Merge branch 'x86/mpparse' into x86/devel

Conflicts:

	arch/x86/Kconfig
	arch/x86/kernel/io_apic_32.c
	arch/x86/kernel/setup_64.c
	arch/x86/mm/init_32.c

Signed-off-by: Ingo Molnar <mingo@elte.hu>

3 files changed, 910 insertions, 32 deletions

diff --git a/arch/x86/kernel/cpu/mtrr/generic.c b/arch/x86/kernel/cpu/mtrr/generic.c
index 5d241ce94a4..509bd3d9eac 100644
--- a/arch/x86/kernel/cpu/mtrr/generic.c
+++ b/arch/x86/kernel/cpu/mtrr/generic.c
@@ -37,7 +37,7 @@ static struct fixed_range_block fixed_range_blocks[] = {
 static unsigned long smp_changes_mask;
 static struct mtrr_state mtrr_state = {};
 static int mtrr_state_set;
-static u64 tom2;
+u64 mtrr_tom2;
 
 #undef MODULE_PARAM_PREFIX
 #define MODULE_PARAM_PREFIX "mtrr."
@@ -139,8 +139,8 @@ u8 mtrr_type_lookup(u64 start, u64 end)
 		}
 	}
 
-	if (tom2) {
-		if (start >= (1ULL<<32) && (end < tom2))
+	if (mtrr_tom2) {
+		if (start >= (1ULL<<32) && (end < mtrr_tom2))
 			return MTRR_TYPE_WRBACK;
 	}
 
@@ -158,6 +158,20 @@ get_mtrr_var_range(unsigned int index, struct mtrr_var_range *vr)
 	rdmsr(MTRRphysMask_MSR(index), vr->mask_lo, vr->mask_hi);
 }
 
+/*  fill the MSR pair relating to a var range  */
+void fill_mtrr_var_range(unsigned int index,
+		u32 base_lo, u32 base_hi, u32 mask_lo, u32 mask_hi)
+{
+	struct mtrr_var_range *vr;
+
+	vr = mtrr_state.var_ranges;
+
+	vr[index].base_lo = base_lo;
+	vr[index].base_hi = base_hi;
+	vr[index].mask_lo = mask_lo;
+	vr[index].mask_hi = mask_hi;
+}
+
 static void
 get_fixed_ranges(mtrr_type * frs)
 {
@@ -213,13 +227,13 @@ void __init get_mtrr_state(void)
 	mtrr_state.enabled = (lo & 0xc00) >> 10;
 
 	if (amd_special_default_mtrr()) {
-		unsigned lo, hi;
+		unsigned low, high;
 		/* TOP_MEM2 */
-		rdmsr(MSR_K8_TOP_MEM2, lo, hi);
-		tom2 = hi;
-		tom2 <<= 32;
-		tom2 |= lo;
-		tom2 &= 0xffffff8000000ULL;
+		rdmsr(MSR_K8_TOP_MEM2, low, high);
+		mtrr_tom2 = high;
+		mtrr_tom2 <<= 32;
+		mtrr_tom2 |= low;
+		mtrr_tom2 &= 0xffffff800000ULL;
 	}
 	if (mtrr_show) {
 		int high_width;
@@ -251,9 +265,9 @@ void __init get_mtrr_state(void)
 			else
 				printk(KERN_INFO "MTRR %u disabled\n", i);
 		}
-		if (tom2) {
+		if (mtrr_tom2) {
 			printk(KERN_INFO "TOM2: %016llx aka %lldM\n",
-					  tom2, tom2>>20);
+					  mtrr_tom2, mtrr_tom2>>20);
 		}
 	}
 	mtrr_state_set = 1;
@@ -328,7 +342,7 @@ static void set_fixed_range(int msr, bool *changed, unsigned int *msrwords)
 
 	if (lo != msrwords[0] || hi != msrwords[1]) {
 		if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
-		    boot_cpu_data.x86 == 15 &&
+		    (boot_cpu_data.x86 >= 0x0f && boot_cpu_data.x86 <= 0x11) &&
 		    ((msrwords[0] | msrwords[1]) & K8_MTRR_RDMEM_WRMEM_MASK))
 			k8_enable_fixed_iorrs();
 		mtrr_wrmsr(msr, msrwords[0], msrwords[1]);
diff --git a/arch/x86/kernel/cpu/mtrr/main.c b/arch/x86/kernel/cpu/mtrr/main.c
index 6a1e278d932..105afe12beb 100644
--- a/arch/x86/kernel/cpu/mtrr/main.c
+++ b/arch/x86/kernel/cpu/mtrr/main.c
@@ -37,6 +37,7 @@
 #include <linux/smp.h>
 #include <linux/cpu.h>
 #include <linux/mutex.h>
+#include <linux/sort.h>
 
