Re: [Xen-devel] Re: [PATCH 1/5] xen: events: use irq_alloc_desc(_at)instead of open-coding an IRQ allocator.

[Stefano Stabellini]

On Tue, 26 Oct 2010, Ian Campbell wrote:
> On Mon, 2010-10-25 at 19:02 +0100, Ian Campbell wrote:
> > 
> > 
> > > What do you see when you pass in a PCI device and say give the guest
> > 32 CPUs??
> > 
> > I can try tomorrow and see, based on what you say above without
> > implementing what I described I suspect the answer will be "carnage". 
> 
> Actually, it looks like multi-vcpu is broken, I only see 1 regardless of
> how many I configured. It's not clear if this is breakage in Linus'
> tree, something I pulled in from one of Jeremy's, yours or Stefano's
> trees or some local pebcak. I'll investigate...
 
I found the bug, it was introduced by:

"xen: use vcpu_ops to setup cpu masks"

I have added the fix at the end of my branch and I am also appending the
fix here.

---


xen: initialize cpu masks for pv guests in xen_smp_init

Pv guests don't have ACPI and need the cpu masks to be set
correctly as early as possible so we call xen_fill_possible_map from
xen_smp_init.
On the other hand the initial domain supports ACPI so in this case we skip
xen_fill_possible_map and rely on it. However Xen might limit the number
of cpus usable by the domain, so we filter those masks during smp
initialization using the VCPUOP_is_up hypercall.
It is important that the filtering is done before
xen_setup_vcpu_info_placement.

Signed-off-by: Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx>

diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c
index 1386767..834dfeb 100644
--- a/arch/x86/xen/smp.c
+++ b/arch/x86/xen/smp.c
@@ -28,6 +28,7 @@
 #include <asm/xen/interface.h>
 #include <asm/xen/hypercall.h>
 
+#include <xen/xen.h>
 #include <xen/page.h>
 #include <xen/events.h>
 
@@ -156,6 +157,25 @@ static void __init xen_fill_possible_map(void)
 {
 	int i, rc;
 
+	if (xen_initial_domain())
+		return;
+
+	for (i = 0; i < nr_cpu_ids; i++) {
+		rc = HYPERVISOR_vcpu_op(VCPUOP_is_up, i, NULL);
+		if (rc >= 0) {
+			num_processors++;
+			set_cpu_possible(i, true);
+		}
+	}
+}
+
+static void __init xen_filter_cpu_maps(void)
+{
+	int i, rc;
+
+	if (!xen_initial_domain())
+		return;
+
 	num_processors = 0;
 	disabled_cpus = 0;
 	for (i = 0; i < nr_cpu_ids; i++) {
@@ -179,6 +199,7 @@ static void __init xen_smp_prepare_boot_cpu(void)
 	   old memory can be recycled */
 	make_lowmem_page_readwrite(xen_initial_gdt);
 
+	xen_filter_cpu_maps();
 	xen_setup_vcpu_info_placement();
 }
 
@@ -195,8 +216,6 @@ static void __init xen_smp_prepare_cpus(unsigned int max_cpus)
 	if (xen_smp_intr_init(0))
 		BUG();
 
-	xen_fill_possible_map();
-
 	if (!alloc_cpumask_var(&xen_cpu_initialized_map, GFP_KERNEL))
 		panic("could not allocate xen_cpu_initialized_map\n");
 
@@ -487,5 +506,6 @@ static const struct smp_ops xen_smp_ops __initdata = {
 void __init xen_smp_init(void)
 {
 	smp_ops = xen_smp_ops;
+	xen_fill_possible_map();
 	xen_init_spinlocks();
 }
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/