Re: [PATCH v5 3/6] arm64: hyperv: Add per-CPU RSI host call infrastructure for CCA Realms

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] arm64: hyperv: kcalloc() violates RSI host call 256-byte alignment
--

commit 15aadbf7702767ee71f473eab885a0900dbd0a76
Author: Kameron Carr <kameroncarr@linux.microsoft.com>

arm64: hyperv: Add per-CPU RSI host call infrastructure for CCA Realms

This commit adds a dynamically allocated array of per-CPU `struct
rsi_host_call` structures for ARM CCA Realms. The memory is allocated using
kcalloc() during hyperv_init() to avoid paying the memory cost on non-Realm
guests.

> diff --git a/arch/arm64/hyperv/mshyperv.c b/arch/arm64/hyperv/mshyperv.c
> index 4fdc26ade1d74..b1657cfabd7d1 100644
> --- a/arch/arm64/hyperv/mshyperv.c
> +++ b/arch/arm64/hyperv/mshyperv.c
[ ... ]
> @@ -85,6 +90,20 @@ static int __init hyperv_init(void)
>  	if (!hyperv_detect_via_acpi() && !hyperv_detect_via_smccc())
>  		return 0;
>  
> +	/*
> +	 * The RSI host-call buffers are only ever used when
> +	 * is_realm_world() is true. Skip the allocation on non-Realm
> +	 * guests. A single contiguous array of nr_cpu_ids entries is
> +	 * allocated; each CPU indexes into it by its processor ID.
> +	 */
> +	if (is_realm_world()) {
> +		hv_hostcall_array = kcalloc(nr_cpu_ids,
> +					    sizeof(struct rsi_host_call),
> +					    GFP_KERNEL);

[Severity: High]
Does kcalloc() satisfy the __aligned(256) requirement of struct rsi_host_call
when SLUB debugging or KASAN is enabled?

The struct definition explicitly demands 256-byte alignment:

include/linux/arm-smccc-rsi.h:
struct rsi_host_call {
	u16 immediate;
	u8 _padding[6];
	u64 gprs[31];
} __aligned(256);

kcalloc() only receives the size of the structure, not the alignment
constraint. When SLUB debugging is enabled, redzoning metadata is prepended to
the object, shifting the base pointer off the 256-byte boundary. A misaligned
Intermediate Physical Address (IPA) might lead to hardware rejection of the
hypercall by the Realm Management Monitor (RMM).

Additionally, since this allocates a single contiguous array for all CPUs,
does it bypass NUMA locality? Using alloc_percpu() would satisfy the alignment
constraint natively and allocate memory on local nodes.

> +		if (!hv_hostcall_array)
> +			return -ENOMEM;
> +	}
> +

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261002215725.2015959-1-kameroncarr@linux.microsoft.com?part=3