 #include <asm/e820.h>
 #include <asm/mtrr.h>
@@ -609,6 +610,787 @@ static struct sysdev_driver mtrr_sysdev_driver = {
 	.resume		= mtrr_restore,
 };
 
+/* should be related to MTRR_VAR_RANGES nums */
+#define RANGE_NUM 256
+
+struct res_range {
+	unsigned long start;
+	unsigned long end;
+};
+
+static int __init
+add_range(struct res_range *range, int nr_range, unsigned long start,
+			      unsigned long end)
+{
+	/* out of slots */
+	if (nr_range >= RANGE_NUM)
+		return nr_range;
+
+	range[nr_range].start = start;
+	range[nr_range].end = end;
+
+	nr_range++;
+
+	return nr_range;
+}
+
+static int __init
+add_range_with_merge(struct res_range *range, int nr_range, unsigned long start,
+			      unsigned long end)
+{
+	int i;
+
+	/* try to merge it with old one */
+	for (i = 0; i < nr_range; i++) {
+		unsigned long final_start, final_end;
+		unsigned long common_start, common_end;
+
+		if (!range[i].end)
+			continue;
+
+		common_start = max(range[i].start, start);
+		common_end = min(range[i].end, end);
+		if (common_start > common_end + 1)
+			continue;
+
+		final_start = min(range[i].start, start);
+		final_end = max(range[i].end, end);
+
+		range[i].start = final_start;
+		range[i].end =  final_end;
+		return nr_range;
+	}
+
+	/* need to add that */
+	return add_range(range, nr_range, start, end);
+}
+
+static void __init
+subtract_range(struct res_range *range, unsigned long start, unsigned long end)
+{
+	int i, j;
+
+	for (j = 0; j < RANGE_NUM; j++) {
+		if (!range[j].end)
+			continue;
+
+		if (start <= range[j].start && end >= range[j].end) {
+			range[j].start = 0;
+			range[j].end = 0;
+			continue;
+		}
+
+		if (start <= range[j].start && end < range[j].end &&
+		    range[j].start < end + 1) {
+			range[j].start = end + 1;
+			continue;
+		}
+
+
+		if (start > range[j].start && end >= range[j].end &&
+		    range[j].end > start - 1) {
+			range[j].end = start - 1;
+			continue;
+		}
+
+		if (start > range[j].start && end < range[j].end) {
+			/* find the new spare */
+			for (i = 0; i < RANGE_NUM; i++) {
+				if (range[i].end == 0)
+					break;
+			}
+			if (i < RANGE_NUM) {
+				range[i].end = range[j].end;
+				range[i].start = end + 1;
+			} else {
+				printk(KERN_ERR "run of slot in ranges\n");
+			}
+			range[j].end = start - 1;
+			continue;
+		}
+	}
+}
+
+static int __init cmp_range(const void *x1, const void *x2)
+{
+	const struct res_range *r1 = x1;
+	const struct res_range *r2 = x2;
+	long start1, start2;
+
+	start1 = r1->start;
+	start2 = r2->start;
+
+	return start1 - start2;
+}
+
+struct var_mtrr_range_state {
+	unsigned long base_pfn;
+	unsigned long size_pfn;
+	mtrr_type type;
+};
+
+struct var_mtrr_range_state __initdata range_state[RANGE_NUM];
+static int __initdata debug_print;
+
+static int __init
+x86_get_mtrr_mem_range(struct res_range *range, int nr_range,
+		       unsigned long extra_remove_base,
+		       unsigned long extra_remove_size)
+{
+	unsigned long i, base, size;
+	mtrr_type type;
+
+	for (i = 0; i < num_var_ranges; i++) {
+		type = range_state[i].type;
+		if (type != MTRR_TYPE_WRBACK)
+			continue;
+		base = range_state[i].base_pfn;
+		size = range_state[i].size_pfn;
+		nr_range = add_range_with_merge(range, nr_range, base,
+						base + size - 1);
+	}
+	if (debug_print) {
+		printk(KERN_DEBUG "After WB checking\n");
+		for (i = 0; i < nr_range; i++)
+			printk(KERN_DEBUG "MTRR MAP PFN: %016lx - %016lx\n",
+				 range[i].start, range[i].end + 1);
+	}
+
+	/* take out UC ranges */
+	for (i = 0; i < num_var_ranges; i++) {
+		type = range_state[i].type;
+		if (type != MTRR_TYPE_UNCACHABLE)
+			continue;
+		size = range_state[i].size_pfn;
+		if (!size)
+			continue;
+		base = range_state[i].base_pfn;
+		subtract_range(range, base, base + size - 1);
+	}
+	if (extra_remove_size)
+		subtract_range(range, extra_remove_base,
+				 extra_remove_base + extra_remove_size  - 1);
+
+	/* get new range num */
+	nr_range = 0;
+	for (i = 0; i < RANGE_NUM; i++) {
+		if (!range[i].end)
+			continue;
+		nr_range++;
+	}
+	if  (debug_print) {
+		printk(KERN_DEBUG "After UC checking\n");
+		for (i = 0; i < nr_range; i++)
+			printk(KERN_DEBUG "MTRR MAP PFN: %016lx - %016lx\n",
