Re: [PATCH] perf/x86/intel/lbr: Zero the xstate buffer on allocation

[Liang, Kan]

On 6/11/2021 9:03 AM, Thomas Gleixner wrote:
> XRSTORS requires a valid xstate buffer to work correctly. XSAVES does not
> guarantee to write a fully valid buffer according to the SDM:
>
>    "XSAVES does not write to any parts of the XSAVE header other than the
>     XSTATE_BV and XCOMP_BV fields."
>
> XRSTORS triggers a #GP:
>
>    "If bytes 63:16 of the XSAVE header are not all zero."
>
> It's dubious at best how this can work at all when the buffer is not zeroed
> before use.

I didn't run into any issues when I did the test. I guess the reserved 
bits in the buffer may always happen to be 0. That's why it didn't set 
off the alarm to me. Thank you very much for the fix.

Thanks,
Kan

> Allocate the buffers with __GFP_ZERO to prevent XRSTORS failure.
>
> Fixes: ce711ea3cab9 ("perf/x86/intel/lbr: Support XSAVES/XRSTORS for LBR context switch")
> Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> Cc: Kan Liang <kan.liang@xxxxxxxxxxxxxxx>
> Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> Cc: x86@xxxxxxxxxx
> ---
>   arch/x86/events/intel/lbr.c |    3 ++-
>   1 file changed, 2 insertions(+), 1 deletion(-)
>
> --- a/arch/x86/events/intel/lbr.c
> +++ b/arch/x86/events/intel/lbr.c
> @@ -731,7 +731,8 @@ void reserve_lbr_buffers(void)
>   		if (!kmem_cache || cpuc->lbr_xsave)
>   			continue;
>   
> -		cpuc->lbr_xsave = kmem_cache_alloc_node(kmem_cache, GFP_KERNEL,
> +		cpuc->lbr_xsave = kmem_cache_alloc_node(kmem_cache,
> +							GFP_KERNEL | __GFP_ZERO,
>   							cpu_to_node(cpu));
>   	}
>   }