From: James Hogan <james.hogan@mips.com>

We shouldn't set the ISA bit in CP0_EPC for nanoMIPS.

Signed-off-by: James Hogan <james.hogan@mips.com>
Signed-off-by: Yongbok Kim <yongbok.kim@mips.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Signed-off-by: Stefan Markovic <smarkovic@wavecomp.com>
---
 target/mips/helper.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/target/mips/helper.c b/target/mips/helper.c
index dc8f2a5..fb47018 100644
--- a/target/mips/helper.c
+++ b/target/mips/helper.c
@@ -656,7 +656,8 @@ target_ulong exception_resume_pc (CPUMIPSState *env)
     target_ulong bad_pc;
     target_ulong isa_mode;
 
-    isa_mode = !!(env->hflags & MIPS_HFLAG_M16);
+    isa_mode = env->hflags & MIPS_HFLAG_M16 &&
+                !(env->insn_flags & ISA_NANOMIPS32);
     bad_pc = env->active_tc.PC | isa_mode;
     if (env->hflags & MIPS_HFLAG_BMASK) {
         /* If the exception was raised from a delay slot, come back to
-- 
2.7.4

On 07/19/2018 05:55 AM, Stefan Markovic wrote:
> From: James Hogan <james.hogan@mips.com>
> 
> We shouldn't set the ISA bit in CP0_EPC for nanoMIPS.
> 
> Signed-off-by: James Hogan <james.hogan@mips.com>
> Signed-off-by: Yongbok Kim <yongbok.kim@mips.com>
> Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
> Signed-off-by: Stefan Markovic <smarkovic@wavecomp.com>
> ---
>  target/mips/helper.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/target/mips/helper.c b/target/mips/helper.c
> index dc8f2a5..fb47018 100644
> --- a/target/mips/helper.c
> +++ b/target/mips/helper.c
> @@ -656,7 +656,8 @@ target_ulong exception_resume_pc (CPUMIPSState *env)
>      target_ulong bad_pc;
>      target_ulong isa_mode;
>  
> -    isa_mode = !!(env->hflags & MIPS_HFLAG_M16);
> +    isa_mode = env->hflags & MIPS_HFLAG_M16 &&
> +                !(env->insn_flags & ISA_NANOMIPS32);

Is there a compelling reason to have MIPS_HFLAG_M16 *set* for ISA_NANOMIPS?

This seems like the 6th or 7th patch that works around M16.  If M16 were always
unset for nanomips, it would have avoided a lot of extra effort, it would seem.
 In addition, looking at the fully patched tree and grepping for HFLAG_M16,
there appear to be further uses that have been missed for the extra NANOMIPS check.


r~