+				 range[i].start, range[i].end + 1);
+	}
+
+	/* sort the ranges */
+	sort(range, nr_range, sizeof(struct res_range), cmp_range, NULL);
+	if  (debug_print) {
+		printk(KERN_DEBUG "After sorting\n");
+		for (i = 0; i < nr_range; i++)
+			printk(KERN_DEBUG "MTRR MAP PFN: %016lx - %016lx\n",
+				 range[i].start, range[i].end + 1);
+	}
+
+	/* clear those is not used */
+	for (i = nr_range; i < RANGE_NUM; i++)
+		memset(&range[i], 0, sizeof(range[i]));
+
+	return nr_range;
+}
+
+static struct res_range __initdata range[RANGE_NUM];
+
+#ifdef CONFIG_MTRR_SANITIZER
+
+static unsigned long __init sum_ranges(struct res_range *range, int nr_range)
+{
+	unsigned long sum;
+	int i;
+
+	sum = 0;
+	for (i = 0; i < nr_range; i++)
+		sum += range[i].end + 1 - range[i].start;
+
+	return sum;
+}
+
+static int enable_mtrr_cleanup __initdata =
+	CONFIG_MTRR_SANITIZER_ENABLE_DEFAULT;
+
+static int __init disable_mtrr_cleanup_setup(char *str)
+{
+	if (enable_mtrr_cleanup != -1)
+		enable_mtrr_cleanup = 0;
+	return 0;
+}
+early_param("disable_mtrr_cleanup", disable_mtrr_cleanup_setup);
+
+static int __init enable_mtrr_cleanup_setup(char *str)
+{
+	if (enable_mtrr_cleanup != -1)
+		enable_mtrr_cleanup = 1;
+	return 0;
+}
+early_param("enble_mtrr_cleanup", enable_mtrr_cleanup_setup);
+
+struct var_mtrr_state {
+	unsigned long	range_startk;
+	unsigned long	range_sizek;
+	unsigned long	chunk_sizek;
+	unsigned long	gran_sizek;
+	unsigned int	reg;
+};
+
+static void __init
+set_var_mtrr(unsigned int reg, unsigned long basek, unsigned long sizek,
+		unsigned char type, unsigned int address_bits)
+{
+	u32 base_lo, base_hi, mask_lo, mask_hi;
+	u64 base, mask;
+
+	if (!sizek) {
+		fill_mtrr_var_range(reg, 0, 0, 0, 0);
+		return;
+	}
+
+	mask = (1ULL << address_bits) - 1;
+	mask &= ~((((u64)sizek) << 10) - 1);
+
+	base  = ((u64)basek) << 10;
+
+	base |= type;
+	mask |= 0x800;
+
+	base_lo = base & ((1ULL<<32) - 1);
+	base_hi = base >> 32;
+
+	mask_lo = mask & ((1ULL<<32) - 1);
+	mask_hi = mask >> 32;
+
+	fill_mtrr_var_range(reg, base_lo, base_hi, mask_lo, mask_hi);
+}
+
+static void __init
+save_var_mtrr(unsigned int reg, unsigned long basek, unsigned long sizek,
+		unsigned char type)
+{
+	range_state[reg].base_pfn = basek >> (PAGE_SHIFT - 10);
+	range_state[reg].size_pfn = sizek >> (PAGE_SHIFT - 10);
+	range_state[reg].type = type;
+}
+
+static void __init
+set_var_mtrr_all(unsigned int address_bits)
+{
+	unsigned long basek, sizek;
+	unsigned char type;
+	unsigned int reg;
+
+	for (reg = 0; reg < num_var_ranges; reg++) {
+		basek = range_state[reg].base_pfn << (PAGE_SHIFT - 10);
+		sizek = range_state[reg].size_pfn << (PAGE_SHIFT - 10);
+		type = range_state[reg].type;
+
+		set_var_mtrr(reg, basek, sizek, type, address_bits);
+	}
+}
+
+static unsigned int __init
+range_to_mtrr(unsigned int reg, unsigned long range_startk,
+	      unsigned long range_sizek, unsigned char type)
+{
+	if (!range_sizek || (reg >= num_var_ranges))
+		return reg;
+
+	while (range_sizek) {
+		unsigned long max_align, align;
+		unsigned long sizek;
+
+		/* Compute the maximum size I can make a range */
+		if (range_startk)
+			max_align = ffs(range_startk) - 1;
+		else
+			max_align = 32;
+		align = fls(range_sizek) - 1;
+		if (align > max_align)
+			align = max_align;
+
+		sizek = 1 << align;
+		if (debug_print)
+			printk(KERN_DEBUG "Setting variable MTRR %d, "
+				"base: %ldMB, range: %ldMB, type %s\n",
+				reg, range_startk >> 10, sizek >> 10,
+				(type == MTRR_TYPE_UNCACHABLE)?"UC":
+				    ((type == MTRR_TYPE_WRBACK)?"WB":"Other")
+				);
+		save_var_mtrr(reg++, range_startk, sizek, type);
+		range_startk += sizek;
+		range_sizek -= sizek;
+		if (reg >= num_var_ranges)
+			break;
+	}
+	return reg;
+}
+
+static unsigned __init
+range_to_mtrr_with_hole(struct var_mtrr_state *state, unsigned long basek,
+			unsigned long sizek)
+{
+	unsigned long hole_basek, hole_sizek;
+	unsigned long second_basek, second_sizek;
+	unsigned long range0_basek, range0_sizek;
+	unsigned long range_basek, range_sizek;
+	unsigned long chunk_sizek;
+	unsigned long gran_sizek;
+
+	hole_basek = 0;
+	hole_sizek = 0;
+	second_basek = 0;
+	second_sizek = 0;
