[RFC PATCH 2/2] ARM: SMP support for mach-virt

On Mon, Dec 03, 2012 at 09:55:35PM +0000, Rob Herring wrote:

On 12/03/2012 11:52 AM, Will Deacon wrote:

quoted

From: Marc Zyngier <redacted>

This patch adds support for SMP to mach-virt.

Signed-off-by: Marc Zyngier <redacted>
Signed-off-by: Will Deacon <redacted>

[...]

quoted

+/*
+ * This provides a "holding pen" into which all secondary cores are held
+ * until we're ready for them to initialise.
+ */
+ENTRY(virt_secondary_startup)
+	mrc	p15, 0, r0, c0, c0, 5
+	and	r0, r0, #15
+	adr	r4, 1f
+	ldmia	r4, {r5, r6}
+	sub	r4, r4, r5
+	add	r6, r6, r4
+pen:	ldr	r7, [r6]
+	cmp	r7, r0
+	bne	pen

Why is the pen is needed? It should only be needed for hotplug on
systems that can't reset their cores. I'd hope you could design good
virtual h/w.

It's not so much about designing good virtual h/w as it is avoiding tying
the platform to it. What we don't want is to mandate that in order to boot
this machine, you *must* implement an emulation of some virtual
power-controller or SMP booting device. If we go down that route, there's
less advantage from having the virtual platform in the first place.

There's also less of a problem with the pen approach to booting because
ultimately the virtual CPUs executing there are just pthreads and will be
scheduled appropriately by the hypervisor (in contrast to a real system
where there may be concerns about power consumption and memory bandwidth).

For hotplug, sure, we could have an *optional* virtio-based device for
dealing with that if we want to. We could even have some early probing code
for it and use it for SMP boot if we find a matching DT node, but we'd still
need to keep the pen code lying around as a fallback.

quoted

+/*
+ * Enumerate the possible CPU set from the device tree.
+ */
+static void __init virt_smp_init_cpus(void)
+{
+	const char *enable_method;
+	struct device_node *dn = NULL;
+	int cpu = 0;
+	u32 release_addr;
+
+	while ((dn = of_find_node_by_type(dn, "cpu"))) {
+		if (cpu >= NR_CPUS)
+			goto next;
+
+		/*
+		 * We currently support only the "spin-table" enable-method.
+		 */
+		enable_method = of_get_property(dn, "enable-method", NULL);
+		if (!enable_method || strcmp(enable_method, "spin-table")) {

Are these documented?

It's part of the EPAPR spec iirc and follows the booting protocol used by
arm64.

Will