[PATCH 3/4] kvmppc: magic page paravirtualization - guest part

[ehrhardt at linux.vnet.ibm.com]

From: Christian Ehrhardt <ehrhardt at linux.vnet.ibm.com>

This patch adds the guest handling for the magic page mechanism. A Hypervisor
can modify the device tree passed to the guest. Using that already existing
interface a guest can simply detect available hypervisor features and agree
on the supported ones using hypercalls.
In this example it is checked for the feature switch "feature,pv-magicpage"
in the hypervisor node and additional data which represents the size the
hypervisor requests in "data,pv-magicpage-size".
When the guest reads that data and wants to support it the memory is allocated
and passed to the hypervisor using the KVM_HCALL_RESERVE_MAGICPAGE hypercall.

Signed-off-by: Christian Ehrhardt <ehrhardt at linux.vnet.ibm.com>
---

[diffstat]
 arch/powerpc/kernel/kvm.c      |   51 +++++++++++++++++++++++++++++++++++++++++
 include/asm-powerpc/fixmap.h   |   10 +++++++-
 include/asm-powerpc/kvm_para.h |   26 ++++++++++++++++++++
 mm/page_alloc.c                |    1
 4 files changed, 87 insertions(+), 1 deletion(-)

[diff]

diff --git a/arch/powerpc/kernel/kvm.c b/arch/powerpc/kernel/kvm.c
--- a/arch/powerpc/kernel/kvm.c
+++ b/arch/powerpc/kernel/kvm.c
@@ -22,9 +22,60 @@
 #include <linux/percpu.h>
 #include <linux/mm.h>
 #include <linux/kvm_para.h>
+#include <linux/bootmem.h>
+#include <asm/fixmap.h>
+
+/*
+ * this is guest memory granted to the hypervisor;
+ * the hypervisor can place data in this area and rewrite
+ * privileged instructions to read from this area without
+ * trapping.
+ * Only the Hypervisor needs to be aware of the structure layout
+ * which makes the guest more felxible - the guest only guarantees
+ * the size which is requested by the hypervisor and read from a
+ * device tree entry.
+ */
+static void *kvm_magicpage;
+
+static void __init kvmppc_register_magic_page(void)
+{
+	unsigned long gvaddr;
+	unsigned long gpaddr;
+	int size;
+	long err;
+
+	size = kvmppc_pv_read_data(KVM_PVDATA_MAGICPAGE_SIZE);
+	if (size < 0) {
+		printk(KERN_ERR "%s: couldn't read size for kvmppc style "
+			"paravirtualization support (got %d)\n",
+			__func__, size);
+		return;
+	}
+
+	/* FIXME Guest SMP needs that percpu which */
+	kvm_magicpage = alloc_bootmem(size);
+	if (!kvm_magicpage) {
+		printk(KERN_ERR "%s - failed to allocate %d bytes\n",
+			 __func__, size);
+		return;
+	}
+	gpaddr = (unsigned long)__pa(kvm_magicpage);
+	gvaddr = fix_to_virt(FIX_KVM_PV);
+
+	err = kvm_hypercall2(KVM_HCALL_RESERVE_MAGICPAGE, gvaddr, gpaddr);
+	if (err)
+		printk(KERN_ERR "%s: couldn't register pv mem\n", __func__);
+	else
+		printk(KERN_NOTICE "%s: registered %d bytes for pv mem support"
+			" (gvaddr 0x%08lx gpaddr 0x%08lx)\n",
+			 __func__, size, gvaddr, gpaddr);
+}
 
 void __init kvm_guest_init(void)
 {
 	if (!kvm_para_available())
 		return;
+
+	if (kvm_para_has_feature(KVM_FEATURE_PPCPV_MAGICPAGE))
+		kvmppc_register_magic_page();
 }
diff --git a/include/asm-powerpc/fixmap.h b/include/asm-powerpc/fixmap.h
--- a/include/asm-powerpc/fixmap.h
+++ b/include/asm-powerpc/fixmap.h
@@ -36,7 +36,7 @@
  *
  * these 'compile-time allocated' memory buffers are
  * fixed-size 4k pages. (or larger if used with an increment
- * highger than 1) use fixmap_set(idx,phys) to associate
+ * higher than 1) use fixmap_set(idx,phys) to associate
  * physical memory with fixmap indices.
  *
  * TLB entries of such buffers will not be flushed across
@@ -44,6 +44,14 @@
  */
 enum fixed_addresses {
 	FIX_HOLE,
+#ifdef CONFIG_KVM_GUEST
+	/*
+	 * reserved virtual address space for paravirtualization - needs to be
+	 *  <=32k away from base address 0 to be able to reach it with
+	 * immediate addressing using base 0 instead of needing a register.
+	 */
+	FIX_KVM_PV,
+#endif
 #ifdef CONFIG_HIGHMEM
 	FIX_KMAP_BEGIN,	/* reserved pte's for temporary kernel mappings */
 	FIX_KMAP_END = FIX_KMAP_BEGIN+(KM_TYPE_NR*NR_CPUS)-1,
diff --git a/include/asm-powerpc/kvm_para.h b/include/asm-powerpc/kvm_para.h
--- a/include/asm-powerpc/kvm_para.h
+++ b/include/asm-powerpc/kvm_para.h
@@ -28,10 +28,18 @@
 
 #define KVM_HYPERCALL_BIN 0x03ffffff
 
+#define KVM_HCALL_RESERVE_MAGICPAGE	0
+
+#define KVM_PVDATA_MAGICPAGE_SIZE	"data,pv-magicpage-size"
+
+/* List of PV features supported, returned as a bitfield */
+#define KVM_FEATURE_PPCPV_MAGICPAGE	0
+
 static struct kvmppc_para_features {
 	char *dtcell;
 	int feature;
 } para_features[] = {
+	{ "feature,pv-magicpage", KVM_FEATURE_PPCPV_MAGICPAGE }
 };
 
 static inline int kvm_para_available(void)
@@ -67,6 +75,24 @@
 	of_node_put(dn);
 
 	return features;
+}
+
+/* reads the specified data field out of the hypervisor node */
+static inline int kvmppc_pv_read_data(char *dtcell)
+{
+	struct device_node *dn;
+	const int *dtval;
+
+	dn = of_find_node_by_path("/hypervisor");
+	if (!dn)
+		return -EINVAL;
+
+	dtval = of_get_property(dn, dtcell, NULL);
+	of_node_put(dn);
+	if (dtval)
+		return *dtval;
+	else
+		return -EINVAL;
 }
 
 void kvm_guest_init(void);
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -550,6 +550,7 @@
 				prefetchw(p + 1);
 			__ClearPageReserved(p);
 			set_page_count(p, 0);
+
 		}
 
 		set_page_refcounted(page);