First shot at factoring SMM code into generic parts and southbridge specific

parts.

This should help to reduce the code duplication for Rudolf's K8/VIA SMM
implementation...

Signed-off-by: Stefan Reinauer <stepan@coresystems.de>
Acked-by: Joseph Smith <joe@settoplinux.org>



git-svn-id: svn://svn.coreboot.org/coreboot/trunk@3870 2b7e53f0-3cfb-0310-b3e9-8179ed1497e1

src/cpu/amd/model_10xxx/Config.lb

@@ -31,6 +31,7 @@ dir /cpu/x86/sse
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
 dir /cpu/x86/pae
+dir /cpu/x86/smm
 dir /cpu/amd/mtrr
 dir /cpu/amd/quadcore
 dir /cpu/amd/microcode

src/cpu/amd/model_fxx/Config.lb

@@ -12,6 +12,7 @@ dir /cpu/x86/sse
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
 dir /cpu/x86/pae
+dir /cpu/x86/smm
 dir /cpu/amd/mtrr
 dir /cpu/amd/dualcore
 dir /cpu/amd/microcode

src/cpu/amd/model_gx1/Config.lb

@@ -3,4 +3,5 @@ dir /cpu/x86/fpu
 dir /cpu/x86/mmx
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
+dir /cpu/x86/smm
 driver model_gx1_init.o

src/cpu/amd/model_gx2/Config.lb

@@ -3,6 +3,7 @@ dir /cpu/x86/fpu
 dir /cpu/x86/mmx
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
+dir /cpu/x86/smm
 driver model_gx2_init.o
 object cpubug.o
-object vsmsetup.o
+object vsmsetup.o

src/cpu/amd/model_lx/Config.lb

@@ -3,6 +3,7 @@ dir /cpu/x86/fpu
 dir /cpu/x86/mmx
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
+dir /cpu/x86/smm
 driver model_lx_init.o
 object cpubug.o
-object vsmsetup.o
+object vsmsetup.o

src/cpu/intel/ep80579/Config.lb

@@ -26,5 +26,6 @@ dir /cpu/x86/mmx
 dir /cpu/x86/sse
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
+dir /cpu/x86/smm
 dir /cpu/intel/microcode
 driver ep80579_init.o

src/cpu/intel/model_69x/Config.lb

@@ -5,5 +5,6 @@ dir /cpu/x86/mmx
 dir /cpu/x86/sse
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
+dir /cpu/x86/smm
 dir /cpu/intel/microcode
 driver model_69x_init.o

src/cpu/intel/model_6dx/Config.lb

@@ -5,5 +5,6 @@ dir /cpu/x86/mmx
 dir /cpu/x86/sse
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
+dir /cpu/x86/smm
 dir /cpu/intel/microcode
 driver model_6dx_init.o

src/cpu/intel/model_6ex/Config.lb

@@ -8,6 +8,7 @@ dir /cpu/x86/mmx
 dir /cpu/x86/sse
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
+dir /cpu/x86/smm
 dir /cpu/intel/microcode
 dir /cpu/intel/hyperthreading
 driver model_6ex_init.o

src/cpu/intel/model_6fx/Config.lb

@@ -8,6 +8,7 @@ dir /cpu/x86/mmx
 dir /cpu/x86/sse
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
+dir /cpu/x86/smm
 dir /cpu/intel/microcode
 dir /cpu/intel/hyperthreading
 driver model_6fx_init.o

src/cpu/intel/model_6xx/Config.lb

@@ -5,5 +5,6 @@ dir /cpu/x86/mmx
 dir /cpu/x86/sse
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
+dir /cpu/x86/smm
 dir /cpu/intel/microcode
 driver model_6xx_init.o

src/cpu/intel/model_f0x/Config.lb

@@ -7,5 +7,6 @@ dir /cpu/x86/mmx
 dir /cpu/x86/sse
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
+dir /cpu/x86/smm
 dir /cpu/intel/microcode
 driver model_f0x_init.o

