[PATCH] target/arm: mark SP_EL1 with ARM_CP_EL3_NO_EL2_KEEP

[PATCH] target/arm: mark SP_EL1 with ARM_CP_EL3_NO_EL2_KEEP

[Jerome Forissier]

SP_EL1 must be kept when EL3 is present but EL2 is not. Therefore mark
it with ARM_CP_EL3_NO_EL2_KEEP.

Fixes: 696ba3771894 ("target/arm: Handle cpreg registration for missing EL")
Signed-off-by: Jerome Forissier <jerome.forissier@linaro.org>
---
 target/arm/helper.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/target/arm/helper.c b/target/arm/helper.c
index d7bc467a2a..328d252a84 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -4971,7 +4971,7 @@ static const ARMCPRegInfo v8_cp_reginfo[] = {
       .fieldoffset = offsetof(CPUARMState, sp_el[0]) },
     { .name = "SP_EL1", .state = ARM_CP_STATE_AA64,
       .opc0 = 3, .opc1 = 4, .crn = 4, .crm = 1, .opc2 = 0,
-      .access = PL2_RW, .type = ARM_CP_ALIAS,
+      .access = PL2_RW, .type = ARM_CP_ALIAS | ARM_CP_EL3_NO_EL2_KEEP,
       .fieldoffset = offsetof(CPUARMState, sp_el[1]) },
     { .name = "SPSel", .state = ARM_CP_STATE_AA64,
       .opc0 = 3, .opc1 = 0, .crn = 4, .crm = 2, .opc2 = 0,
-- 
2.34.1

Re: [PATCH] target/arm: mark SP_EL1 with ARM_CP_EL3_NO_EL2_KEEP

[Richard Henderson]

On 9/27/22 05:00, Jerome Forissier wrote:
> SP_EL1 must be kept when EL3 is present but EL2 is not. Therefore mark
> it with ARM_CP_EL3_NO_EL2_KEEP.
> 
> Fixes: 696ba3771894 ("target/arm: Handle cpreg registration for missing EL")
> Signed-off-by: Jerome Forissier <jerome.forissier@linaro.org>

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

This certainly looks right, as how else would EL3 set up EL1 as desired. I suppose this 
error comes from a slight mis-read of R_RJFFP, in that SP_EL1 may be considered 
"associated" with EL1 and not EL[23].

r~

> ---
>   target/arm/helper.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/target/arm/helper.c b/target/arm/helper.c
> index d7bc467a2a..328d252a84 100644
> --- a/target/arm/helper.c
> +++ b/target/arm/helper.c
> @@ -4971,7 +4971,7 @@ static const ARMCPRegInfo v8_cp_reginfo[] = {
>         .fieldoffset = offsetof(CPUARMState, sp_el[0]) },
>       { .name = "SP_EL1", .state = ARM_CP_STATE_AA64,
>         .opc0 = 3, .opc1 = 4, .crn = 4, .crm = 1, .opc2 = 0,
> -      .access = PL2_RW, .type = ARM_CP_ALIAS,
> +      .access = PL2_RW, .type = ARM_CP_ALIAS | ARM_CP_EL3_NO_EL2_KEEP,
>         .fieldoffset = offsetof(CPUARMState, sp_el[1]) },
>       { .name = "SPSel", .state = ARM_CP_STATE_AA64,
>         .opc0 = 3, .opc1 = 0, .crn = 4, .crm = 2, .opc2 = 0,

Re: [PATCH] target/arm: mark SP_EL1 with ARM_CP_EL3_NO_EL2_KEEP

[Peter Maydell]

On Wed, 28 Sept 2022 at 15:56, Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> On 9/27/22 05:00, Jerome Forissier wrote:
> > SP_EL1 must be kept when EL3 is present but EL2 is not. Therefore mark
> > it with ARM_CP_EL3_NO_EL2_KEEP.
> >
> > Fixes: 696ba3771894 ("target/arm: Handle cpreg registration for missing EL")
> > Signed-off-by: Jerome Forissier <jerome.forissier@linaro.org>
>
> Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
>
> This certainly looks right, as how else would EL3 set up EL1 as desired. I suppose this
> error comes from a slight mis-read of R_RJFFP, in that SP_EL1 may be considered
> "associated" with EL1 and not EL[23].

Oops. Applied to target-arm.next, and
Cc: qemu-stable@nongnu.org
since this bug made it into 7.1 :-(

thanks
-- PMM