[RFC KVM 19/27] kvm/isolation: initialize the KVM page table with core mappings

[Alexandre Chartre <alexandre.chartre@oracle.com>]

The KVM page table is initialized with adding core memory mappings:
the kernel text, the per-cpu memory, the kvm module, the cpu_entry_area,
%esp fixup stacks, IRQ stacks.

Signed-off-by: Alexandre Chartre <alexandre.chartre@oracle.com>
---
 arch/x86/kernel/cpu/common.c |    2 +
 arch/x86/kvm/isolation.c     |  131 ++++++++++++++++++++++++++++++++++++++++++
 arch/x86/kvm/isolation.h     |   10 +++
 include/linux/percpu.h       |    2 +
 mm/percpu.c                  |    6 +-
 5 files changed, 149 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index 3764054..0fa44b1 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -1511,6 +1511,8 @@ static __init int setup_clearcpuid(char *arg)
 EXPORT_PER_CPU_SYMBOL(current_task);
 
 DEFINE_PER_CPU(struct irq_stack *, hardirq_stack_ptr);
+EXPORT_PER_CPU_SYMBOL_GPL(hardirq_stack_ptr);
+
 DEFINE_PER_CPU(unsigned int, irq_count) __visible = -1;
 
 DEFINE_PER_CPU(int, __preempt_count) = INIT_PREEMPT_COUNT;
diff --git a/arch/x86/kvm/isolation.c b/arch/x86/kvm/isolation.c
index 2052abf..cf5ee0d 100644
--- a/arch/x86/kvm/isolation.c
+++ b/arch/x86/kvm/isolation.c
@@ -10,6 +10,8 @@
 #include <linux/printk.h>
 #include <linux/slab.h>
 
+#include <asm/cpu_entry_area.h>
+#include <asm/processor.h>
 #include <asm/mmu_context.h>
 #include <asm/pgalloc.h>
 
@@ -88,6 +90,8 @@ struct mm_struct kvm_mm = {
 DEFINE_STATIC_KEY_FALSE(kvm_isolation_enabled);
 EXPORT_SYMBOL(kvm_isolation_enabled);
 
+static void kvm_isolation_uninit_page_table(void);
+static void kvm_isolation_uninit_mm(void);
 static void kvm_clear_mapping(void *ptr, size_t size,
 			      enum page_table_level level);
 
@@ -1024,10 +1028,130 @@ int kvm_copy_percpu_mapping(void *percpu_ptr, size_t size)
 EXPORT_SYMBOL(kvm_copy_percpu_mapping);
 
 
+static int kvm_isolation_init_page_table(void)
+{
+	void *stack;
+	int cpu, rv;
+
+	/*
+	 * Copy the mapping for all the kernel text. We copy at the PMD
+	 * level since the PUD is shared with the module mapping space.
+	 */
+	rv = kvm_copy_mapping((void *)__START_KERNEL_map, KERNEL_IMAGE_SIZE,
+	     PGT_LEVEL_PMD);
+	if (rv)
+		goto out_uninit_page_table;
+
+	/* copy the mapping of per cpu memory */
+	rv = kvm_copy_mapping(pcpu_base_addr, pcpu_unit_size * pcpu_nr_units,
+	     PGT_LEVEL_PMD);
+	if (rv)
+		goto out_uninit_page_table;
+
+	/*
+	 * Copy the mapping for cpu_entry_area and %esp fixup stacks
+	 * (this is based on the PTI userland address space, but probably
+	 * not needed because the KVM address space is not directly
+	 * enterered from userspace). They can both be copied at the P4D
+	 * level since they each have a dedicated P4D entry.
+	 */
+	rv = kvm_copy_mapping((void *)CPU_ENTRY_AREA_PER_CPU, P4D_SIZE,
+	     PGT_LEVEL_P4D);
+	if (rv)
+		goto out_uninit_page_table;
+
+#ifdef CONFIG_X86_ESPFIX64
+	rv = kvm_copy_mapping((void *)ESPFIX_BASE_ADDR, P4D_SIZE,
+	     PGT_LEVEL_P4D);
+	if (rv)
+		goto out_uninit_page_table;
+#endif
+
+#ifdef CONFIG_VMAP_STACK
+	/*
+	 * Interrupt stacks are vmap'ed with guard pages, so we need to
+	 * copy mappings.
+	 */
+	for_each_possible_cpu(cpu) {
+		stack = per_cpu(hardirq_stack_ptr, cpu);
+		pr_debug("IRQ Stack %px\n", stack);
+		if (!stack)
+			continue;
+		rv = kvm_copy_ptes(stack - IRQ_STACK_SIZE, IRQ_STACK_SIZE);
+		if (rv)
+			goto out_uninit_page_table;
+	}
+
+#endif
+
+	/* copy mapping of the current module (kvm) */
+	rv = kvm_copy_module_mapping();
+	if (rv)
+		goto out_uninit_page_table;
+
+	return 0;
+
+out_uninit_page_table:
+	kvm_isolation_uninit_page_table();
+	return rv;
+}
+
+/*
+ * Free all buffers used by the kvm page table. These buffers are stored
+ * in the kvm_pgt_dgroup_list.
+ */
+static void kvm_isolation_uninit_page_table(void)
+{
+	struct pgt_directory_group *dgroup, *dgroup_next;
+	enum page_table_level level;
+	void *ptr;
+	int i;
+
+	mutex_lock(&kvm_pgt_dgroup_lock);
+
+	list_for_each_entry_safe(dgroup, dgroup_next,
+				 &kvm_pgt_dgroup_list, list) {
+
+		for (i = 0; i < dgroup->count; i++) {
+			ptr = dgroup->directory[i].ptr;
+			level = dgroup->directory[i].level;
+
+			switch (dgroup->directory[i].level) {
+
+			case PGT_LEVEL_PTE:
+				kvm_pte_free(NULL, ptr);
+				break;
+
+			case PGT_LEVEL_PMD:
+				kvm_pmd_free(NULL, ptr);
+				break;
+
+			case PGT_LEVEL_PUD:
+				kvm_pud_free(NULL, ptr);
+				break;
+
+			case PGT_LEVEL_P4D:
+				kvm_p4d_free(NULL, ptr);
+				break;
+
+			default:
+				pr_err("unexpected page directory %d for %px\n",
+				       level, ptr);
+			}
+		}
+
+		list_del(&dgroup->list);
+		kfree(dgroup);
+	}
+
+	mutex_unlock(&kvm_pgt_dgroup_lock);
+}
+
 static int kvm_isolation_init_mm(void)
 {
 	pgd_t *kvm_pgd;
 	gfp_t gfp_mask;
+	int rv;
 