src/cpu/intel/model_f1x/Config.lb

@@ -7,5 +7,6 @@ dir /cpu/x86/mmx
 dir /cpu/x86/sse
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
+dir /cpu/x86/smm
 dir /cpu/intel/microcode
 driver model_f1x_init.o

src/cpu/intel/model_f2x/Config.lb

@@ -7,6 +7,7 @@ dir /cpu/x86/mmx
 dir /cpu/x86/sse
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
+dir /cpu/x86/smm
 dir /cpu/intel/microcode
 dir /cpu/intel/hyperthreading
 driver model_f2x_init.o

src/cpu/intel/model_f3x/Config.lb

@@ -7,6 +7,7 @@ dir /cpu/x86/mmx
 dir /cpu/x86/sse
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
+dir /cpu/x86/smm
 dir /cpu/intel/microcode
 dir /cpu/intel/hyperthreading
 driver model_f3x_init.o

src/cpu/intel/model_f4x/Config.lb

@@ -7,6 +7,7 @@ dir /cpu/x86/mmx
 dir /cpu/x86/sse
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
+dir /cpu/x86/smm
 dir /cpu/intel/microcode
 dir /cpu/intel/hyperthreading
 driver model_f4x_init.o

src/cpu/via/model_c3/Config.lb

@@ -26,5 +26,6 @@ dir /cpu/x86/mmx
 dir /cpu/x86/sse
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
+dir /cpu/x86/smm
 dir /cpu/intel/microcode
 driver model_c3_init.o

src/cpu/via/model_c7/Config.lb

@@ -26,5 +26,6 @@ dir /cpu/x86/mmx
 dir /cpu/x86/sse
 dir /cpu/x86/lapic
 dir /cpu/x86/cache
+dir /cpu/x86/smm
 dir /cpu/intel/microcode
 driver model_c7_init.o

src/cpu/x86/smm/Config.lb

@@ -0,0 +1,47 @@
+##
+## This file is part of the coreboot project.
+##
+## Copyright (C) 2008 coresystems GmbH
+##
+## This program is free software; you can redistribute it and/or modify
+## it under the terms of the GNU General Public License as published by
+## the Free Software Foundation; either version 2 of the License, or
+## (at your option) any later version.
+##
+## This program is distributed in the hope that it will be useful,
+## but WITHOUT ANY WARRANTY; without even the implied warranty of
+## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+## GNU General Public License for more details.
+##
+## You should have received a copy of the GNU General Public License
+## along with this program; if not, write to the Free Software
+## Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301 USA
+##
+
+uses HAVE_SMI_HANDLER
+
+if HAVE_SMI_HANDLER
+	object smmrelocate.S
+
+	smmobject smmhandler.S
+	smmobject smihandler.o
+
+	makerule smm.o
+		depends	"$(SMM-OBJECTS) printk.o vtxprintf.o $(LIBGCC_FILE_NAME)" 
+		action	"$(CC) $(DISTRO_LFLAGS) -nostdlib -r -o $@ $(SMM-OBJECTS) printk.o vtxprintf.o $(LIBGCC_FILE_NAME)"
+	end
+
+	makerule smm
+		depends	"smm.o $(TOP)/src/cpu/x86/smm/smm.ld ldoptions" 
+		action	"$(CC) $(DISTRO_LFLAGS) -nostdlib -nostartfiles -static -o smm.elf -T $(TOP)/src/cpu/x86/smm/smm.ld smm.o"
+		action 	"$(CROSS_COMPILE)nm -n smm.elf | sort > smm.map"
+		action  "$(OBJCOPY) -O binary smm.elf smm"
+	end
+
+	makerule smm_bin.c
+		depends "smm"
+		action "(echo 'unsigned char smm[] = {'; od -vtx1 smm | sed -e 's,^[0-9]* *,,' -e 's:[0-9a-f][0-9a-f] :0x&,:g' -e 's:[0-9a-f][0-9a-f]$$:0x&,:'; echo '}; unsigned int smm_len = '; wc -c smm |awk '{print $$1;}' ; echo ';')  > smm_bin.c"
+	end
+
+	object ./smm_bin.o
+end

