KVM: ARM: Hypervisor initialization

Sets up KVM code to handle all exceptions taken to Hyp mode.

When the kernel is booted in Hyp mode, calling an hvc instruction with r0
pointing to the new vectors, the HVBAR is changed to the the vector pointers.
This allows subsystems (like KVM here) to execute code in Hyp-mode with the
MMU disabled.

We initialize other Hyp-mode registers and enables the MMU for Hyp-mode from
the id-mapped hyp initialization code. Afterwards, the HVBAR is changed to
point to KVM Hyp vectors used to catch guest faults and to switch to Hyp mode
to perform a world-switch into a KVM guest.

Also provides memory mapping code to map required code pages, data structures,
and I/O regions  accessed in Hyp mode at the same virtual address as the host
kernel virtual addresses, but which conforms to the architectural requirements
for translations in Hyp mode. This interface is added in arch/arm/kvm/arm_mmu.c
and comprises:
 - create_hyp_mappings(from, to);
 - create_hyp_io_mappings(from, to, phys_addr);
 - free_hyp_pmds();

Reviewed-by: Will Deacon <will.deacon@arm.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <c.dall@virtualopensystems.com>

4 files changed, 577 insertions, 0 deletions

diff --git a/arch/arm/kvm/arm.c b/arch/arm/kvm/arm.c
index d3506b4001a..2c6b780e78a 100644
--- a/arch/arm/kvm/arm.c
+++ b/arch/arm/kvm/arm.c
@@ -34,11 +34,21 @@
 #include <asm/ptrace.h>
 #include <asm/mman.h>
 #include <asm/cputype.h>
+#include <asm/tlbflush.h>
+#include <asm/virt.h>
+#include <asm/kvm_arm.h>
+#include <asm/kvm_asm.h>
+#include <asm/kvm_mmu.h>
 
 #ifdef REQUIRES_VIRT
 __asm__(".arch_extension	virt");
 #endif
 
+static DEFINE_PER_CPU(unsigned long, kvm_arm_hyp_stack_page);
+static struct vfp_hard_struct __percpu *kvm_host_vfp_state;
+static unsigned long hyp_default_vectors;
+
+
 int kvm_arch_hardware_enable(void *garbage)
 {
 	return 0;
@@ -331,9 +341,171 @@ long kvm_arch_vm_ioctl(struct file *filp,
 	return -EINVAL;
 }
 
