On Wed, 1 Aug 2012, Konrad Rzeszutek Wilk wrote:
On Thu, Jul 26, 2012 at 04:33:49PM +0100, Stefano Stabellini wrote:
Check for a "/xen" node in the device tree, if it is present set xen_domain_type to XEN_HVM_DOMAIN and continue initialization.
Map the real shared info page using XENMEM_add_to_physmap with XENMAPSPACE_shared_info.
Signed-off-by: Stefano Stabellini stefano.stabellini@eu.citrix.com
arch/arm/xen/enlighten.c | 56 ++++++++++++++++++++++++++++++++++++++++++++++ 1 files changed, 56 insertions(+), 0 deletions(-)
diff --git a/arch/arm/xen/enlighten.c b/arch/arm/xen/enlighten.c index d27c2a6..8c923af 100644 --- a/arch/arm/xen/enlighten.c +++ b/arch/arm/xen/enlighten.c @@ -5,6 +5,9 @@ #include <asm/xen/hypervisor.h> #include <asm/xen/hypercall.h> #include <linux/module.h> +#include <linux/of.h> +#include <linux/of_irq.h> +#include <linux/of_address.h> struct start_info _xen_start_info; struct start_info *xen_start_info = &_xen_start_info; @@ -33,3 +36,56 @@ int xen_remap_domain_mfn_range(struct vm_area_struct *vma, return -ENOSYS; } EXPORT_SYMBOL_GPL(xen_remap_domain_mfn_range);
+/*
- == Xen Device Tree format ==
- /xen node;
- compatible "arm,xen";
- one interrupt for Xen event notifications;
- one memory region to map the grant_table.
- */
+static int __init xen_guest_init(void) +{
- int cpu;
- struct xen_add_to_physmap xatp;
- static struct shared_info *shared_info_page = 0;
- struct device_node *node;
- node = of_find_compatible_node(NULL, NULL, "arm,xen");
- if (!node) {
pr_info("No Xen support\n");I don't think the pr_info is appropiate here?
Yes, you are right. In fact I had already turned it into a pr_debug.
return 0;Should this be -ENODEV?
Considering that xen_guest_init is called by a core_initcall, I didn't want to return an error just because Xen is not present on the platform.
- }
- xen_domain_type = XEN_HVM_DOMAIN;
- if (!shared_info_page)
shared_info_page = (struct shared_info *)get_zeroed_page(GFP_KERNEL);- if (!shared_info_page) {
pr_err("not enough memory");\n
OK
return -ENOMEM;- }
- xatp.domid = DOMID_SELF;
- xatp.idx = 0;
- xatp.space = XENMAPSPACE_shared_info;
- xatp.gpfn = __pa(shared_info_page) >> PAGE_SHIFT;
- if (HYPERVISOR_memory_op(XENMEM_add_to_physmap, &xatp))
BUG();- HYPERVISOR_shared_info = (struct shared_info *)shared_info_page;
- /* xen_vcpu is a pointer to the vcpu_info struct in the shared_info
* page, we use it in the event channel upcall and in some pvclock* related functions. We don't need the vcpu_info placement* optimizations because we don't use any pv_mmu or pv_irq op on* HVM.* When xen_hvm_init_shared_info is run at boot time only vcpu 0 is* online but xen_hvm_init_shared_info is run at resume time too and* in that case multiple vcpus might be online. */- for_each_online_cpu(cpu) {
per_cpu(xen_vcpu, cpu) =&HYPERVISOR_shared_info->vcpu_info[cpu];- }
- return 0;
This above looks stringly similar to the x86 one. Could it be abstracted away to share the same code? Or is that something that ought to be done later on when there is more meat on the bone?
Actually I had to remove these three lines because on ARM we are going to have just one vcpu_info struct in the shared_info page and then rely on VCPUOP_register_vcpu_info.