+	chunk_sizek = state->chunk_sizek;
+	gran_sizek = state->gran_sizek;
+
+	/* align with gran size, prevent small block used up MTRRs */
+	range_basek = ALIGN(state->range_startk, gran_sizek);
+	if ((range_basek > basek) && basek)
+		return second_sizek;
+	state->range_sizek -= (range_basek - state->range_startk);
+	range_sizek = ALIGN(state->range_sizek, gran_sizek);
+
+	while (range_sizek > state->range_sizek) {
+		range_sizek -= gran_sizek;
+		if (!range_sizek)
+			return 0;
+	}
+	state->range_sizek = range_sizek;
+
+	/* try to append some small hole */
+	range0_basek = state->range_startk;
+	range0_sizek = ALIGN(state->range_sizek, chunk_sizek);
+	if (range0_sizek == state->range_sizek) {
+		if (debug_print)
+			printk(KERN_DEBUG "rangeX: %016lx - %016lx\n",
+				range0_basek<<10,
+				(range0_basek + state->range_sizek)<<10);
+		state->reg = range_to_mtrr(state->reg, range0_basek,
+				state->range_sizek, MTRR_TYPE_WRBACK);
+		return 0;
+	}
+
+	range0_sizek -= chunk_sizek;
+	if (range0_sizek && sizek) {
+	    while (range0_basek + range0_sizek > (basek + sizek)) {
+		range0_sizek -= chunk_sizek;
+		if (!range0_sizek)
+			break;
+	    }
+	}
+
+	if (range0_sizek) {
+		if (debug_print)
+			printk(KERN_DEBUG "range0: %016lx - %016lx\n",
+				range0_basek<<10,
+				(range0_basek + range0_sizek)<<10);
+		state->reg = range_to_mtrr(state->reg, range0_basek,
+				range0_sizek, MTRR_TYPE_WRBACK);
+
+	}
+
+	range_basek = range0_basek + range0_sizek;
+	range_sizek = chunk_sizek;
+
+	if (range_basek + range_sizek > basek &&
+	    range_basek + range_sizek <= (basek + sizek)) {
+		/* one hole */
+		second_basek = basek;
+		second_sizek = range_basek + range_sizek - basek;
+	}
+
+	/* if last piece, only could one hole near end */
+	if ((second_basek || !basek) &&
+	    range_sizek - (state->range_sizek - range0_sizek) - second_sizek <
+	    (chunk_sizek >> 1)) {
+		/*
+		 * one hole in middle (second_sizek is 0) or at end
+		 * (second_sizek is 0 )
+		 */
+		hole_sizek = range_sizek - (state->range_sizek - range0_sizek)
+				 - second_sizek;
+		hole_basek = range_basek + range_sizek - hole_sizek
+				 - second_sizek;
+	} else {
+		/* fallback for big hole, or several holes */
+		range_sizek = state->range_sizek - range0_sizek;
+		second_basek = 0;
+		second_sizek = 0;
+	}
+
+	if (debug_print)
+		printk(KERN_DEBUG "range: %016lx - %016lx\n", range_basek<<10,
+			 (range_basek + range_sizek)<<10);
+	state->reg = range_to_mtrr(state->reg, range_basek, range_sizek,
+					 MTRR_TYPE_WRBACK);
+	if (hole_sizek) {
+		if (debug_print)
+			printk(KERN_DEBUG "hole: %016lx - %016lx\n",
+				 hole_basek<<10, (hole_basek + hole_sizek)<<10);
+		state->reg = range_to_mtrr(state->reg, hole_basek, hole_sizek,
+						 MTRR_TYPE_UNCACHABLE);
+
+	}
+
+	return second_sizek;
+}
+
+static void __init
+set_var_mtrr_range(struct var_mtrr_state *state, unsigned long base_pfn,
+		   unsigned long size_pfn)
+{
+	unsigned long basek, sizek;
+	unsigned long second_sizek = 0;
+
+	if (state->reg >= num_var_ranges)
+		return;
+
+	basek = base_pfn << (PAGE_SHIFT - 10);
+	sizek = size_pfn << (PAGE_SHIFT - 10);
+
+	/* See if I can merge with the last range */
+	if ((basek <= 1024) ||
+	    (state->range_startk + state->range_sizek == basek)) {
+		unsigned long endk = basek + sizek;
+		state->range_sizek = endk - state->range_startk;
+		return;
+	}
+	/* Write the range mtrrs */
+	if (state->range_sizek != 0)
+		second_sizek = range_to_mtrr_with_hole(state, basek, sizek);
+
+	/* Allocate an msr */
+	state->range_startk = basek + second_sizek;
+	state->range_sizek  = sizek - second_sizek;
+}
+
+/* mininum size of mtrr block that can take hole */
+static u64 mtrr_chunk_size __initdata = (256ULL<<20);
+
+static int __init parse_mtrr_chunk_size_opt(char *p)
+{
+	if (!p)
+		return -EINVAL;
+	mtrr_chunk_size = memparse(p, &p);
+	return 0;
+}
+early_param("mtrr_chunk_size", parse_mtrr_chunk_size_opt);
+
+/* granity of mtrr of block */
+static u64 mtrr_gran_size __initdata;
+
+static int __init parse_mtrr_gran_size_opt(char *p)
+{
+	if (!p)
+		return -EINVAL;
+	mtrr_gran_size = memparse(p, &p);
+	return 0;
+}
+early_param("mtrr_gran_size", parse_mtrr_gran_size_opt);