src/cpu/x86/smm/smihandler.c

@@ -0,0 +1,203 @@
+/*
+ * This file is part of the coreboot project.
+ *
+ * Copyright (C) 2008 coresystems GmbH
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; version 2 of
+ * the License.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston,
+ * MA 02110-1301 USA
+ */
+
+#include <arch/io.h>
+#include <arch/romcc_io.h>
+#include <console/console.h>
+#include <cpu/x86/cache.h>
+#include <cpu/x86/smm.h>
+
+void southbridge_smi_set_eos(void);
+
+#define DEBUG_SMI
+
+typedef enum { SMI_LOCKED, SMI_UNLOCKED } smi_semaphore;
+
+/* SMI multiprocessing semaphore */
+static volatile smi_semaphore smi_handler_status = SMI_UNLOCKED;
+
+static int smi_obtain_lock(void)
+{
+	u8 ret = SMI_LOCKED;
+
+	asm volatile (
+		"movb %2, %%al\n"
+		"xchgb %%al, %1\n"
+		"movb %%al, %0\n"
+		: "=g" (ret), "=m" (smi_handler_status)
+		: "g" (SMI_LOCKED)
+		: "eax"
+	);
+
+	return (ret == SMI_UNLOCKED);
+}
+
+static void smi_release_lock(void)
+{
+	asm volatile (
+		"movb %1, %%al\n"
+		"xchgb %%al, %0\n"
+		: "=m" (smi_handler_status)
+		: "g" (SMI_UNLOCKED)
+		: "eax"
+	);
+}
+
+#define LAPIC_ID 0xfee00020
+static inline __attribute__((always_inline)) unsigned long nodeid(void)
+{
+	return (*((volatile unsigned long *)(LAPIC_ID)) >> 24);
+}
+
+/* ********************* smi_util ************************* */
+
+/* Data */
+#define UART_RBR 0x00
+#define UART_TBR 0x00
+
+/* Control */
+#define UART_IER 0x01
+#define UART_IIR 0x02
+#define UART_FCR 0x02
+#define UART_LCR 0x03
+#define UART_MCR 0x04
+#define UART_DLL 0x00
+#define UART_DLM 0x01
+
+/* Status */
+#define UART_LSR 0x05
+#define UART_MSR 0x06
+#define UART_SCR 0x07
+
+static int uart_can_tx_byte(void)
+{
+	return inb(TTYS0_BASE + UART_LSR) & 0x20;
+}
+
+static void uart_wait_to_tx_byte(void)
+{
+	while(!uart_can_tx_byte()) 
+	;
+}
+
+static void uart_wait_until_sent(void)
+{
+	while(!(inb(TTYS0_BASE + UART_LSR) & 0x40))
+	; 
+}
+
+static void uart_tx_byte(unsigned char data)
+{
+	uart_wait_to_tx_byte();
+	outb(data, TTYS0_BASE + UART_TBR);
+	/* Make certain the data clears the fifos */
+	uart_wait_until_sent();
+}
+
+void console_tx_flush(void)
+{
+	uart_wait_to_tx_byte();
+}
+
+void console_tx_byte(unsigned char byte)
+{
+	if (byte == '\n')
+		uart_tx_byte('\r');
+	uart_tx_byte(byte);
+}
+
+/* ********************* smi_util ************************* */
+
+
+void io_trap_handler(int smif)
+{
+	southbridge_io_trap_handler(smif);
+}
+
+/**
+ * @brief Set the EOS bit
+ */
+static void smi_set_eos(void)
+{
+	southbridge_smi_set_eos();
+}
+
+/**
+ * @brief Interrupt handler for SMI#
+ *
+ * @param smm_revision revision of the smm state save map
+ */
+
+void smi_handler(u32 smm_revision)
+{
+	u8 reg8;
+	u16 pmctrl;
+	u16 pm1_sts;
+	u32 smi_sts, gpe0_sts, tco_sts;
+	unsigned int node;
+	smm_state_save_area_t state_save;
+
+	/* Are we ok to execute the handler? */
+	if (!smi_obtain_lock())
+		return;
+
+	node=nodeid();
+
+#ifdef DEBUG_SMI
+	console_loglevel = DEFAULT_CONSOLE_LOGLEVEL;
+#else
+	console_loglevel = 1;
+#endif
+
+	printk_debug("\nSMI# #%d\n", node);
+
+	switch (smm_revision) {
+	case 0x00030007:
+		state_save.type = LEGACY;
+		state_save.legacy_state_save = (legacy_smm_state_save_area_t *)
+			(0xa8000 + 0x7e00 - (node * 0x400));
+		break;
+	case 0x00030100:
+		state_save.type = EM64T;
+		state_save.em64t_state_save = (em64t_smm_state_save_area_t *)
+			(0xa8000 + 0x7d00 - (node * 0x400));
+		break;
+	case 0x00030064:
+		state_save.type = AMD64;
+		state_save.amd64_state_save = (amd64_smm_state_save_area_t *)
+			(0xa8000 + 0x7e00 - (node * 0x400));
+		break;
+	default:
+		printk_debug("smm_revision: 0x%08x\n", smm_revision);
+		printk_debug("SMI# not supported on your CPU\n");
+		/* Don't release lock, so no further SMI will happen,
+		 * if we don't handle it anyways.
+		 */
+		return;
+	}
+
+	southbridge_smi_handler(node, &state_save);
+
+	smi_release_lock();
+
+	/* De-assert SMI# signal to allow another SMI */
+	smi_set_eos();
+}

