testers are generally expected not to use
development snapshots.
Also I find it suspicious that you say that the message has since
vanished, as I can clearly reproduce it with current head (2.29.51):
$ cat oddfpr.s
.module mips64r2
foo:
ldc1 $f1, 0($2)
$ as -o oddfpr.o oddfpr.s
oddfpr.s: Assembler messages:
oddfpr.s:3: Warning: float register should be even, was 1
$
Can you send me a .s file produced from r4k_fpu.S along with compiler
flags used?
> This appears to be because .set mips64r2 does not change the FPU ABI to
> 64-bit when -march=mips64r2 (or e.g. -march=xlp) is provided on the
> command line on that toolchain, from the default FPU ABI of 32-bit due
> to the -mabi=32. This makes access to the odd FPU registers invalid.
Correct, the purpose of `.set arch=mips64r2', which is the canonical form
`.set mips64r2' is equivalent to, and indeed any `.set arch=...'
pseudo-op, is only to change the set of instructions accepted within the
limits determined by the ABI chosen with `-mabi=...' (or its equivalent
short forms) and not the ABI itself.
The patch itself LGTM.
Reviewed-by: Maciej W. Rozycki <macro@xxxxxxxx>
Maciej