+
+static int nr_mtrr_spare_reg __initdata =
+				 CONFIG_MTRR_SANITIZER_SPARE_REG_NR_DEFAULT;
+
+static int __init parse_mtrr_spare_reg(char *arg)
+{
+	if (arg)
+		nr_mtrr_spare_reg = simple_strtoul(arg, NULL, 0);
+	return 0;
+}
+
+early_param("mtrr_spare_reg_nr", parse_mtrr_spare_reg);
+
+static int __init
+x86_setup_var_mtrrs(struct res_range *range, int nr_range,
+		    u64 chunk_size, u64 gran_size)
+{
+	struct var_mtrr_state var_state;
+	int i;
+	int num_reg;
+
+	var_state.range_startk	= 0;
+	var_state.range_sizek	= 0;
+	var_state.reg		= 0;
+	var_state.chunk_sizek	= chunk_size >> 10;
+	var_state.gran_sizek	= gran_size >> 10;
+
+	memset(range_state, 0, sizeof(range_state));
+
+	/* Write the range etc */
+	for (i = 0; i < nr_range; i++)
+		set_var_mtrr_range(&var_state, range[i].start,
+				   range[i].end - range[i].start + 1);
+
+	/* Write the last range */
+	if (var_state.range_sizek != 0)
+		range_to_mtrr_with_hole(&var_state, 0, 0);
+
+	num_reg = var_state.reg;
+	/* Clear out the extra MTRR's */
+	while (var_state.reg < num_var_ranges) {
+		save_var_mtrr(var_state.reg, 0, 0, 0);
+		var_state.reg++;
+	}
+
+	return num_reg;
+}
+
+struct mtrr_cleanup_result {
+	unsigned long gran_sizek;
+	unsigned long chunk_sizek;
+	unsigned long lose_cover_sizek;
+	unsigned int num_reg;
+	int bad;
+};
+
+/*
+ * gran_size: 1M, 2M, ..., 2G
+ * chunk size: gran_size, ..., 4G
+ * so we need (2+13)*6
+ */
+#define NUM_RESULT	90
+#define PSHIFT		(PAGE_SHIFT - 10)
+
+static struct mtrr_cleanup_result __initdata result[NUM_RESULT];
+static struct res_range __initdata range_new[RANGE_NUM];
+static unsigned long __initdata min_loss_pfn[RANGE_NUM];
+
+static int __init mtrr_cleanup(unsigned address_bits)
+{
+	unsigned long extra_remove_base, extra_remove_size;
+	unsigned long i, base, size, def, dummy;
+	mtrr_type type;
+	int nr_range, nr_range_new;
+	u64 chunk_size, gran_size;
+	unsigned long range_sums, range_sums_new;
+	int index_good;
+	int num_reg_good;
+
+	/* extra one for all 0 */
+	int num[MTRR_NUM_TYPES + 1];
+
+	if (!is_cpu(INTEL) || enable_mtrr_cleanup < 1)
+		return 0;
+	rdmsr(MTRRdefType_MSR, def, dummy);
+	def &= 0xff;
+	if (def != MTRR_TYPE_UNCACHABLE)
+		return 0;
+
+	/* get it and store it aside */
+	memset(range_state, 0, sizeof(range_state));
+	for (i = 0; i < num_var_ranges; i++) {
+		mtrr_if->get(i, &base, &size, &type);
+		range_state[i].base_pfn = base;
+		range_state[i].size_pfn = size;
+		range_state[i].type = type;
+	}
+
+	/* check entries number */
+	memset(num, 0, sizeof(num));
+	for (i = 0; i < num_var_ranges; i++) {
+		type = range_state[i].type;
+		size = range_state[i].size_pfn;
+		if (type >= MTRR_NUM_TYPES)
+			continue;
+		if (!size)
+			type = MTRR_NUM_TYPES;
+		num[type]++;
+	}
+
+	/* check if we got UC entries */
+	if (!num[MTRR_TYPE_UNCACHABLE])
+		return 0;
+
+	/* check if we only had WB and UC */
+	if (num[MTRR_TYPE_WRBACK] + num[MTRR_TYPE_UNCACHABLE] !=
+		num_var_ranges - num[MTRR_NUM_TYPES])
+		return 0;
+
+	memset(range, 0, sizeof(range));
+	extra_remove_size = 0;
+	if (mtrr_tom2) {
+		extra_remove_base = 1 << (32 - PAGE_SHIFT);
+		extra_remove_size =
+			(mtrr_tom2 >> PAGE_SHIFT) - extra_remove_base;
+	}
+	nr_range = x86_get_mtrr_mem_range(range, 0, extra_remove_base,
+					  extra_remove_size);
+	range_sums = sum_ranges(range, nr_range);
+	printk(KERN_INFO "total RAM coverred: %ldM\n",
+	       range_sums >> (20 - PAGE_SHIFT));
+
+	if (mtrr_chunk_size && mtrr_gran_size) {
+		int num_reg;
+
+		debug_print = 1;
+		/* convert ranges to var ranges state */
+		num_reg = x86_setup_var_mtrrs(range, nr_range, mtrr_chunk_size,
+					      mtrr_gran_size);
+
+		/* we got new setting in range_state, check it */
+		memset(range_new, 0, sizeof(range_new));
+		nr_range_new = x86_get_mtrr_mem_range(range_new, 0,
+						      extra_remove_base,
+						      extra_remove_size);
+		range_sums_new = sum_ranges(range_new, nr_range_new);
+
+		i = 0;
+		result[i].chunk_sizek = mtrr_chunk_size >> 10;
+		result[i].gran_sizek = mtrr_gran_size >> 10;
+		result[i].num_reg = num_reg;
+		if (range_sums < range_sums_new) {
+			result[i].lose_cover_sizek =
+				(range_sums_new - range_sums) << PSHIFT;
