[PATCH] x86/sev: Fix SNP guest kdump hang/softlockup/panic

[Tom Lendacky]

On 4/21/25 15:44, Ashish Kalra wrote:
> From: Ashish Kalra <ashish.kalra at amd.com>
> 

No need to add to what Boris already said on the commit message and
comments, so just looking at the code.

> +
> +		/*
> +		 * Issue AP destroy on all APs (to ensure they are kicked out
> +		 * of guest mode) to allow using RMPADJUST to remove the VMSA
> +		 * tag on VMSA pages especially for guests that allow HLT to
> +		 * not be intercepted.
> +		 */
> +
> +		local_irq_save(flags);
> +
> +		ghcb = __sev_get_ghcb(&state);
> +
> +		vc_ghcb_invalidate(ghcb);
> +		ghcb_set_rax(ghcb, vmsa->sev_features);
> +
> +		/* Issue VMGEXIT AP Destroy NAE event */
> +
> +		ghcb_set_sw_exit_code(ghcb, SVM_VMGEXIT_AP_CREATION);
> +		ghcb_set_sw_exit_info_1(ghcb,
> +					((u64)apic_id << 32)	|
> +					((u64)snp_vmpl << 16)	|
> +					SVM_VMGEXIT_AP_DESTROY);
> +		ghcb_set_sw_exit_info_2(ghcb, __pa(vmsa));
> +
> +		sev_es_wr_ghcb_msr(__pa(ghcb));
> +		VMGEXIT();
> +
> +		if (!ghcb_sw_exit_info_1_is_valid(ghcb) ||
> +		    lower_32_bits(ghcb->save.sw_exit_info_1)) {
> +			pr_err("SNP AP Destroy error\n");
> +			local_irq_restore(flags);
> +			return;
> +		}
> +
> +		__sev_put_ghcb(&state);
> +
> +		local_irq_restore(flags);

It looks like this whole block above is very similar to the block in
wakeup_cpu_via_vmgexit(). It makes sense to create a single function
that takes either SVM_VMGEXIT_AP_CREATE or SVM_VMGEXIT_AP_DESTROY as an
argument and processes appropriately.

> +
> +		snp_cleanup_vmsa(vmsa, apic_id);
> +	}
> +}
> +
>  void snp_kexec_finish(void)
>  {
>  	struct sev_es_runtime_data *data;
> @@ -987,6 +1088,8 @@ void snp_kexec_finish(void)
>  	if (!IS_ENABLED(CONFIG_KEXEC_CORE))
>  		return;
>  
> +	sev_snp_shutdown_all_aps();
> +
>  	unshare_all_memory();
>  
>  	/*
> @@ -1254,6 +1357,8 @@ static int wakeup_cpu_via_vmgexit(u32 apic_id, unsigned long start_ip)
>  	/* Record the current VMSA page */
>  	per_cpu(sev_vmsa, cpu) = vmsa;
>  
> +	per_cpu(sev_vcpu_apic_id, cpu) = apic_id;

Is this really needed? Can't you use the cpuid_to_apicid array or create
a function in the topology.c file to return the APIC ID?

Thanks,
Tom

> +
>  	return ret;
>  }
>