Re: [RFC PATCH] x86/mce: Make mce_rdmsrl() do a plain RDMSR only

["Luck, Tony" <tony.luck@intel.com>]

On Wed, Sep 09, 2020 at 01:30:22PM +0200, Borislav Petkov wrote:
> I guess something as straightforward as this:

Do we think there will be other places where we want this
MSR-or-die behaviour?  If there are, then most of this
belongs elsewhere from arch/x86/kernel/cpu/mce/core.c

> ---
> diff --git a/arch/x86/kernel/cpu/mce/core.c b/arch/x86/kernel/cpu/mce/core.c
> index 0ba24dfffdb2..9893caaf2696 100644
> --- a/arch/x86/kernel/cpu/mce/core.c
> +++ b/arch/x86/kernel/cpu/mce/core.c
> @@ -373,10 +373,27 @@ static int msr_to_offset(u32 msr)
>  	return -1;
>  }
>  
> +__visible bool ex_handler_rdmsr_fault(const struct exception_table_entry *fixup,
> +				      struct pt_regs *regs, int trapnr,
> +				      unsigned long error_code,
> +				      unsigned long fault_addr)
> +{
> +	if (pr_warn_once("MSR access error: RDMSR from 0x%x at rIP: 0x%lx (%pS)\n",

The "_once" version seems a little pointless when the next statement in the function
is "panic()".

"warn" seems understated for an error that is going to crash the system.
Just go for "pr_emerg()".

There seems no consistency on using "rIP" or "RIP" ... but I think "RIP"
is slightly ahead.

> +			 (unsigned int)regs->cx, regs->ip, (void *)regs->ip))
> +		show_stack_regs(regs);
> +
> +	panic("MCA Architectural violation!\n");

nitpick: I don't thing Architectural needs to be capitalized.

> +
> +	while (true)
> +		cpu_relax();

Ugh. Is this why you have warn_once() ... because panic might return?

Above comments also apply to the wrmsr path.

-Tony