x86, realmode: Move SMP trampoline to unified realmode code

Migrated SMP trampoline code to the real mode blob.
SMP trampoline code is not yet removed from
.x86_trampoline because it is needed by the wakeup
code.

[ hpa: always enable compiling startup_32_smp in head_32.S... it is
  only a few instructions which go into .init on UP builds, and it makes
  the rest of the code less #ifdef ugly. ]

Signed-off-by: Jarkko Sakkinen <jarkko.sakkinen@intel.com>
Link: http://lkml.kernel.org/r/1336501366-28617-6-git-send-email-jarkko.sakkinen@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>

arch/x86/realmode/rm/Makefile

@@ -13,6 +13,7 @@ always := realmode.bin
 
 realmode-y			+= header.o
 realmode-$(CONFIG_X86_32)	+= reboot_32.o
+realmode-y			+= trampoline_$(BITS).o
 
 targets	+= $(realmode-y)
 

arch/x86/realmode/rm/header.S

@@ -15,5 +15,16 @@ ENTRY(real_mode_header)
 		.long	pa_end
 #ifdef CONFIG_X86_32
 		.long	pa_machine_real_restart_asm
+#endif
+		/* SMP trampoline */
+		.long	pa_trampoline_data
+		.long	pa_trampoline_status
+#ifdef CONFIG_X86_32
+		.long	pa_startup_32_smp
+		.long	pa_boot_gdt
+#else
+		.long	pa_startup_64_smp
+		.long	pa_level3_ident_pgt
+		.long	pa_level3_kernel_pgt
 #endif
 END(real_mode_header)

arch/x86/realmode/rm/trampoline_32.S

@@ -0,0 +1,86 @@
+/*
+ *
+ *	Trampoline.S	Derived from Setup.S by Linus Torvalds
+ *
+ *	4 Jan 1997 Michael Chastain: changed to gnu as.
+ *
+ *	This is only used for booting secondary CPUs in SMP machine
+ *
+ *	Entry: CS:IP point to the start of our code, we are
+ *	in real mode with no stack, but the rest of the
+ *	trampoline page to make our stack and everything else
+ *	is a mystery.
+ *
+ *	We jump into arch/x86/kernel/head_32.S.
+ *
+ *	On entry to trampoline_data, the processor is in real mode
+ *	with 16-bit addressing and 16-bit data.  CS has some value
+ *	and IP is zero.  Thus, we load CS to the physical segment
+ *	of the real mode code before doing anything further.
+ *
+ *	The structure real_mode_header includes entries that need
+ *	to be set up before executing this code:
+ *
+ *	startup_32_smp
+ *	boot_gdt
+ */
+
+#include <linux/linkage.h>
+#include <linux/init.h>
+#include <asm/segment.h>
+#include <asm/page_types.h>
+
+	.text
+	.code16
+	.globl trampoline_data
+
+	.balign PAGE_SIZE
+trampoline_data:
+	wbinvd			# Needed for NUMA-Q should be harmless for others
+
+	.byte	0xea		# ljmpw
+	.word	1f		# Offset
+	.word	real_mode_seg	# Segment
+1:
+	mov	%cs, %ax	# Code and data in the same place
+	mov	%ax, %ds
+
+	cli			# We should be safe anyway
+
+	movl	$0xA5A5A5A5, trampoline_status
+				# write marker for master knows we're running
+
+	/* GDT tables in non default location kernel can be beyond 16MB and
+	 * lgdt will not be able to load the address as in real mode default
+	 * operand size is 16bit. Use lgdtl instead to force operand size
+	 * to 32 bit.
+	 */
+
+	lidtl	boot_idt_descr		# load idt with 0, 0
+	lgdtl	boot_gdt_descr		# load gdt with whatever is appropriate
+
+	xor	%ax, %ax
+	inc	%ax			# protected mode (PE) bit
+	lmsw	%ax			# into protected mode
+
+	# flush prefetch and jump to startup_32_smp in arch/i386/kernel/head.S
+	ljmpl	*(startup_32_smp)
+
+	.data
+	.globl startup_32_smp, boot_gdt, trampoline_status
+
+boot_gdt_descr:
+	.word	__BOOT_DS + 7			# gdt limit
+boot_gdt:
+	.long	0				# gdt base
+
+boot_idt_descr:
+	.word	0				# idt limit = 0
+	.long	0				# idt base = 0L
+
+trampoline_status:
+	.long	0
+
+startup_32_smp:
+	.long	0x00000000
+	.word	__BOOT_CS, 0

