Re: [PATCH v2 16/21] x86/virt/seamldr: Handle TDX Module update failures

[Chao Gao <chao.gao@intel.com>]

> /*
>  * See multi_cpu_stop() from where this multi-cpu state-machine was
>  * adopted, and the rationale for touch_nmi_watchdog()
>@@ -293,8 +301,13 @@ static int do_seamldr_install_module(void *params)
> 				break;
> 			}
> 
>-			if (ret)
>+			if (ret) {
> 				atomic_inc(&tdp_data.failed);
>+				if (curstate > TDP_SHUTDOWN) {
>+					tdx_module_set_error();
>+					print_update_failure_message();
>+				}
>+			}

I found a bug here.

If an error is non-fatal (e.g., occurs before shutting down the TDX module),
the TDX Module remains functional and TDs can continue to run. So, there's no
need to set the TDX module status to erorr or print the error message.

But, in this implementation, a failing CPU doesn't exit the do-while() loop in
do_seamldr_install_module(). Instead, all CPUs fast-forward to the TDP_DONE
state and execute the do-while() body once more. Then, the failing CPU reaches
the above if() statement again with a non-zero "ret" and "curstate=TDP_DONE",
causing it to incorrectly set the module to error and print the error message.

To fix this issue, apply the following diff:

diff --git a/arch/x86/virt/vmx/tdx/seamldr.c b/arch/x86/virt/vmx/tdx/seamldr.c
index a72f6b0b27e9..e525bbd16610 100644
--- a/arch/x86/virt/vmx/tdx/seamldr.c
+++ b/arch/x86/virt/vmx/tdx/seamldr.c
@@ -258,16 +258,8 @@ static void set_target_state(enum tdp_state state)
 /* Last one to ack a state moves to the next state. */
 static void ack_state(void)
 {
-	if (atomic_dec_and_test(&tdp_data.thread_ack)) {
-		/*
-		 * If an error occurred, abort the update by skipping to
-		 * the final state
-		 */
-		if (atomic_read(&tdp_data.failed))
-			set_target_state(TDP_DONE);
-		else
-			set_target_state(tdp_data.state + 1);
-	}
+	if (atomic_dec_and_test(&tdp_data.thread_ack))
+		set_target_state(tdp_data.state + 1);
 }
 
 static void print_update_failure_message(void)
@@ -331,7 +323,7 @@ static int do_seamldr_install_module(void *params)
			touch_nmi_watchdog();
			rcu_momentary_eqs();
		}
-	} while (curstate != TDP_DONE);
+	} while (curstate != TDP_DONE && !atomic_read(&tdp_data.failed));
 
	return ret;
 }