+			result[i].bad = 1;
+		} else
+			result[i].lose_cover_sizek =
+				(range_sums - range_sums_new) << PSHIFT;
+
+		printk(KERN_INFO "%sgran_size: %ldM \tchunk_size: %ldM \t",
+			 result[i].bad?"*BAD*":" ", result[i].gran_sizek >> 10,
+			 result[i].chunk_sizek >> 10);
+		printk(KERN_CONT "num_reg: %d  \tlose cover RAM: %s%ldM \n",
+			 result[i].num_reg, result[i].bad?"-":"",
+			 result[i].lose_cover_sizek >> 10);
+		if (!result[i].bad) {
+			set_var_mtrr_all(address_bits);
+			return 1;
+		}
+		printk(KERN_INFO "invalid mtrr_gran_size or mtrr_chunk_size, "
+		       "will find optimal one\n");
+		debug_print = 0;
+		memset(result, 0, sizeof(result[0]));
+	}
+
+	i = 0;
+	memset(min_loss_pfn, 0xff, sizeof(min_loss_pfn));
+	memset(result, 0, sizeof(result));
+	for (gran_size = (1ULL<<20); gran_size < (1ULL<<32); gran_size <<= 1) {
+		for (chunk_size = gran_size; chunk_size < (1ULL<<33);
+		     chunk_size <<= 1) {
+			int num_reg;
+
+			if (debug_print)
+				printk(KERN_INFO
+			       "\ngran_size: %lldM   chunk_size_size: %lldM\n",
+				       gran_size >> 20, chunk_size >> 20);
+			if (i >= NUM_RESULT)
+				continue;
+
+			/* convert ranges to var ranges state */
+			num_reg = x86_setup_var_mtrrs(range, nr_range,
+							 chunk_size, gran_size);
+
+			/* we got new setting in range_state, check it */
+			memset(range_new, 0, sizeof(range_new));
+			nr_range_new = x86_get_mtrr_mem_range(range_new, 0,
+					 extra_remove_base, extra_remove_size);
+			range_sums_new = sum_ranges(range_new, nr_range_new);
+
+			result[i].chunk_sizek = chunk_size >> 10;
+			result[i].gran_sizek = gran_size >> 10;
+			result[i].num_reg = num_reg;
+			if (range_sums < range_sums_new) {
+				result[i].lose_cover_sizek =
+					(range_sums_new - range_sums) << PSHIFT;
+				result[i].bad = 1;
+			} else
+				result[i].lose_cover_sizek =
+					(range_sums - range_sums_new) << PSHIFT;
+
+			/* double check it */
+			if (!result[i].bad && !result[i].lose_cover_sizek) {
+				if (nr_range_new != nr_range ||
+					memcmp(range, range_new, sizeof(range)))
+						result[i].bad = 1;
+			}
+
+			if (!result[i].bad && (range_sums - range_sums_new <
+					       min_loss_pfn[num_reg])) {
+				min_loss_pfn[num_reg] =
+					range_sums - range_sums_new;
+			}
+			i++;
+		}
+	}
+
+	/* print out all */
+	for (i = 0; i < NUM_RESULT; i++) {
+		printk(KERN_INFO "%sgran_size: %ldM \tchunk_size: %ldM \t",
+		       result[i].bad?"*BAD* ":" ", result[i].gran_sizek >> 10,
+		       result[i].chunk_sizek >> 10);
+		printk(KERN_CONT "num_reg: %d \tlose RAM: %s%ldM\n",
+		       result[i].num_reg, result[i].bad?"-":"",
+		       result[i].lose_cover_sizek >> 10);
+	}
+
+	/* try to find the optimal index */
+	if (nr_mtrr_spare_reg >= num_var_ranges)
+		nr_mtrr_spare_reg = num_var_ranges - 1;
+	num_reg_good = -1;
+	for (i = num_var_ranges - nr_mtrr_spare_reg; i > 0; i--) {
+		if (!min_loss_pfn[i]) {
+			num_reg_good = i;
+			break;
+		}
+	}
+
+	index_good = -1;
+	if (num_reg_good != -1) {
+		for (i = 0; i < NUM_RESULT; i++) {
+			if (!result[i].bad &&
+			    result[i].num_reg == num_reg_good &&
+			    !result[i].lose_cover_sizek) {
+				index_good = i;
+				break;
+			}
+		}
+	}
+
+	if (index_good != -1) {
+		printk(KERN_INFO "Found optimal setting for mtrr clean up\n");
+		i = index_good;
+		printk(KERN_INFO "gran_size: %ldM \tchunk_size: %ldM \t",
+				result[i].gran_sizek >> 10,
+				result[i].chunk_sizek >> 10);
+		printk(KERN_CONT "num_reg: %d \tlose RAM: %ldM\n",
+				result[i].num_reg,
+				result[i].lose_cover_sizek >> 10);
+		/* convert ranges to var ranges state */
+		chunk_size = result[i].chunk_sizek;
+		chunk_size <<= 10;
+		gran_size = result[i].gran_sizek;
+		gran_size <<= 10;
+		debug_print = 1;
+		x86_setup_var_mtrrs(range, nr_range, chunk_size, gran_size);
+		set_var_mtrr_all(address_bits);
+		return 1;
+	}
+
+	printk(KERN_INFO "mtrr_cleanup: can not find optimal value\n");
+	printk(KERN_INFO "please specify mtrr_gran_size/mtrr_chunk_size\n");
+
+	return 0;
+}
+#else
+static int __init mtrr_cleanup(unsigned address_bits)
+{
+	return 0;
+}
+#endif
+
+static int __initdata changed_by_mtrr_cleanup;
+
 static int disable_mtrr_trim;
 