arch/x86/realmode/rm/trampoline_64.S

@@ -0,0 +1,175 @@
+/*
+ *
+ *	Trampoline.S	Derived from Setup.S by Linus Torvalds
+ *
+ *	4 Jan 1997 Michael Chastain: changed to gnu as.
+ *	15 Sept 2005 Eric Biederman: 64bit PIC support
+ *
+ *	Entry: CS:IP point to the start of our code, we are
+ *	in real mode with no stack, but the rest of the
+ *	trampoline page to make our stack and everything else
+ *	is a mystery.
+ *
+ *	On entry to trampoline_data, the processor is in real mode
+ *	with 16-bit addressing and 16-bit data.  CS has some value
+ *	and IP is zero.  Thus, data addresses need to be absolute
+ *	(no relocation) and are taken with regard to r_base.
+ *
+ *	With the addition of trampoline_level4_pgt this code can
+ *	now enter a 64bit kernel that lives at arbitrary 64bit
+ *	physical addresses.
+ *
+ *	If you work on this file, check the object module with objdump
+ *	--full-contents --reloc to make sure there are no relocation
+ *	entries.
+ */
+
+#include <linux/linkage.h>
+#include <linux/init.h>
+#include <asm/pgtable_types.h>
+#include <asm/page_types.h>
+#include <asm/msr.h>
+#include <asm/segment.h>
+#include <asm/processor-flags.h>
+
+	.text
+	.balign PAGE_SIZE
+	.code16
+
+ENTRY(trampoline_data)
+	cli			# We should be safe anyway
+	wbinvd
+
+	.byte	0xea		# ljmpw
+	.word	1f		# Offset
+	.word	real_mode_seg	# Segment
+1:
+	mov	%cs, %ax	# Code and data in the same place
+	mov	%ax, %ds
+	mov	%ax, %es
+	mov	%ax, %ss
+
+	movl	$0xA5A5A5A5, trampoline_status
+	# write marker for master knows we're running
+
+	# Setup stack
+	movw	$trampoline_stack_end, %sp
+
+	call	verify_cpu		# Verify the cpu supports long mode
+	testl   %eax, %eax		# Check for return code
+	jnz	no_longmode
+
+	/*
+	 * GDT tables in non default location kernel can be beyond 16MB and
+	 * lgdt will not be able to load the address as in real mode default
+	 * operand size is 16bit. Use lgdtl instead to force operand size
+	 * to 32 bit.
+	 */
+
+	lidtl	tidt	# load idt with 0, 0
+	lgdtl	tgdt	# load gdt with whatever is appropriate
+
+	mov	$X86_CR0_PE, %ax	# protected mode (PE) bit
+	lmsw	%ax			# into protected mode
+
+	# flush prefetch and jump to startup_32
+	ljmpl	*(startup_32_vector)
+
+no_longmode:
+	hlt
+	jmp no_longmode
+#include "../kernel/verify_cpu.S"
+
+	.code32
+	.balign 4
+ENTRY(startup_32)
+	movl	$__KERNEL_DS, %eax	# Initialize the %ds segment register
+	movl	%eax, %ds
+
+	movl	$X86_CR4_PAE, %eax
+	movl	%eax, %cr4		# Enable PAE mode
+
+	movl	pa_startup_64_smp, %esi
+	movl	pa_startup_64_smp_high, %edi
+
+					# Setup trampoline 4 level pagetables
+	leal	pa_trampoline_level4_pgt, %eax
+	movl	%eax, %cr3
+
+	movl	$MSR_EFER, %ecx
+	movl	$(1 << _EFER_LME), %eax	# Enable Long Mode
+	xorl	%edx, %edx
+	wrmsr
+
+	# Enable paging and in turn activate Long Mode
+	# Enable protected mode
+	movl	$(X86_CR0_PG | X86_CR0_PE), %eax
+	movl	%eax, %cr0
+
+	/*
+	 * At this point we're in long mode but in 32bit compatibility mode
+	 * with EFER.LME = 1, CS.L = 0, CS.D = 1 (and in turn
+	 * EFER.LMA = 1). Now we want to jump in 64bit mode, to do that we use
+	 * the new gdt/idt that has __KERNEL_CS with CS.L = 1.
+	 */
+	ljmpl	*(pa_startup_64_vector)
+
+	.code64
+	.balign 4
+ENTRY(startup_64)
+	# Now jump into the kernel using virtual addresses
+	movl	%edi, %eax
+	shlq	$32, %rax
+	addl	%esi, %eax
+	jmp	*%rax
+
+	# Careful these need to be in the same 64K segment as the above;
+tidt:
+	.word	0			# idt limit = 0
+	.word	0, 0			# idt base = 0L
+
+	# Duplicate the global descriptor table
+	# so the kernel can live anywhere
+	.balign 4
+	.globl tgdt
+tgdt:
+	.short	tgdt_end - tgdt		# gdt limit
+	.long	pa_tgdt
+	.short	0
+	.quad	0x00cf9b000000ffff	# __KERNEL32_CS
+	.quad	0x00af9b000000ffff	# __KERNEL_CS
+	.quad	0x00cf93000000ffff	# __KERNEL_DS
+tgdt_end:
+
+	.balign 4
+startup_32_vector:
+	.long	pa_startup_32
+	.word	__KERNEL32_CS, 0
+
+	.balign 4
+	.globl startup_64_vector
+startup_64_vector:
+	.long	pa_startup_64
+	.word	__KERNEL_CS, 0
+
+	.data
+
+	.balign 4
+ENTRY(trampoline_status)
+	.long	0
+
+trampoline_stack:
+	.org 0x1000
+trampoline_stack_end:
+
+	.globl	level3_ident_pgt
+	.globl	level3_kernel_pgt
+ENTRY(trampoline_level4_pgt)
+	level3_ident_pgt:	.quad	0
+	.fill 510,8,0
+	level3_kernel_pgt:	.quad	0
+
+	.globl	startup_64_smp
+	.globl	startup_64_smp_high
+startup_64_smp:		.long 0
+startup_64_smp_high:	.long 0