x86: Add support for starting 64-bit kernel

Add code to jump to a 64-bit Linux kernel. We need to set up a flat page
table structure, a new GDT and then go through a few hoops in the right
order.

Signed-off-by: Simon Glass <sjg@chromium.org>

arch/x86/cpu/Makefile

@@ -10,4 +10,4 @@
 
 extra-y	= start.o
 obj-$(CONFIG_X86_RESET_VECTOR) += resetvec.o start16.o
-obj-y	+= interrupts.o cpu.o
+obj-y	+= interrupts.o cpu.o call64.o

arch/x86/cpu/call64.S

@@ -0,0 +1,93 @@
+/*
+ * (C) Copyright 2014 Google, Inc
+ * Copyright (C) 1991, 1992, 1993  Linus Torvalds
+ *
+ * Parts of this copied from Linux arch/x86/boot/compressed/head_64.S
+ *
+ * SPDX-License-Identifier:	GPL-2.0+
+ */
+
+#include <asm/global_data.h>
+#include <asm/msr-index.h>
+#include <asm/processor-flags.h>
+
+.code32
+.globl cpu_call64
+cpu_call64:
+	/*
+	 * cpu_call64(ulong pgtable, ulong setup_base, ulong target)
+	 *
+	 * eax - pgtable
+	 * edx - setup_base
+	 * ecx - target
+	 */
+	cli
+	push	%ecx		/* arg2 = target */
+	push	%edx		/* arg1 = setup_base */
+	mov	%eax, %ebx
+
+	/* Load new GDT with the 64bit segments using 32bit descriptor */
+	leal	gdt, %eax
+	movl	%eax, gdt+2
+	lgdt	gdt
+
+	/* Enable PAE mode */
+	movl	$(X86_CR4_PAE), %eax
+	movl	%eax, %cr4
+
+	/* Enable the boot page tables */
+	leal	(%ebx), %eax
+	movl	%eax, %cr3
+
+	/* Enable Long mode in EFER (Extended Feature Enable Register) */
+	movl	$MSR_EFER, %ecx
+	rdmsr
+	btsl	$_EFER_LME, %eax
+	wrmsr
+
+	/* After gdt is loaded */
+	xorl	%eax, %eax
+	lldt	%ax
+	movl    $0x20, %eax
+	ltr	%ax
+
+	/*
+	 * Setup for the jump to 64bit mode
+	 *
+	 * When the jump is performed we will be in long mode but
+	 * in 32bit compatibility mode with EFER.LME = 1, CS.L = 0, CS.D = 1
+	 * (and in turn EFER.LMA = 1). To jump into 64bit mode we use
+	 * the new gdt/idt that has __KERNEL_CS with CS.L = 1.
+	 * We place all of the values on our mini stack so lret can
+	 * used to perform that far jump. See the gdt below.
+	 */
+	pop	%esi			/* setup_base */
+
+	pushl	$0x10
+	leal	lret_target, %eax
+	pushl	%eax
+
+	/* Enter paged protected Mode, activating Long Mode */
+	movl	$(X86_CR0_PG | X86_CR0_PE), %eax
+	movl	%eax, %cr0
+
+	/* Jump from 32bit compatibility mode into 64bit mode. */
+	lret
+
+code64:
+lret_target:
+	pop	%eax			/* target */
+	mov	%eax, %eax		/* Clear bits 63:32 */
+	jmp	*%eax			/* Jump to the 64-bit target */
+
+	.data
+gdt:
+	.word	gdt_end - gdt
+	.long	gdt
+	.word	0
+	.quad	0x0000000000000000	/* NULL descriptor */
+	.quad	0x00af9a000000ffff	/* __KERNEL_CS */
+	.quad	0x00cf92000000ffff	/* __KERNEL_DS */
+	.quad	0x0080890000000000	/* TS descriptor */
+	.quad   0x0000000000000000	/* TS continued */
+gdt_end:

arch/x86/cpu/cpu.c

@@ -18,7 +18,10 @@
 
 #include <common.h>
 #include <command.h>
+#include <errno.h>
+#include <malloc.h>
 #include <asm/control_regs.h>
+#include <asm/cpu.h>
 #include <asm/processor.h>
 #include <asm/processor-flags.h>
 #include <asm/interrupt.h>
@@ -339,3 +342,45 @@ int print_cpuinfo(void)
 
 	return 0;
 }
+
+#define PAGETABLE_SIZE		(6 * 4096)
+
+/**
+ * build_pagetable() - build a flat 4GiB page table structure for 64-bti mode
+ *
+ * @pgtable: Pointer to a 24iKB block of memory
+ */
+static void build_pagetable(uint32_t *pgtable)
+{
+	uint i;
+
+	memset(pgtable, '\0', PAGETABLE_SIZE);
+
+	/* Level 4 needs a single entry */
+	pgtable[0] = (uint32_t)&pgtable[1024] + 7;
+
+	/* Level 3 has one 64-bit entry for each GiB of memory */
+	for (i = 0; i < 4; i++) {
+		pgtable[1024 + i * 2] = (uint32_t)&pgtable[2048] +
+							0x1000 * i + 7;
+	}
+
+	/* Level 2 has 2048 64-bit entries, each repesenting 2MiB */
+	for (i = 0; i < 2048; i++)
+		pgtable[2048 + i * 2] = 0x183 + (i << 21UL);
+}
+
+int cpu_jump_to_64bit(ulong setup_base, ulong target)
+{
+	uint32_t *pgtable;
+
+	pgtable = memalign(4096, PAGETABLE_SIZE);
+	if (!pgtable)
+		return -ENOMEM;
+
+	build_pagetable(pgtable);
+	cpu_call64((ulong)pgtable, setup_base, target);
+	free(pgtable);
+
+	return -EFAULT;
+}

arch/x86/include/asm/cpu.h

@@ -26,4 +26,30 @@ void cpu_disable_paging_pae(void);
  */
 int cpu_has_64bit(void);
 
+/**
+ * cpu_call64() - Jump to a 64-bit Linux kernel (internal function)
+ *
+ * The kernel is uncompressed and the 64-bit entry point is expected to be
+ * at @target.
+ *
+ * This function is used internally - see cpu_jump_to_64bit() for a more
+ * useful function.
+ *
+ * @pgtable:	Address of 24KB area containing the page table
+ * @setup_base:	Pointer to the setup.bin information for the kernel
+ * @target:	Pointer to the start of the kernel image
+ */
+void cpu_call64(ulong pgtable, ulong setup_base, ulong target);
+
+/**
+ * cpu_jump_to_64bit() - Jump to a 64-bit Linux kernel
+ *
+ * The kernel is uncompressed and the 64-bit entry point is expected to be
+ * at @target.
+ *
+ * @setup_base:	Pointer to the setup.bin information for the kernel
+ * @target:	Pointer to the start of the kernel image
+ */
+int cpu_jump_to_64bit(ulong setup_base, ulong target);
+
 #endif