src/cpu/x86/smm/smm.ld

@@ -0,0 +1,53 @@
+
+/* Maximum number of CPUs/cores */
+CPUS = 4;
+
+SECTIONS
+{
+	/* This is the actual SMM handler. 
+	 *
+	 * We just put code, rodata, data and bss all in a row.
+	 */
+	. = 0xa0000;
+	.handler (.): {
+		/* Assembler stub */
+		*(.handler)
+
+		/* C code of the SMM handler */
+		*(.text);
+		*(.text.*);
+
+		/* C read-only data of the SMM handler */
+		. = ALIGN(16);
+		*(.rodata)
+		*(.rodata.*)
+
+		/* C read-write data of the SMM handler */
+		 . = ALIGN(4);
+		*(.data)
+
+		/* C uninitialized data of the SMM handler */
+		 . = ALIGN(4);
+		*(.bss)
+		*(.sbss)
+
+		/* What is this? */
+		*(COMMON)
+		 . = ALIGN(4);
+	}
+
+	/* We are using the ASEG interleaved to stuff the SMM handlers
+	 * for all CPU cores in there. The jump table redirects the execution
+	 * to the actual SMM handler
+	 */
+	. = 0xa8000 - (( CPUS - 1) * 0x400);
+	.jumptable : {
+		*(.jumptable)
+	}	
+
+	/DISCARD/ : {
+		*(.comment)
+		*(.note)
+		*(.note.*)
+	}
+}

