Message-ID: <8634bdbgaz.wl-maz@kernel.org>
Date: Thu, 03 Jul 2025 13:38:12 +0100
From: Marc Zyngier <maz@...nel.org>
To: perlarsen@...gle.com
Cc: Oliver Upton <oliver.upton@...ux.dev>,
	Joey Gouly <joey.gouly@....com>,
	Suzuki K Poulose <suzuki.poulose@....com>,
	Zenghui Yu <yuzenghui@...wei.com>,
	Catalin Marinas <catalin.marinas@....com>,
	Will Deacon <will@...nel.org>,
	Sudeep Holla <sudeep.holla@....com>,
	linux-arm-kernel@...ts.infradead.org,
	kvmarm@...ts.linux.dev,
	linux-kernel@...r.kernel.org,
	ahomescu@...gle.com,
	armellel@...gle.com,
	arve@...roid.com,
	ayrton@...gle.com,
	qperret@...gle.com,
	sebastianene@...gle.com,
	qwandor@...gle.com
Subject: Re: [PATCH v7 2/5] KVM: arm64: Use SMCCC 1.2 for FF-A initialization and in host handler

On Tue, 01 Jul 2025 23:06:35 +0100,
Per Larsen via B4 Relay <devnull+perlarsen.google.com@...nel.org> wrote:
> 
> From: Per Larsen <perlarsen@...gle.com>
> 
> SMCCC 1.1 and prior allows four registers to be sent back as a result
> of an FF-A interface. SMCCC 1.2 increases the number of results that can
> be sent back to 8 and 16 for 32-bit and 64-bit SMC/HVCs respectively.
> 
> FF-A 1.0 references SMCCC 1.2 (reference [4] on page xi) and FF-A 1.2
> explicitly requires SMCCC 1.2 so it should be safe to use this version
> unconditionally. Moreover, it is simpler to implement FF-A features
> without having to worry about compatibility with SMCCC 1.1 and older.
> 
> Update the FF-A initialization and host handler code to use SMCCC 1.2.
> 
> Signed-off-by: Per Larsen <perlarsen@...gle.com>
> ---
>  arch/arm64/kvm/hyp/nvhe/Makefile |   1 +
>  arch/arm64/kvm/hyp/nvhe/ffa.c    | 193 +++++++++++++++++++++++++--------------
>  2 files changed, 125 insertions(+), 69 deletions(-)
> 
> diff --git a/arch/arm64/kvm/hyp/nvhe/Makefile b/arch/arm64/kvm/hyp/nvhe/Makefile
> index a76522d63c3e630795db5972a99abc3d24bc5e26..f859a8fb41a25effea1edd977bef889423153399 100644
> --- a/arch/arm64/kvm/hyp/nvhe/Makefile
> +++ b/arch/arm64/kvm/hyp/nvhe/Makefile
> @@ -27,6 +27,7 @@ hyp-obj-y := timer-sr.o sysreg-sr.o debug-sr.o switch.o tlb.o hyp-init.o host.o
>  	 cache.o setup.o mm.o mem_protect.o sys_regs.o pkvm.o stacktrace.o ffa.o
>  hyp-obj-y += ../vgic-v3-sr.o ../aarch32.o ../vgic-v2-cpuif-proxy.o ../entry.o \
>  	 ../fpsimd.o ../hyp-entry.o ../exception.o ../pgtable.o
> +hyp-obj-y += ../../../kernel/smccc-call.o
>  hyp-obj-$(CONFIG_LIST_HARDENED) += list_debug.o
>  hyp-obj-y += $(lib-objs)
>  
> diff --git a/arch/arm64/kvm/hyp/nvhe/ffa.c b/arch/arm64/kvm/hyp/nvhe/ffa.c
> index 2c199d40811efb5bfae199c4a67d8ae3d9307357..65d241ba32403d014b43cc4ef4d5bf9693813809 100644
> --- a/arch/arm64/kvm/hyp/nvhe/ffa.c
> +++ b/arch/arm64/kvm/hyp/nvhe/ffa.c
> @@ -71,36 +71,68 @@ static u32 hyp_ffa_version;
>  static bool has_version_negotiated;
>  static hyp_spinlock_t version_lock;
>  
> -static void ffa_to_smccc_error(struct arm_smccc_res *res, u64 ffa_errno)
> +static void ffa_to_smccc_error(struct arm_smccc_1_2_regs *res, u64 ffa_errno)
>  {
> -	*res = (struct arm_smccc_res) {
> +	*res = (struct arm_smccc_1_2_regs) {
>  		.a0	= FFA_ERROR,
>  		.a2	= ffa_errno,
>  	};
>  }
>  
> -static void ffa_to_smccc_res_prop(struct arm_smccc_res *res, int ret, u64 prop)
> +static void ffa_to_smccc_res_prop(struct arm_smccc_1_2_regs *res, int ret, u64 prop)
>  {
>  	if (ret == FFA_RET_SUCCESS) {
> -		*res = (struct arm_smccc_res) { .a0 = FFA_SUCCESS,
> -						.a2 = prop };
> +		*res = (struct arm_smccc_1_2_regs) { .a0 = FFA_SUCCESS,
> +						      .a2 = prop };
>  	} else {
>  		ffa_to_smccc_error(res, ret);
>  	}
>  }
>  
> -static void ffa_to_smccc_res(struct arm_smccc_res *res, int ret)
> +static void ffa_to_smccc_res(struct arm_smccc_1_2_regs *res, int ret)
>  {
>  	ffa_to_smccc_res_prop(res, ret, 0);
>  }
>  
>  static void ffa_set_retval(struct kvm_cpu_context *ctxt,
> -			   struct arm_smccc_res *res)
> +			   struct arm_smccc_1_2_regs *res)
>  {
> +	DECLARE_REG(u64, func_id, ctxt, 0);
>  	cpu_reg(ctxt, 0) = res->a0;
>  	cpu_reg(ctxt, 1) = res->a1;
>  	cpu_reg(ctxt, 2) = res->a2;
>  	cpu_reg(ctxt, 3) = res->a3;
> +	cpu_reg(ctxt, 4) = res->a4;
> +	cpu_reg(ctxt, 5) = res->a5;
> +	cpu_reg(ctxt, 6) = res->a6;
> +	cpu_reg(ctxt, 7) = res->a7;

