Re: [PATCH] x86/mce: Add support for PHYSADDRV and PHYSADDRVALIDSUPPORTED bits

[Yazen Ghannam <yazen.ghannam@amd.com>]

On Tue, Jul 29, 2025 at 08:46:57PM +0000, Avadhut Naik wrote:
> Starting with Zen6, AMD's Scalable MCA systems will incorporate two new
> bits in MCA_STATUS and MCA_CONFIG MSRs. These bits will indicate if a
> valid System Physical Address (SPA) is present in MCA_ADDR.
> 
> PhysAddrValidSupported bit (MCA_CONFIG[11]) serves as the architectural
> indicator and states if PhysAddrV bit (MCA_STATUS[54]) is Reserved or
> if it indicates validity of SPA in MCA_ADDR.
> 
> PhysAddrV bit (MCA_STATUS[54]) advertises if MCA_ADDR contains valid
> SPA or if it is implementation specific.

The commit message is missing an imperative statement. You can describe
how and why these bits are helpful.

For example: "Use and prefer MCA_STATUS[PhysAddrV] when checking for a
usable address." or something like this.

> 
> Signed-off-by: Avadhut Naik <avadhut.naik@amd.com>
> ---
>  arch/x86/include/asm/mce.h    |  2 ++
>  arch/x86/kernel/cpu/mce/amd.c | 16 +++++++++++++---
>  2 files changed, 15 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/x86/include/asm/mce.h b/arch/x86/include/asm/mce.h
> index 6c77c03139f7..387cf250525f 100644
> --- a/arch/x86/include/asm/mce.h
> +++ b/arch/x86/include/asm/mce.h
> @@ -48,6 +48,7 @@
>  
>  /* AMD-specific bits */
>  #define MCI_STATUS_TCC		BIT_ULL(55)  /* Task context corrupt */
> +#define MCI_STATUS_PADDRVAL	BIT_ULL(54)  /* Valid System Physical Address */
>  #define MCI_STATUS_SYNDV	BIT_ULL(53)  /* synd reg. valid */
>  #define MCI_STATUS_DEFERRED	BIT_ULL(44)  /* uncorrected error, deferred exception */
>  #define MCI_STATUS_POISON	BIT_ULL(43)  /* access poisonous data */
> @@ -62,6 +63,7 @@
>   */
>  #define MCI_CONFIG_MCAX		0x1
>  #define MCI_CONFIG_FRUTEXT	BIT_ULL(9)
> +#define MCI_CONFIG_PAVALID	BIT_ULL(11)
>  #define MCI_IPID_MCATYPE	0xFFFF0000
>  #define MCI_IPID_HWID		0xFFF
>  
> diff --git a/arch/x86/kernel/cpu/mce/amd.c b/arch/x86/kernel/cpu/mce/amd.c
> index 5c4eb28c3ac9..6ac222aec28d 100644
> --- a/arch/x86/kernel/cpu/mce/amd.c
> +++ b/arch/x86/kernel/cpu/mce/amd.c
> @@ -748,9 +748,9 @@ bool amd_mce_is_memory_error(struct mce *m)
>  }
>  
>  /*
> - * AMD systems do not have an explicit indicator that the value in MCA_ADDR is
> - * a system physical address. Therefore, individual cases need to be detected.
> - * Future cases and checks will be added as needed.
> + * Some AMD systems have an explicit indicator that the value in MCA_ADDR is a
> + * system physical address. Individual cases though, need to be detected for
> + * other systems. Future cases will be added as needed.
>   *
>   * 1) General case
>   *	a) Assume address is not usable.
> @@ -764,11 +764,21 @@ bool amd_mce_is_memory_error(struct mce *m)
>   *	a) Reported in legacy bank 4 with extended error code (XEC) 8.
>   *	b) MCA_STATUS[43] is *not* defined as poison in legacy bank 4. Therefore,
>   *	   this bit should not be checked.
> + * 4) MCI_STATUS_PADDRVAL is set
> + *	a)Will provide a valid system physical address.

Space after "a)".

>   *
>   * NOTE: SMCA UMC memory errors fall into case #1.
>   */
>  bool amd_mce_usable_address(struct mce *m)
>  {
> +	u64 smca_config;
> +
> +	rdmsrl(MSR_AMD64_SMCA_MCx_CONFIG(m->bank), smca_config);

Newline here.

> +	if (smca_config & MCI_CONFIG_PAVALID) {
> +		if(m->status & MCI_STATUS_PADDRVAL)
> +			return true;
> +		return false;
> +	}

Newline here.

Also, the entire hunk above should go after the !SMCA case below and
before the other SMCA cases.

Furthermore, the hunk can be simplified to this:

	if (smca_config & MCI_CONFIG_PAVALID)
		return m->status & MCI_STATUS_PADDRVAL;

Also also, the bit names are uncannily similar. I think they should be
the same (except for the prefix) or clearly different.

MCI_CONFIG_PADDRV/MCI_STATUS_PADDRV

MCI_CONFIG_PHYS_ADDRV_SUPP/MCI_STATUS_PADDRVAL

>  	/* Check special northbridge case 3) first. */
>  	if (!mce_flags.smca) {
>  		if (legacy_mce_is_memory_error(m))
> 
> base-commit: d69139008b6dcd9c18483e956f61d187b0c214a2
> -- 

We should revisit saving MCA_CONFIG during logging. We end up reading it
multiple times during error processing.

I realize this came up before in a previous patch set. IIRC, the point
of contention was exposing MCA_CONFIG to userspace. We don't necessarily
have to do that. We can just use it in the kernel data structures.

Thanks,
Yazen