Re: [Patch V1 3/3] x86, mce: Handling LMCE events

[Borislav Petkov <bp@alien8.de>]

On Fri, May 29, 2015 at 09:28:02AM -0700, Ashok Raj wrote:
> This patch has handling changes to do_machine_check() to process MCE
> signaled as local MCE. Typically only recoverable errors (SRAR) type
> error will be Signaled as LMCE. But architecture does not restrict to
> only those errors.
> 
> When errors are signaled as LMCE, there is no need for the MCE handler to
> perform rendezvous with other logical processors unlike earlier processors
> that would broadcast machine check errors.
> 
> See http://www.intel.com/sdm Volume 3, Chapter 15 for more information
> on MSR's and documentation on Local MCE.
> 
> Signed-off-by: Ashok Raj <ashok.raj@intel.com>
> ---
>  arch/x86/kernel/cpu/mcheck/mce.c       | 25 ++++++++++++++++++++++---
>  arch/x86/kernel/cpu/mcheck/mce_intel.c |  1 +
>  2 files changed, 23 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c
> index d10aada..c130391 100644
> --- a/arch/x86/kernel/cpu/mcheck/mce.c
> +++ b/arch/x86/kernel/cpu/mcheck/mce.c
> @@ -1047,6 +1047,7 @@ void do_machine_check(struct pt_regs *regs, long error_code)
>  	char *msg = "Unknown";
>  	u64 recover_paddr = ~0ull;
>  	int flags = MF_ACTION_REQUIRED;
> +	int lmce = 0;
>  
>  	prev_state = ist_enter(regs);
>  
> @@ -1074,11 +1075,19 @@ void do_machine_check(struct pt_regs *regs, long error_code)
>  		kill_it = 1;
>  
>  	/*
> +	 * Check if this MCE is signaled to only this logical processor
> +	 */
> +	if (m.mcgstatus & MCG_STATUS_LMCES)
> +		lmce = 1;

	else
		/*
		 * Go through ...
		 * ...
		 */
		order = mce_start(&no_way_out);


> +	/*
>  	 * Go through all the banks in exclusion of the other CPUs.
>  	 * This way we don't report duplicated events on shared banks
>  	 * because the first one to see it will clear it.
> +	 * If this is a Local MCE, then no need to perform rendezvous.
>  	 */
> -	order = mce_start(&no_way_out);
> +	if (!lmce)
> +		order = mce_start(&no_way_out);
> +
>  	for (i = 0; i < cfg->banks; i++) {
>  		__clear_bit(i, toclear);
>  		if (!test_bit(i, valid_banks))
-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--