These instructions use zero as the discriminator, not SP.

Signed-off-by: Peter Collingbourne <pcc@google.com>
---
 target/arm/translate-a64.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c
index 8c0764957c..c996ca1393 100644
--- a/target/arm/translate-a64.c
+++ b/target/arm/translate-a64.c
@@ -3429,9 +3429,11 @@ static void disas_ldst_pac(DisasContext *s, uint32_t insn,
 
     if (s->pauth_active) {
         if (use_key_a) {
-            gen_helper_autda(dirty_addr, cpu_env, dirty_addr, cpu_X[31]);
+            gen_helper_autda(dirty_addr, cpu_env, dirty_addr,
+                             new_tmp_a64_zero(s));
         } else {
-            gen_helper_autdb(dirty_addr, cpu_env, dirty_addr, cpu_X[31]);
+            gen_helper_autdb(dirty_addr, cpu_env, dirty_addr,
+                             new_tmp_a64_zero(s));
         }
     }
 
-- 
2.28.0.163.g6104cc2f0b6-goog

On Mon, Aug 3, 2020 at 3:27 PM Peter Collingbourne <pcc@google.com> wrote:
>
> These instructions use zero as the discriminator, not SP.

Oh, there is no such thing as STRAA/STRAB. I must have been confused
by the name of the function, disas_ldst_pac. I will send a v2 with a
fixed commit message, and another patch to rename the function to
disas_ld_pac.

Peter

>
> Signed-off-by: Peter Collingbourne <pcc@google.com>
> ---
>  target/arm/translate-a64.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c
> index 8c0764957c..c996ca1393 100644
> --- a/target/arm/translate-a64.c
> +++ b/target/arm/translate-a64.c
> @@ -3429,9 +3429,11 @@ static void disas_ldst_pac(DisasContext *s, uint32_t insn,
>
>      if (s->pauth_active) {
>          if (use_key_a) {
> -            gen_helper_autda(dirty_addr, cpu_env, dirty_addr, cpu_X[31]);
> +            gen_helper_autda(dirty_addr, cpu_env, dirty_addr,
> +                             new_tmp_a64_zero(s));
>          } else {
> -            gen_helper_autdb(dirty_addr, cpu_env, dirty_addr, cpu_X[31]);
> +            gen_helper_autdb(dirty_addr, cpu_env, dirty_addr,
> +                             new_tmp_a64_zero(s));
>          }
>      }
>
> --
> 2.28.0.163.g6104cc2f0b6-goog
>

On 8/3/20 5:21 PM, Peter Collingbourne wrote:
> On Mon, Aug 3, 2020 at 3:27 PM Peter Collingbourne <pcc@google.com> wrote:
>>
>> These instructions use zero as the discriminator, not SP.
> 
> Oh, there is no such thing as STRAA/STRAB. I must have been confused
> by the name of the function, disas_ldst_pac. I will send a v2 with a
> fixed commit message, and another patch to rename the function to
> disas_ld_pac.

It's called decode_ldst_pac because the Arm ARM section is called "Load/store
register (pac)".  Page C4-311 in the F.a revision.

But yes, there are only loads defined in the section.


r~

On Tue, Aug 4, 2020 at 8:41 AM Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> On 8/3/20 5:21 PM, Peter Collingbourne wrote:
> > On Mon, Aug 3, 2020 at 3:27 PM Peter Collingbourne <pcc@google.com> wrote:
> >>
> >> These instructions use zero as the discriminator, not SP.
> >
> > Oh, there is no such thing as STRAA/STRAB. I must have been confused
> > by the name of the function, disas_ldst_pac. I will send a v2 with a
> > fixed commit message, and another patch to rename the function to
> > disas_ld_pac.
>
> It's called decode_ldst_pac because the Arm ARM section is called "Load/store
> register (pac)".  Page C4-311 in the F.a revision.
>
> But yes, there are only loads defined in the section.

I see. Arguably the ARM ARM section is misnamed then. There is a
sibling section named "Load register (literal)", so there is precedent
for naming a section after the types of instructions that are actually
supported. I will send mail to errata@arm.com to see if the section
can be renamed.

Peter