src/cpu/x86/smm/smmhandler.S

@@ -0,0 +1,201 @@
+/*
+ * This file is part of the coreboot project.
+ *
+ * Copyright (C) 2008 coresystems GmbH
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; version 2 of
+ * the License.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston,
+ * MA 02110-1301 USA
+ */
+
+/* NOTE: This handler assumes the SMM window goes from 0xa0000
+ * to 0xaffff. In fact, at least on Intel Core CPUs (i945 chipset)
+ * the SMM window is 128K big, covering 0xa0000 to 0xbffff.
+ * So there is a lot of potential for growth in here. Let's stick
+ * to 64k if we can though.
+ */
+
+/*
+ * +--------------------------------+ 0xaffff
+ * |  Save State Map Node 0         |
+ * |  Save State Map Node 1         |
+ * |  Save State Map Node 2         |
+ * |  Save State Map Node 3         |
+ * |  ...                           |
+ * +--------------------------------+ 0xaf000
+ * |                                |
+ * |                                |
+ * |                                |
+ * +--------------------------------+ 0xa8400
+ * | SMM Entry Node 0 (+ stack)     | 
+ * +--------------------------------+ 0xa8000
+ * | SMM Entry Node 1 (+ stack)     | 
+ * | SMM Entry Node 2 (+ stack)     | 
+ * | SMM Entry Node 3 (+ stack)     | 
+ * | ...                            |
+ * +--------------------------------+ 0xa7400
+ * |                                |
+ * | SMM Handler                    |
+ * |                                |
+ * +--------------------------------+ 0xa0000
+ *
+ */
+
+#include <arch/asm.h>
+
+#define LAPIC_ID 0xfee00020
+
+/* SMM_HANDLER_OFFSET is the 16bit offset within the ASEG
+ * at which smm_handler_start lives. At the moment the handler
+ * lives right at 0xa0000, so the offset is 0. 
+ */
+
+#define SMM_HANDLER_OFFSET 0x0000
+
+/* initially SMM is some sort of real mode. Let gcc know
+ * how to treat the SMM handler stub
+ */
+
+.section ".handler", "a", @progbits
+
+.code16
+
+/**
+ * SMM code to enable protected mode and jump to the
+ * C-written function void smi_handler(u32 smm_revision)
+ *
+ * All the bad magic is not all that bad after all.
+ */
+smm_handler_start:
+	movw    $(smm_gdtptr16 - smm_handler_start + SMM_HANDLER_OFFSET), %bx
+	data32  lgdt %cs:(%bx)
+
+	movl    %cr0, %eax
+	andl    $0x7FFAFFD1, %eax /* PG,AM,WP,NE,TS,EM,MP = 0 */
+	orl     $0x60000001, %eax /* CD, NW, PE = 1 */
+	movl    %eax, %cr0
+
+	/* Enable protected mode */
+	data32  ljmp    $0x08, $1f
+
+.code32
+1:
+	/* Use flat data segment */
+	movw    $0x10, %ax
+	movw    %ax, %ds
+	movw    %ax, %es