+static void cpu_init_hyp_mode(void *vector)
+{
+	unsigned long long pgd_ptr;
+	unsigned long pgd_low, pgd_high;
+	unsigned long hyp_stack_ptr;
+	unsigned long stack_page;
+	unsigned long vector_ptr;
+
+	/* Switch from the HYP stub to our own HYP init vector */
+	__hyp_set_vectors((unsigned long)vector);
+
+	pgd_ptr = (unsigned long long)kvm_mmu_get_httbr();
+	pgd_low = (pgd_ptr & ((1ULL << 32) - 1));
+	pgd_high = (pgd_ptr >> 32ULL);
+	stack_page = __get_cpu_var(kvm_arm_hyp_stack_page);
+	hyp_stack_ptr = stack_page + PAGE_SIZE;
+	vector_ptr = (unsigned long)__kvm_hyp_vector;
+
+	/*
+	 * Call initialization code, and switch to the full blown
+	 * HYP code. The init code doesn't need to preserve these registers as
+	 * r1-r3 and r12 are already callee save according to the AAPCS.
+	 * Note that we slightly misuse the prototype by casing the pgd_low to
+	 * a void *.
+	 */
+	kvm_call_hyp((void *)pgd_low, pgd_high, hyp_stack_ptr, vector_ptr);
+}
+
+/**
+ * Inits Hyp-mode on all online CPUs
+ */
+static int init_hyp_mode(void)
+{
+	phys_addr_t init_phys_addr;
+	int cpu;
+	int err = 0;
+
+	/*
+	 * Allocate Hyp PGD and setup Hyp identity mapping
+	 */
+	err = kvm_mmu_init();
+	if (err)
+		goto out_err;
+
+	/*
+	 * It is probably enough to obtain the default on one
+	 * CPU. It's unlikely to be different on the others.
+	 */
+	hyp_default_vectors = __hyp_get_vectors();
+
+	/*
+	 * Allocate stack pages for Hypervisor-mode
+	 */
+	for_each_possible_cpu(cpu) {
+		unsigned long stack_page;
+
+		stack_page = __get_free_page(GFP_KERNEL);
+		if (!stack_page) {
+			err = -ENOMEM;
+			goto out_free_stack_pages;
+		}
+
+		per_cpu(kvm_arm_hyp_stack_page, cpu) = stack_page;
+	}
+
+	/*
+	 * Execute the init code on each CPU.
+	 *
+	 * Note: The stack is not mapped yet, so don't do anything else than
+	 * initializing the hypervisor mode on each CPU using a local stack
+	 * space for temporary storage.
+	 */
+	init_phys_addr = virt_to_phys(__kvm_hyp_init);
+	for_each_online_cpu(cpu) {
+		smp_call_function_single(cpu, cpu_init_hyp_mode,
+					 (void *)(long)init_phys_addr, 1);
+	}
+
+	/*
+	 * Unmap the identity mapping
+	 */
+	kvm_clear_hyp_idmap();
+
+	/*
+	 * Map the Hyp-code called directly from the host
+	 */
+	err = create_hyp_mappings(__kvm_hyp_code_start, __kvm_hyp_code_end);
+	if (err) {
+		kvm_err("Cannot map world-switch code\n");
+		goto out_free_mappings;
+	}
+
+	/*
+	 * Map the Hyp stack pages
+	 */
+	for_each_possible_cpu(cpu) {
+		char *stack_page = (char *)per_cpu(kvm_arm_hyp_stack_page, cpu);
+		err = create_hyp_mappings(stack_page, stack_page + PAGE_SIZE);
+
+		if (err) {
+			kvm_err("Cannot map hyp stack\n");
+			goto out_free_mappings;
+		}
+	}
+
+	/*
+	 * Map the host VFP structures
+	 */
+	kvm_host_vfp_state = alloc_percpu(struct vfp_hard_struct);
+	if (!kvm_host_vfp_state) {
+		err = -ENOMEM;
+		kvm_err("Cannot allocate host VFP state\n");
+		goto out_free_mappings;
+	}
+
+	for_each_possible_cpu(cpu) {
+		struct vfp_hard_struct *vfp;
+
+		vfp = per_cpu_ptr(kvm_host_vfp_state, cpu);
+		err = create_hyp_mappings(vfp, vfp + 1);
+
+		if (err) {
+			kvm_err("Cannot map host VFP state: %d\n", err);
+			goto out_free_vfp;
+		}
+	}
+
+	kvm_info("Hyp mode initialized successfully\n");
+	return 0;
+out_free_vfp:
+	free_percpu(kvm_host_vfp_state);
+out_free_mappings:
+	free_hyp_pmds();
+out_free_stack_pages:
+	for_each_possible_cpu(cpu)
+		free_page(per_cpu(kvm_arm_hyp_stack_page, cpu));
+out_err:
+	kvm_err("error initializing Hyp mode: %d\n", err);
+	return err;
+}
+
+/**
+ * Initialize Hyp-mode and memory mappings on all CPUs.
+ */
 int kvm_arch_init(void *opaque)
 {
+	int err;
+
+	if (!is_hyp_mode_available()) {
+		kvm_err("HYP mode not available\n");
+		return -ENODEV;
+	}
+
+	if (kvm_target_cpu() < 0) {
+		kvm_err("Target CPU not supported!\n");
+		return -ENODEV;
+	}
+
+	err = init_hyp_mode();
+	if (err)
+		goto out_err;
+
 	return 0;
+out_err:
+	return err;
 }
 