 	gfp_mask = GFP_KERNEL | __GFP_ZERO;
 	kvm_pgd = (pgd_t *)__get_free_pages(gfp_mask, PGD_ALLOCATION_ORDER);
@@ -1054,6 +1178,12 @@ static int kvm_isolation_init_mm(void)
 	mm_init_cpumask(&kvm_mm);
 	init_new_context(NULL, &kvm_mm);
 
+	rv = kvm_isolation_init_page_table();
+	if (rv) {
+		kvm_isolation_uninit_mm();
+		return rv;
+	}
+
 	return 0;
 }
 
@@ -1065,6 +1195,7 @@ static void kvm_isolation_uninit_mm(void)
 
 	destroy_context(&kvm_mm);
 
+	kvm_isolation_uninit_page_table();
 	kvm_free_all_range_mapping();
 
 #ifdef CONFIG_PAGE_TABLE_ISOLATION
diff --git a/arch/x86/kvm/isolation.h b/arch/x86/kvm/isolation.h
index 3ef2060..1f79e28 100644
--- a/arch/x86/kvm/isolation.h
+++ b/arch/x86/kvm/isolation.h
@@ -3,6 +3,16 @@
 #define ARCH_X86_KVM_ISOLATION_H
 
 #include <linux/kvm_host.h>
+#include <linux/export.h>
+
+/*
+ * Copy the memory mapping for the current module. This is defined as a
+ * macro to ensure it is expanded in the module making the call so that
+ * THIS_MODULE has the correct value.
+ */
+#define kvm_copy_module_mapping()			\
+	(kvm_copy_ptes(THIS_MODULE->core_layout.base,	\
+	    THIS_MODULE->core_layout.size))
 
 DECLARE_STATIC_KEY_FALSE(kvm_isolation_enabled);
 
diff --git a/include/linux/percpu.h b/include/linux/percpu.h
index 70b7123..fb0ab9a 100644
--- a/include/linux/percpu.h
+++ b/include/linux/percpu.h
@@ -70,6 +70,8 @@
 
 extern void *pcpu_base_addr;
 extern const unsigned long *pcpu_unit_offsets;
+extern int pcpu_unit_size;
+extern int pcpu_nr_units;
 
 struct pcpu_group_info {
 	int			nr_units;	/* aligned # of units */
diff --git a/mm/percpu.c b/mm/percpu.c
index 68dd2e7..b68b3d8 100644
--- a/mm/percpu.c
+++ b/mm/percpu.c
@@ -119,8 +119,10 @@
 #endif	/* CONFIG_SMP */
 
 static int pcpu_unit_pages __ro_after_init;
-static int pcpu_unit_size __ro_after_init;
-static int pcpu_nr_units __ro_after_init;
+int pcpu_unit_size __ro_after_init;
+EXPORT_SYMBOL(pcpu_unit_size);
+int pcpu_nr_units __ro_after_init;
+EXPORT_SYMBOL(pcpu_nr_units);
 static int pcpu_atom_size __ro_after_init;
 int pcpu_nr_slots __ro_after_init;
 static size_t pcpu_chunk_struct_size __ro_after_init;
-- 
1.7.1