On 5/28/24 02:55, Kirill A. Shutemov wrote:

+/* Stop new private<->shared conversions */
+static void tdx_kexec_begin(bool crash)
+{
+	/*
+	 * Crash kernel reaches here with interrupts disabled: can't wait for
+	 * conversions to finish.
+	 *
+	 * If race happened, just report and proceed.
+	 */
+	if (!set_memory_enc_stop_conversion(!crash))
+		pr_warn("Failed to stop shared<->private conversions\n");
+}

I don't like having to pass 'crash' in here.

If interrupts are the problem we have ways of testing for those directly.

If it's being in an oops that's a problem, we have 'oops_in_progress'
for that.

In other words, I'd much rather this function (or better yet
set_memory_enc_stop_conversion() itself) use some existing API to change
its behavior in a crash rather than have the context be passed down and
twiddled through several levels of function calls.

There are a ton of these in the console code:

	if (oops_in_progress)
		foo_trylock();
	else
		foo_lock();

To me, that's a billion times more clear than a 'wait' argument that
gets derives from who-knows-what that I have to trace through ten levels
of function calls.