 static int __init disable_mtrr_trim_setup(char *str)
@@ -648,6 +1430,19 @@ int __init amd_special_default_mtrr(void)
 	return 0;
 }
 
+static u64 __init real_trim_memory(unsigned long start_pfn,
+				   unsigned long limit_pfn)
+{
+	u64 trim_start, trim_size;
+	trim_start = start_pfn;
+	trim_start <<= PAGE_SHIFT;
+	trim_size = limit_pfn;
+	trim_size <<= PAGE_SHIFT;
+	trim_size -= trim_start;
+
+	return e820_update_range(trim_start, trim_size, E820_RAM,
+				E820_RESERVED);
+}
 /**
  * mtrr_trim_uncached_memory - trim RAM not covered by MTRRs
  * @end_pfn: ending page frame number
@@ -663,8 +1458,11 @@ int __init mtrr_trim_uncached_memory(unsigned long end_pfn)
 {
 	unsigned long i, base, size, highest_pfn = 0, def, dummy;
 	mtrr_type type;
-	u64 trim_start, trim_size;
+	int nr_range;
+	u64 total_trim_size;
 
+	/* extra one for all 0 */
+	int num[MTRR_NUM_TYPES + 1];
 	/*
 	 * Make sure we only trim uncachable memory on machines that
 	 * support the Intel MTRR architecture:
@@ -676,14 +1474,22 @@ int __init mtrr_trim_uncached_memory(unsigned long end_pfn)
 	if (def != MTRR_TYPE_UNCACHABLE)
 		return 0;
 
