Re: [PATCH 1/2] arm64: add HWCAP3_LS64_ACCDATA
From: Siddharth Chintamaneni
Date: Sun Nov 22 2026 - 14:42:24 EST
On Tue, 17 Nov 2026 at 14:22, Arnd Bergmann <arnd@xxxxxxxxxx> wrote:
>
> From: Arnd Bergmann <arnd@xxxxxxxx>
>
> LS64_ACCDATA extends the LS64 feature with the user-space ls64bv0 instruction
> and the EL1 ACCDATA register. Like LS64, it was added with Armv8.7, and it
> roughly corresponds to the 'enqcmd' feature for shared hardware workqueues
> on x86 CPUs as described in the kernel's SVA documentation. [1]
A few small notes on this patch (the first three were also highlighted by the
Sashiko review [1]):
- Nit: s/ls64bv0/st64bv0/ in the commit message.
- Should this come as Patch 2/2 after enabling st64bv0 in the kernel so
HWCAP3_LS64_ACCDATA is not advertised to userspace at an intermediate
commit?
>
> Userspace can check availability of HWCAP3_LS64_ACCDATA to see if the shared
> hardware workqueue feature is possible on a given driver.
>
> Link: https://www.kernel.org/doc/html/v7.3/arch/x86/sva.html#enqcmd [1]
> Signed-off-by: Arnd Bergmann <arnd@xxxxxxxx>
> ---
> Documentation/arch/arm64/elf_hwcaps.rst | 7 +++++++
> arch/arm64/include/uapi/asm/hwcap.h | 1 +
> arch/arm64/kernel/cpufeature.c | 1 +
> arch/arm64/kernel/cpuinfo.c | 1 +
> 4 files changed, 10 insertions(+)
>
> diff --git a/Documentation/arch/arm64/elf_hwcaps.rst b/Documentation/arch/arm64/elf_hwcaps.rst
> index f60ca5612daa..f63035696654 100644
> --- a/Documentation/arch/arm64/elf_hwcaps.rst
> +++ b/Documentation/arch/arm64/elf_hwcaps.rst
> @@ -478,6 +478,13 @@ HWCAP3_SVE_LUT6
> Functionality implied by ID_AA64PFR0_EL1.SVE == 0b0001 and
> ID_AA64ISAR2_EL1.LUT == 0b0010.
>
> +HWCAP3_LS64_ACCDATA
> + Functionality implied by ID_AA64ISAR1_EL1.LS64 >= 0b0011. Note that
> + the function of instruction st64bv0 requires support by CPU, system
> + and target (device) memory location and HWCAP3_LS64_ACCDATA implies
> + the support of CPU. User should only use st64bv0 on supported target
> + (device) memory location, otherwise fallback to the non-atomic
> + alternatives.
>
> 4. Unused AT_HWCAP bits
> -----------------------
> diff --git a/arch/arm64/include/uapi/asm/hwcap.h b/arch/arm64/include/uapi/asm/hwcap.h
> index 10272ddb4d6f..c85c445275f6 100644
> --- a/arch/arm64/include/uapi/asm/hwcap.h
> +++ b/arch/arm64/include/uapi/asm/hwcap.h
> @@ -155,5 +155,6 @@
> #define HWCAP3_F16F32DOT (1UL << 9)
> #define HWCAP3_F16F32MM (1UL << 10)
> #define HWCAP3_SVE_LUT6 (1UL << 11)
> +#define HWCAP3_LS64_ACCDATA (1UL << 12)
>
> #endif /* _UAPI__ASM_HWCAP_H */
> diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
> index 952f70a48190..bbd809c83a6c 100644
> --- a/arch/arm64/kernel/cpufeature.c
> +++ b/arch/arm64/kernel/cpufeature.c
> @@ -3226,6 +3226,7 @@ static const struct arm64_cpu_capabilities arm64_elf_hwcaps[] = {
> HWCAP_CAP(ID_AA64ISAR1_EL1, DGH, IMP, CAP_HWCAP, KERNEL_HWCAP_DGH),
> HWCAP_CAP(ID_AA64ISAR1_EL1, I8MM, IMP, CAP_HWCAP, KERNEL_HWCAP_I8MM),
> HWCAP_CAP(ID_AA64ISAR1_EL1, LS64, LS64, CAP_HWCAP, KERNEL_HWCAP_LS64),
> + HWCAP_CAP(ID_AA64ISAR1_EL1, LS64, LS64_ACCDATA, CAP_HWCAP, KERNEL_HWCAP_LS64_ACCDATA),
> HWCAP_CAP(ID_AA64ISAR2_EL1, LUT, IMP, CAP_HWCAP, KERNEL_HWCAP_LUT),
> HWCAP_CAP(ID_AA64ISAR3_EL1, FAMINMAX, IMP, CAP_HWCAP, KERNEL_HWCAP_FAMINMAX),
> HWCAP_CAP(ID_AA64ISAR3_EL1, LSFE, IMP, CAP_HWCAP, KERNEL_HWCAP_LSFE),
> diff --git a/arch/arm64/kernel/cpuinfo.c b/arch/arm64/kernel/cpuinfo.c
> index 94e6cd211d70..009de5255d2d 100644
> --- a/arch/arm64/kernel/cpuinfo.c
> +++ b/arch/arm64/kernel/cpuinfo.c
> @@ -158,6 +158,7 @@ static const char *const hwcap_str[] = {
> [KERNEL_HWCAP_F16F32DOT] = "f16f32dot",
> [KERNEL_HWCAP_F16F32MM] = "f16f32mm",
> [KERNEL_HWCAP_SVE_LUT6] = "svelut6",
> + [KERNEL_HWCAP_LS64_ACCDATA] = "ls64_accdata",
should we drop underscore to maintain consistency?
> };
>
> #ifdef CONFIG_COMPAT
> --
> 2.53.0
>
[1] https://sashiko.dev/#/patchset/20261117222224.1465328-1-arnd%40kernel.org