From DEN0028G 2.6:

<quote>
Registers W4-W7 must be preserved unless they contain results, as
specified in the function definition.
</quote>

On what grounds can you blindly change these registers?

> +
> +	/*
> +	 * DEN0028C 2.6: SMC32/HVC32 call from aarch64 must preserve x8-x30.
> +	 *
> +	 * The most straightforward approach is to look at the function ID
> +	 * sent by the caller. However, the caller could send FFA_MSG_WAIT
> +	 * which is a 32-bit interface but the reply could very well be 64-bit
> +	 * such as FFA_FN64_MSG_SEND_DIRECT_REQ or FFA_MSG_SEND_DIRECT_REQ2.
> +	 *
> +	 * Instead, we could look at the function ID in the response (a0) but
> +	 * that doesn't work either as FFA_VERSION responses put the version
> +	 * number (or error code) in w0.
> +	 *
> +	 * Set x8-x17 iff response contains 64-bit function ID in a0.
> +	 */
> +	if (func_id != FFA_VERSION && ARM_SMCCC_IS_64(res->a0)) {
> +		cpu_reg(ctxt, 8) = res->a8;
> +		cpu_reg(ctxt, 9) = res->a9;
> +		cpu_reg(ctxt, 10) = res->a10;
> +		cpu_reg(ctxt, 11) = res->a11;
> +		cpu_reg(ctxt, 12) = res->a12;
> +		cpu_reg(ctxt, 13) = res->a13;
> +		cpu_reg(ctxt, 14) = res->a14;
> +		cpu_reg(ctxt, 15) = res->a15;
> +		cpu_reg(ctxt, 16) = res->a16;
> +		cpu_reg(ctxt, 17) = res->a17;
> +	}
>  }

I don't see how that can ever work.

Irrespective of how FFA_MSG_WAIT actually works (and I couldn't find
anything in the spec that supports the above), the requester will
fully expect its registers to be preserved based on the initial
function type, and that alone. No ifs, no buts.

If what you describe can happen (please provide a convincing example),
then the spec is doomed.

	M.

-- 
Without deviation from the norm, progress is not possible.

Powered by blists - more mailing lists