Re: [PATCH v4 02/10] cpu/bugs: Allow forcing Automatic IBRS with SNP active using spectre_v2=eibrs

[Pawan Gupta]

On Tue, Aug 04, 2026 at 06:56:03PM -0500, Kim Phillips wrote:
...
> @@ -2192,7 +2193,12 @@ static void __init spectre_v2_select_mitigation(void)
>  			break;
>  		fallthrough;
>  	case SPECTRE_V2_CMD_FORCE:
> -		if (boot_cpu_has(X86_FEATURE_IBRS_ENHANCED)) {
> +		/*
> +		 * Don't use AutoIBRS when SNP is enabled because it degrades
> +		 * host userspace indirect branch performance.
> +		 */
> +		if (boot_cpu_has(X86_FEATURE_IBRS_ENHANCED) &&
> +		    !boot_cpu_has(X86_FEATURE_SEV_SNP)) {
>  			spectre_v2_enabled = SPECTRE_V2_EIBRS;
>  			break;
>  		}

Can we also account for retpoline availability here? If both SNP and
retpoline are available, then only avoid AutoIBRS:

          case SPECTRE_V2_CMD_FORCE:
                  /*
                   * Prefer retpoline when SNP is enabled because Auto-IBRS
                   * degrades host userspace indirect branch performance.
                   */
                  if (boot_cpu_has(X86_FEATURE_IBRS_ENHANCED) &&
                     !(boot_cpu_has(X86_FEATURE_SEV_SNP) &&
                       IS_ENABLED(CONFIG_MITIGATION_RETPOLINE))) {
                          spectre_v2_enabled = SPECTRE_V2_EIBRS;
                          break;
                  }

With this, patch 1 & 3 can be omitted. Also avoids the ping pong: we don't
want AutoIBRS -> We don't have retpoline -> Lets have AutoIBRS.