 /* NOP: Compiling as a module not supported */
diff --git a/arch/arm/kvm/init.S b/arch/arm/kvm/init.S
index 1dc8926e26d..9f37a79b880 100644
--- a/arch/arm/kvm/init.S
+++ b/arch/arm/kvm/init.S
@@ -15,5 +15,100 @@
  * along with this program; if not, write to the Free Software
  * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
  */
+
+#include <linux/linkage.h>
+#include <asm/unified.h>
 #include <asm/asm-offsets.h>
 #include <asm/kvm_asm.h>
+#include <asm/kvm_arm.h>
+
+/********************************************************************
+ * Hypervisor initialization
+ *   - should be called with:
+ *       r0,r1 = Hypervisor pgd pointer
+ *       r2 = top of Hyp stack (kernel VA)
+ *       r3 = pointer to hyp vectors
+ */
+
+	.text
+	.pushsection    .hyp.idmap.text,"ax"
+	.align 5
+__kvm_hyp_init:
+	.globl __kvm_hyp_init
+
+	@ Hyp-mode exception vector
+	W(b)	.
+	W(b)	.
+	W(b)	.
+	W(b)	.
+	W(b)	.
+	W(b)	__do_hyp_init
+	W(b)	.
+	W(b)	.
+
+__do_hyp_init:
+	@ Set the HTTBR to point to the hypervisor PGD pointer passed
+	mcrr	p15, 4, r0, r1, c2
+
+	@ Set the HTCR and VTCR to the same shareability and cacheability
+	@ settings as the non-secure TTBCR and with T0SZ == 0.
+	mrc	p15, 4, r0, c2, c0, 2	@ HTCR
+	ldr	r12, =HTCR_MASK
+	bic	r0, r0, r12
+	mrc	p15, 0, r1, c2, c0, 2	@ TTBCR
+	and	r1, r1, #(HTCR_MASK & ~TTBCR_T0SZ)
+	orr	r0, r0, r1
+	mcr	p15, 4, r0, c2, c0, 2	@ HTCR
+
+	mrc	p15, 4, r1, c2, c1, 2	@ VTCR
+	ldr	r12, =VTCR_MASK
+	bic	r1, r1, r12
+	bic	r0, r0, #(~VTCR_HTCR_SH)	@ clear non-reusable HTCR bits
+	orr	r1, r0, r1
+	orr	r1, r1, #(KVM_VTCR_SL0 | KVM_VTCR_T0SZ | KVM_VTCR_S)
+	mcr	p15, 4, r1, c2, c1, 2	@ VTCR
+
+	@ Use the same memory attributes for hyp. accesses as the kernel
+	@ (copy MAIRx ro HMAIRx).
+	mrc	p15, 0, r0, c10, c2, 0
+	mcr	p15, 4, r0, c10, c2, 0
+	mrc	p15, 0, r0, c10, c2, 1
+	mcr	p15, 4, r0, c10, c2, 1
+
+	@ Set the HSCTLR to:
+	@  - ARM/THUMB exceptions: Kernel config (Thumb-2 kernel)
+	@  - Endianness: Kernel config
+	@  - Fast Interrupt Features: Kernel config
+	@  - Write permission implies XN: disabled
+	@  - Instruction cache: enabled
+	@  - Data/Unified cache: enabled
+	@  - Memory alignment checks: enabled
+	@  - MMU: enabled (this code must be run from an identity mapping)
+	mrc	p15, 4, r0, c1, c0, 0	@ HSCR
+	ldr	r12, =HSCTLR_MASK
+	bic	r0, r0, r12
+	mrc	p15, 0, r1, c1, c0, 0	@ SCTLR
+	ldr	r12, =(HSCTLR_EE | HSCTLR_FI | HSCTLR_I | HSCTLR_C)
+	and	r1, r1, r12
+ ARM(	ldr	r12, =(HSCTLR_M | HSCTLR_A)			)
+ THUMB(	ldr	r12, =(HSCTLR_M | HSCTLR_A | HSCTLR_TE)		)
+	orr	r1, r1, r12
+	orr	r0, r0, r1
+	isb
+	mcr	p15, 4, r0, c1, c0, 0	@ HSCR
+	isb
+
+	@ Set stack pointer and return to the kernel
+	mov	sp, r2
+
+	@ Set HVBAR to point to the HYP vectors
+	mcr	p15, 4, r3, c12, c0, 0	@ HVBAR
+
+	eret
+
+	.ltorg
+
+	.globl __kvm_hyp_init_end
+__kvm_hyp_init_end:
+
+	.popsection
