Re: [RFC PATCH v2 1/3] powerpc/prom_init: Replace linux,sml-base/sml-size with linux,sml-log

From: Jerry Snitselaar <jsnitsel@redhat.com>
To: Stefan Berger <stefanb@linux.ibm.com>
Cc: devicetree@vger.kernel.org, rnsastry@linux.ibm.com,
	linux-kernel@vger.kernel.org, jarkko@kernel.org,
	linux-integrity@vger.kernel.org, viparash@in.ibm.com,
	linuxppc-dev@lists.ozlabs.org, peterhuewe@gmx.de
Subject: Re: [RFC PATCH v2 1/3] powerpc/prom_init: Replace linux,sml-base/sml-size with linux,sml-log
Date: Mon, 11 Mar 2024 10:47:19 -0700	[thread overview]
Message-ID: <s2a42qzx4pp7fvgde4p5t7pcwsop4wlm266e2gdgr2gb5y77cn@udrmavhiwqoo> (raw)
In-Reply-To: <20240311132030.1103122-2-stefanb@linux.ibm.com>

On Mon, Mar 11, 2024 at 09:20:28AM -0400, Stefan Berger wrote:
> linux,sml-base holds the address of a buffer with the TPM log. This
> buffer may become invalid after a kexec. To avoid accessing an invalid
> address or corrupted buffer, embed the whole TPM log in the device tree
> property linux,sml-log. This helps to protect the log since it is
> properly carried across a kexec soft reboot with both of the kexec
> syscalls.
> 
> Avoid having the firmware ingest the whole TPM log when calling
> prom_setprop but only create the linux,sml-log property as a place holder.
> Insert the actual TPM log during the tree flattening phase.
> 
> Fixes: 4a727429abec ("PPC64: Add support for instantiating SML from Open Firmware")
> Suggested-by: Michael Ellerman <mpe@ellerman.id.au>
> Signed-off-by: Stefan Berger <stefanb@linux.ibm.com>
> ---
>  arch/powerpc/kernel/prom_init.c | 27 +++++++++++++++++++--------
>  1 file changed, 19 insertions(+), 8 deletions(-)
> 
> diff --git a/arch/powerpc/kernel/prom_init.c b/arch/powerpc/kernel/prom_init.c
> index e67effdba85c..6f7ca72013c2 100644
> --- a/arch/powerpc/kernel/prom_init.c
> +++ b/arch/powerpc/kernel/prom_init.c
> @@ -211,6 +211,8 @@ static cell_t __prombss regbuf[1024];
>  
>  static bool  __prombss rtas_has_query_cpu_stopped;
>  
> +static u64 __prombss sml_base;
> +static u32 __prombss sml_size;

Should inside an #ifdef CONFIG_PPC64 since prom_instantiate_sml() is?

>  
>  /*
>   * Error results ... some OF calls will return "-1" on error, some
> @@ -1954,17 +1956,15 @@ static void __init prom_instantiate_sml(void)
>  	}
>  	prom_printf(" done\n");
>  
> -	reserve_mem(base, size);
> -
> -	prom_setprop(ibmvtpm_node, "/vdevice/vtpm", "linux,sml-base",
> -		     &base, sizeof(base));
> -	prom_setprop(ibmvtpm_node, "/vdevice/vtpm", "linux,sml-size",
> -		     &size, sizeof(size));
> -
> -	prom_debug("sml base     = 0x%llx\n", base);
> +	/* Add property now, defer adding log to tree flattening phase */
> +	prom_setprop(ibmvtpm_node, "/vdevice/vtpm", "linux,sml-log",
> +		     NULL, 0);
>  	prom_debug("sml size     = 0x%x\n", size);
>  
>  	prom_debug("prom_instantiate_sml: end...\n");
> +
> +	sml_base = base;
> +	sml_size = size;
>  }
>  
>  /*
> @@ -2645,6 +2645,17 @@ static void __init scan_dt_build_struct(phandle node, unsigned long *mem_start,
>  		}
>  		prev_name = sstart + soff;
>  
> +		if (!prom_strcmp("linux,sml-log", pname)) {
> +			/* push property head */
> +			dt_push_token(OF_DT_PROP, mem_start, mem_end);
> +			dt_push_token(sml_size, mem_start, mem_end);
> +			dt_push_token(soff, mem_start, mem_end);
> +			/* push property content */
> +			valp = make_room(mem_start, mem_end, sml_size, 1);
> +			memcpy(valp, (void *)sml_base, sml_size);
> +			*mem_start = ALIGN(*mem_start, 4);
> +			continue;
> +		}

Same question as above.

Regards,
Jerry

>  		/* get length */
>  		l = call_prom("getproplen", 2, 1, node, pname);
>  
> -- 
> 2.43.0
> 