+	movw    %ax, %ss
+	movw    %ax, %fs
+	movw    %ax, %gs
+
+	/* Get this CPU's LAPIC ID */
+	movl $LAPIC_ID, %esi
+	movl (%esi), %ecx
+	shr  $24, %ecx
+	
+	/* calculate stack offset by multiplying the APIC ID
+	 * by 1024 (0x400), and save that offset in ebp.
+	 */
+	shl $10, %ecx
+	movl %ecx, %ebp
+
+	/* We put the stack for each core right above 
+	 * its SMM entry point. Core 0 starts at 0xa8000, 
+	 * we spare 0x10 bytes for the jump to be sure.
+	 */
+	movl $0xa8010, %eax
+	subl %ecx, %eax		/* subtract offset, see above */
+	movl %eax, %ebx		/* Save bottom of stack in ebx */
+
+#define SMM_STACK_SIZE	(0x400 - 0x10)
+	/* clear stack */
+	cld
+	movl	%eax, %edi
+	movl	$(SMM_STACK_SIZE >> 2), %ecx
+	xorl	%eax, %eax
+	rep	stosl
+
+	/* set new stack */
+	addl	$SMM_STACK_SIZE, %ebx
+	movl	%ebx, %esp
+
+	/* Get SMM revision */
+	movl $0xa8000 + 0x7efc, %ebx	/* core 0 address */
+	subl %ebp, %ebx			/* subtract core X offset */
+	movl (%ebx), %eax
+	pushl %eax
+
+	/* Call 32bit C handler */
+	call smi_handler
+
+	/* To return, just do rsm. It will "clean up" protected mode */
+	rsm
+
+.code16
+
+.align	4, 0xff
+
+smm_gdtptr16:
+	.word	smm_gdt_end - smm_gdt - 1
+	.long	smm_gdt - smm_handler_start + 0xa0000 + SMM_HANDLER_OFFSET
+
+.code32
+
+smm_gdt:
+	/* The first GDT entry can not be used. Keep it zero */
+	.long	0x00000000, 0x00000000
+
+	/* gdt selector 0x08, flat code segment */
+	.word	0xffff, 0x0000		
+	.byte	0x00, 0x9b, 0xcf, 0x00 /* G=1 and 0x0f, 4GB limit */	
+
+	/* gdt selector 0x10, flat data segment */
+	.word	0xffff, 0x0000		
+	.byte	0x00, 0x93, 0xcf, 0x00
+
+smm_gdt_end:
+
+
+.section ".jumptable", "a", @progbits
+
+/* This is the SMM jump table. All cores use the same SMM handler
+ * for simplicity. But SMM Entry needs to be different due to the 
+ * save state area. The jump table makes sure all CPUs jump into the
+ * real handler on SMM entry.
+ */
+
+/* This code currently supports up to 4 CPU cores. If more than 4 CPU cores
+ * shall be used, below table has to be updated, as well as smm.ld
+ */
+
+/* GNU AS/LD will always generate code that assumes CS is 0xa000. In reality
+ * CS will be set to SMM_BASE[19:4] though. Knowing that the smm handler is the
+ * first thing in the ASEG, we do a far jump here, to set CS to 0xa000.
+ */
+
+.code16
+jumptable:
+	/* core 3 */
+	ljmp $0xa000, $SMM_HANDLER_OFFSET 
+.align 1024, 0x00
+	/* core 2 */
+	ljmp $0xa000, $SMM_HANDLER_OFFSET 
+.align 1024, 0x00
+	/* core 1 */
+	ljmp $0xa000, $SMM_HANDLER_OFFSET 
+.align 1024, 0x00
+	/* core 0 */
+	ljmp $0xa000, $SMM_HANDLER_OFFSET
+.align 1024, 0x00
+