diff --git a/arch/arm/kvm/interrupts.S b/arch/arm/kvm/interrupts.S
index 1dc8926e26d..d10a8075409 100644
--- a/arch/arm/kvm/interrupts.S
+++ b/arch/arm/kvm/interrupts.S
@@ -15,5 +15,67 @@
  * along with this program; if not, write to the Free Software
  * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
  */
+
+#include <linux/linkage.h>
+#include <linux/const.h>
+#include <asm/unified.h>
+#include <asm/page.h>
 #include <asm/asm-offsets.h>
 #include <asm/kvm_asm.h>
+#include <asm/kvm_arm.h>
+
+	.text
+
+__kvm_hyp_code_start:
+	.globl __kvm_hyp_code_start
+
+/********************************************************************
+ * Flush per-VMID TLBs
+ */
+ENTRY(__kvm_flush_vm_context)
+	bx	lr
+ENDPROC(__kvm_flush_vm_context)
+
+/********************************************************************
+ *  Hypervisor world-switch code
+ */
+ENTRY(__kvm_vcpu_run)
+	bx	lr
+
+/********************************************************************
+ *  Call function in Hyp mode
+ *
+ *
+ * u64 kvm_call_hyp(void *hypfn, ...);
+ *
+ * This is not really a variadic function in the classic C-way and care must
+ * be taken when calling this to ensure parameters are passed in registers
+ * only, since the stack will change between the caller and the callee.
+ *
+ * Call the function with the first argument containing a pointer to the
+ * function you wish to call in Hyp mode, and subsequent arguments will be
+ * passed as r0, r1, and r2 (a maximum of 3 arguments in addition to the
+ * function pointer can be passed).  The function being called must be mapped
+ * in Hyp mode (see init_hyp_mode in arch/arm/kvm/arm.c).  Return values are
+ * passed in r0 and r1.
+ *
+ * The calling convention follows the standard AAPCS:
+ *   r0 - r3: caller save
+ *   r12:     caller save
+ *   rest:    callee save
+ */
+ENTRY(kvm_call_hyp)
+	hvc	#0
+	bx	lr
+
+/********************************************************************
+ * Hypervisor exception vector and handlers
+ */
+
+	.align 5
+__kvm_hyp_vector:
+	.globl __kvm_hyp_vector
+	nop
+
+__kvm_hyp_code_end:
+	.globl	__kvm_hyp_code_end
diff --git a/arch/arm/kvm/mmu.c b/arch/arm/kvm/mmu.c
index 10ed4643269..4decdb61801 100644
--- a/arch/arm/kvm/mmu.c
+++ b/arch/arm/kvm/mmu.c
@@ -15,3 +15,251 @@
  * along with this program; if not, write to the Free Software
  * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
  */
+
+#include <linux/mman.h>
+#include <linux/kvm_host.h>
+#include <linux/io.h>
+#include <asm/idmap.h>
+#include <asm/pgalloc.h>
+#include <asm/kvm_arm.h>
+#include <asm/kvm_mmu.h>
+#include <asm/mach/map.h>
+
+extern char  __hyp_idmap_text_start[], __hyp_idmap_text_end[];
+
+static DEFINE_MUTEX(kvm_hyp_pgd_mutex);
+
+static void kvm_set_pte(pte_t *pte, pte_t new_pte)
+{
+	pte_val(*pte) = new_pte;
+	/*
+	 * flush_pmd_entry just takes a void pointer and cleans the necessary
