Re: [PATCHv3 27/32] x86/mm/cpa: Generailize __set_memory_enc_pgtable()

[Brijesh Singh]

On 2/18/22 10:17, Kirill A. Shutemov wrote:
...
>   
> -	/*
> -	 * Before changing the encryption attribute, we need to flush caches.
> -	 */
> -	cpa_flush(&cpa, !this_cpu_has(X86_FEATURE_SME_COHERENT));
> +	/* Flush caches as needed before changing the encryption attribute. */
> +	if (x86_platform.cc->enc_tlb_flush_required(enc))
> +		cpa_flush(&cpa, x86_platform.cc->enc_cache_flush_required());
>   
>   	ret = __change_page_attr_set_clr(&cpa, 1);
>   
> @@ -2027,7 +2026,8 @@ static int __set_memory_enc_pgtable(unsigned long addr, int numpages, bool enc)
>   	 * Notify hypervisor that a given memory range is mapped encrypted
>   	 * or decrypted.
>   	 */
> -	notify_range_enc_status_changed(addr, numpages, enc);
> +	if (!ret)
> +		ret = x86_platform.cc->enc_status_changed(addr, numpages, enc);
>   

Boris has given similar comment on SNP series. This area of code have a 
very similar requirement. Boris proposed it here [1] and I followed up 
[2] with SNP specific comment. It will be good to get your feedback and 
do a generic implementation outside of the SNP/TDX series.

[1] https://lore.kernel.org/linux-mm/Ygz88uacbwuTTNat@xxxxxxx/
[2] 
https://lore.kernel.org/linux-mm/20220216160457.1748381-1-brijesh.singh@xxxxxxx/T/

thanks