src/cpu/x86/smm/smmrelocate.S

@@ -0,0 +1,168 @@
+/*
+ * This file is part of the coreboot project.
+ *
+ * Copyright (C) 2008 coresystems GmbH
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; version 2 of
+ * the License.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston,
+ * MA 02110-1301 USA
+ */
+
+#include <arch/asm.h>
+#include "../../../../src/northbridge/intel/i945/ich7.h"
+
+#undef DEBUG_SMM_RELOCATION
+//#define DEBUG_SMM_RELOCATION
+
+#define LAPIC_ID 0xfee00020
+
+.global smm_relocation_start
+.global smm_relocation_end
+
+/* initially SMM is some sort of real mode. */
+.code16
+
+/**
+ * This trampoline code relocates SMBASE to 0xa0000 - ( lapicid * 0x400 )
+ *
+ * Why 0x400? It is a safe value to cover the save state area per CPU. On
+ * current AMD CPUs this area is _documented_ to be 0x200 bytes. On Intel
+ * Core 2 CPUs the _documented_ parts of the save state area is 48 bytes
+ * bigger, effectively sizing our data structures 0x300 bytes.
+ *
+ * LAPICID	SMBASE		SMM Entry 	SAVE STATE
+ *    0		0xa0000		0xa8000		0xafd00
+ *    1		0x9fc00		0xa7c00		0xaf900
+ *    2		0x9f800		0xa7800		0xaf500
+ *    3		0x9f400		0xa7400		0xaf100
+ *    4		0x9f000		0xa7000		0xaed00
+ *    5		0x9ec00		0xa6c00		0xae900
+ *    6		0x9e800		0xa6800		0xae500
+ *    7		0x9e400		0xa6400		0xae100
+ *    8		0x9e000		0xa6000		0xadd00
+ *    9		0x9dc00		0xa5c00		0xad900
+ *   10		0x9d800		0xa5800		0xad500
+ *   11		0x9d400		0xa5400		0xad100
+ *   12		0x9d000		0xa5000		0xacd00
+ *   13		0x9cc00		0xa4c00		0xac900
+ *   14		0x9c800		0xa4800		0xac500
+ *   15		0x9c400		0xa4400		0xac100
+ *    .		   .		   .		   .
+ *    .		   .		   .		   .
+ *    .		   .		   .		   .
+ *   31		0x98400		0xa0400		0xa8100
+ *
+ * With 32 cores, the SMM handler would need to fit between
+ * 0xa0000-0xa0400 and the stub plus stack would need to go
+ * at 0xa8000-0xa8100 (example for core 0). That is not enough.
+ *
+ * This means we're basically limited to 16 cpu cores before 
+ * we need to use the TSEG/HSEG for the actual SMM handler plus stack.
+ * When we exceed 32 cores, we also need to put SMBASE to TSEG/HSEG.
+ *
+ * If we figure out the documented values above are safe to use,
+ * we could pack the structure above even more, so we could use the
+ * scheme to pack save state areas for 63 AMD CPUs or 58 Intel CPUs
+ * in the ASEG.
+ *
+ * Note: Some versions of Pentium M need their SMBASE aligned to 32k.
+ * On those the above only works for up to 2 cores. But for now we only
+ * care fore Core (2) Duo/Solo
+ *
+ */
+
+smm_relocation_start:
+	/* Check revision to see if AMD64 style SMM_BASE
+	 *   Intel Core Solo/Duo:  0x30007
+	 *   Intel Core2 Solo/Duo: 0x30100
+	 *   AMD64:                0x3XX64
+	 * This check does not make much sense, unless someone ports
+	 * SMI handling to AMD64 CPUs.
+	 */
+
+	mov $0x38000 + 0x7efc, %ebx
+	addr32 mov (%ebx), %al
+	cmp $0x64, %al
+	je 1f
+ 
+	mov $0x38000 + 0x7ef8, %ebx
+	jmp smm_relocate
+1:
+	mov $0x38000 + 0x7f00, %ebx
+
+smm_relocate:
+	/* Get this CPU's LAPIC ID */
+	movl $LAPIC_ID, %esi
+	addr32 movl (%esi), %ecx
+	shr  $24, %ecx
+	
+	/* calculate offset by multiplying the 
+	 * apic ID by 1024 (0x400)
+	 */
+	movl %ecx, %edx
+	shl $10, %edx
+
+	movl $0xa0000, %eax
+	subl %edx, %eax	/* subtract offset, see above */
+
+	addr32 movl %eax, (%ebx)
+
+
+	/* The next section of code is hardware specific */
+
+	/* Clear SMI status */
+	movw $(DEFAULT_PMBASE + 0x34), %dx
+	inw %dx, %ax
+	outw %ax, %dx
+
+	/* Clear PM1 status */
+	movw $(DEFAULT_PMBASE + 0x00), %dx
+	inw %dx, %ax
+	outw %ax, %dx
+
+	/* Set EOS bit so other SMIs can occur */
+	movw $(DEFAULT_PMBASE + 0x30), %dx
+	inl %dx, %eax
+	orl $(1 << 1), %eax
+	outl %eax, %dx
+
+	/* End of hardware specific section. */
+#ifdef DEBUG_SMM_RELOCATION
+	/* print [SMM-x] so we can determine if CPUx went to SMM */
+	movw $TTYS0_BASE, %dx
+	mov $'[', %al
+	outb %al, %dx
+	mov $'S', %al
+	outb %al, %dx
+	mov $'M', %al
+	outb %al, %dx
+	outb %al, %dx
+	movb $'-', %al
+	outb %al, %dx
+	/* calculate ascii of cpu number. More than 9 cores? -> FIXME */
+	movb %cl, %al
+	addb $'0', %al 
+	outb %al, %dx
+	mov $']', %al
+	outb %al, %dx
+	mov $'\r', %al
+	outb %al, %dx
+	mov $'\n', %al
+	outb %al, %dx
+#endif
+
+	/* That's it. return */
+	rsm
+smm_relocation_end:
+