+	 * cache entries, so we can reuse the function for ptes.
+	 */
+	flush_pmd_entry(pte);
+}
+
+static void free_ptes(pmd_t *pmd, unsigned long addr)
+{
+	pte_t *pte;
+	unsigned int i;
+
+	for (i = 0; i < PTRS_PER_PMD; i++, addr += PMD_SIZE) {
+		if (!pmd_none(*pmd) && pmd_table(*pmd)) {
+			pte = pte_offset_kernel(pmd, addr);
+			pte_free_kernel(NULL, pte);
+		}
+		pmd++;
+	}
+}
+
+/**
+ * free_hyp_pmds - free a Hyp-mode level-2 tables and child level-3 tables
+ *
+ * Assumes this is a page table used strictly in Hyp-mode and therefore contains
+ * only mappings in the kernel memory area, which is above PAGE_OFFSET.
+ */
+void free_hyp_pmds(void)
+{
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	unsigned long addr;
+
+	mutex_lock(&kvm_hyp_pgd_mutex);
+	for (addr = PAGE_OFFSET; addr != 0; addr += PGDIR_SIZE) {
+		pgd = hyp_pgd + pgd_index(addr);
+		pud = pud_offset(pgd, addr);
+
+		if (pud_none(*pud))
+			continue;
+		BUG_ON(pud_bad(*pud));
+
+		pmd = pmd_offset(pud, addr);
+		free_ptes(pmd, addr);
+		pmd_free(NULL, pmd);
+		pud_clear(pud);
+	}
+	mutex_unlock(&kvm_hyp_pgd_mutex);
+}
+
+static void create_hyp_pte_mappings(pmd_t *pmd, unsigned long start,
+				    unsigned long end)
+{
+	pte_t *pte;
+	unsigned long addr;
+	struct page *page;
+
+	for (addr = start & PAGE_MASK; addr < end; addr += PAGE_SIZE) {
+		pte = pte_offset_kernel(pmd, addr);
+		BUG_ON(!virt_addr_valid(addr));
+		page = virt_to_page(addr);
+		kvm_set_pte(pte, mk_pte(page, PAGE_HYP));
+	}
+}
+
+static void create_hyp_io_pte_mappings(pmd_t *pmd, unsigned long start,
+				       unsigned long end,
+				       unsigned long *pfn_base)
+{
+	pte_t *pte;
+	unsigned long addr;
+
+	for (addr = start & PAGE_MASK; addr < end; addr += PAGE_SIZE) {
+		pte = pte_offset_kernel(pmd, addr);
+		BUG_ON(pfn_valid(*pfn_base));
+		kvm_set_pte(pte, pfn_pte(*pfn_base, PAGE_HYP_DEVICE));
+		(*pfn_base)++;
+	}
+}
+
+static int create_hyp_pmd_mappings(pud_t *pud, unsigned long start,
+				   unsigned long end, unsigned long *pfn_base)
+{
+	pmd_t *pmd;
+	pte_t *pte;
+	unsigned long addr, next;
+
+	for (addr = start; addr < end; addr = next) {
+		pmd = pmd_offset(pud, addr);
+
+		BUG_ON(pmd_sect(*pmd));
+
+		if (pmd_none(*pmd)) {
+			pte = pte_alloc_one_kernel(NULL, addr);
+			if (!pte) {
+				kvm_err("Cannot allocate Hyp pte\n");
+				return -ENOMEM;
+			}
+			pmd_populate_kernel(NULL, pmd, pte);
+		}
+
+		next = pmd_addr_end(addr, end);
+
+		/*
+		 * If pfn_base is NULL, we map kernel pages into HYP with the
+		 * virtual address. Otherwise, this is considered an I/O
+		 * mapping and we map the physical region starting at
+		 * *pfn_base to [start, end[.
+		 */
+		if (!pfn_base)
+			create_hyp_pte_mappings(pmd, addr, next);
+		else
+			create_hyp_io_pte_mappings(pmd, addr, next, pfn_base);
+	}
+
+	return 0;
+}
+
