Re: [PATCH] arm64: entry: drop the unused tmp2 argument of mte_set_kernel_gcr()

From: Bradley Morgan

Date: Wed Oct 07 2026 - 13:51:53 EST


On 7 October 2026 18:45:01 BST, Bradley Morgan <brads@xxxxxxxxxxxxxx>
wrote:
>mte_set_kernel_gcr() takes a tmp2 argument the macro never uses. It
>used to, the body passed \tmp and \tmp2 to mte_set_gcr, then
>commit 82868247897b ("arm64: kasan: mte: use a constant kernel
>GCR_EL1 value") swapped that for
>
> mov \tmp, KERNEL_GCR_EL1
>
>and tmp2 stayed in the declaration and in the one caller, which
>passes x23 for it.
>
>Drop it, both places.
>
>Signed-off-by: Bradley Morgan <brads@xxxxxxxxxxxxxx>
>
>diff --git a/arch/arm64/kernel/entry.S b/arch/arm64/kernel/entry.S
>index 0e1d9c3c6a93..2baefe7a82d3 100644
>--- a/arch/arm64/kernel/entry.S
>+++ b/arch/arm64/kernel/entry.S
>@@ -171,7 +171,7 @@
> #endif
> .endm
>
>- .macro mte_set_kernel_gcr, tmp, tmp2
>+ .macro mte_set_kernel_gcr, tmp
> #ifdef CONFIG_KASAN_HW_TAGS
> alternative_cb ARM64_ALWAYS_SYSTEM, kasan_hw_tags_enable
> b 1f
>@@ -258,7 +258,7 @@
>
> apply_ssbd 1, x22, x23
>
>- mte_set_kernel_gcr x22, x23
>+ mte_set_kernel_gcr x22
>
> /*
> * Any non-self-synchronizing system register updates required for
>---

AHH. What happened here, ugh! Why was it sent twice, stupid me man. Let me
go resend it PROPERLY, god sake Bradley..

> arch/arm64/kernel/entry.S | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
>diff --git a/arch/arm64/kernel/entry.S b/arch/arm64/kernel/entry.S
>index f63049ac32dc..c3106dee61e1 100644
>--- a/arch/arm64/kernel/entry.S
>+++ b/arch/arm64/kernel/entry.S
>@@ -171,7 +171,7 @@ alternative_else_nop_endif
> #endif
> .endm
>
>- .macro mte_set_kernel_gcr, tmp, tmp2
>+ .macro mte_set_kernel_gcr, tmp
> #ifdef CONFIG_KASAN_HW_TAGS
> alternative_cb ARM64_ALWAYS_SYSTEM, kasan_hw_tags_enable
> b 1f
>@@ -258,7 +258,7 @@ alternative_else_nop_endif
>
> apply_ssbd 1, x22, x23
>
>- mte_set_kernel_gcr x22, x23
>+ mte_set_kernel_gcr x22
>
> /*
> * Any non-self-synchronizing system register updates required for
>

--- Thanks!
"I'm not a very positive person" - Linus torvalds