+static int __create_hyp_mappings(void *from, void *to, unsigned long *pfn_base)
+{
+	unsigned long start = (unsigned long)from;
+	unsigned long end = (unsigned long)to;
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	unsigned long addr, next;
+	int err = 0;
+
+	BUG_ON(start > end);
+	if (start < PAGE_OFFSET)
+		return -EINVAL;
+
+	mutex_lock(&kvm_hyp_pgd_mutex);
+	for (addr = start; addr < end; addr = next) {
+		pgd = hyp_pgd + pgd_index(addr);
+		pud = pud_offset(pgd, addr);
+
+		if (pud_none_or_clear_bad(pud)) {
+			pmd = pmd_alloc_one(NULL, addr);
+			if (!pmd) {
+				kvm_err("Cannot allocate Hyp pmd\n");
+				err = -ENOMEM;
+				goto out;
+			}
+			pud_populate(NULL, pud, pmd);
+		}
+
+		next = pgd_addr_end(addr, end);
+		err = create_hyp_pmd_mappings(pud, addr, next, pfn_base);
+		if (err)
+			goto out;
+	}
+out:
+	mutex_unlock(&kvm_hyp_pgd_mutex);
+	return err;
+}
+
+/**
+ * create_hyp_mappings - map a kernel virtual address range in Hyp mode
+ * @from:	The virtual kernel start address of the range
+ * @to:		The virtual kernel end address of the range (exclusive)
+ *
+ * The same virtual address as the kernel virtual address is also used in
+ * Hyp-mode mapping to the same underlying physical pages.
+ *
+ * Note: Wrapping around zero in the "to" address is not supported.
+ */
+int create_hyp_mappings(void *from, void *to)
+{
+	return __create_hyp_mappings(from, to, NULL);
+}
+
+/**
+ * create_hyp_io_mappings - map a physical IO range in Hyp mode
+ * @from:	The virtual HYP start address of the range
+ * @to:		The virtual HYP end address of the range (exclusive)
+ * @addr:	The physical start address which gets mapped
+ */
+int create_hyp_io_mappings(void *from, void *to, phys_addr_t addr)
+{
+	unsigned long pfn = __phys_to_pfn(addr);
+	return __create_hyp_mappings(from, to, &pfn);
+}
+
+int kvm_handle_guest_abort(struct kvm_vcpu *vcpu, struct kvm_run *run)
+{
+	return -EINVAL;
+}
+
+phys_addr_t kvm_mmu_get_httbr(void)
+{
+	VM_BUG_ON(!virt_addr_valid(hyp_pgd));
+	return virt_to_phys(hyp_pgd);
+}
+
+int kvm_mmu_init(void)
+{
+	return hyp_pgd ? 0 : -ENOMEM;
+}
+
+/**
+ * kvm_clear_idmap - remove all idmaps from the hyp pgd
+ *
+ * Free the underlying pmds for all pgds in range and clear the pgds (but
+ * don't free them) afterwards.
+ */
+void kvm_clear_hyp_idmap(void)
+{
+	unsigned long addr, end;
+	unsigned long next;
+	pgd_t *pgd = hyp_pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+
+	addr = virt_to_phys(__hyp_idmap_text_start);
+	end = virt_to_phys(__hyp_idmap_text_end);
+
+	pgd += pgd_index(addr);
+	do {
+		next = pgd_addr_end(addr, end);
+		if (pgd_none_or_clear_bad(pgd))
+			continue;
+		pud = pud_offset(pgd, addr);
+		pmd = pmd_offset(pud, addr);
+
+		pud_clear(pud);
+		clean_pmd_entry(pmd);
+		pmd_free(NULL, (pmd_t *)((unsigned long)pmd & PAGE_MASK));
+	} while (pgd++, addr = next, addr < end);
+}