-	if (amd_special_default_mtrr())
-		return 0;
+	/* get it and store it aside */
+	memset(range_state, 0, sizeof(range_state));
+	for (i = 0; i < num_var_ranges; i++) {
+		mtrr_if->get(i, &base, &size, &type);
+		range_state[i].base_pfn = base;
+		range_state[i].size_pfn = size;
+		range_state[i].type = type;
+	}
 
 	/* Find highest cached pfn */
 	for (i = 0; i < num_var_ranges; i++) {
-		mtrr_if->get(i, &base, &size, &type);
+		type = range_state[i].type;
 		if (type != MTRR_TYPE_WRBACK)
 			continue;
+		base = range_state[i].base_pfn;
+		size = range_state[i].size_pfn;
 		if (highest_pfn < base + size)
 			highest_pfn = base + size;
 	}
@@ -698,22 +1504,65 @@ int __init mtrr_trim_uncached_memory(unsigned long end_pfn)
 		return 0;
 	}
 
-	if (highest_pfn < end_pfn) {
+	/* check entries number */
+	memset(num, 0, sizeof(num));
+	for (i = 0; i < num_var_ranges; i++) {
+		type = range_state[i].type;
+		if (type >= MTRR_NUM_TYPES)
+			continue;
+		size = range_state[i].size_pfn;
+		if (!size)
+			type = MTRR_NUM_TYPES;
+		num[type]++;
+	}
+
+	/* no entry for WB? */
+	if (!num[MTRR_TYPE_WRBACK])
+		return 0;
+
+	/* check if we only had WB and UC */
+	if (num[MTRR_TYPE_WRBACK] + num[MTRR_TYPE_UNCACHABLE] !=
+		num_var_ranges - num[MTRR_NUM_TYPES])
+		return 0;
+
+	memset(range, 0, sizeof(range));
+	nr_range = 0;
+	if (mtrr_tom2) {
+		range[nr_range].start = (1ULL<<(32 - PAGE_SHIFT));
+		range[nr_range].end = (mtrr_tom2 >> PAGE_SHIFT) - 1;
+		if (highest_pfn < range[nr_range].end + 1)
+			highest_pfn = range[nr_range].end + 1;
+		nr_range++;
+	}
+	nr_range = x86_get_mtrr_mem_range(range, nr_range, 0, 0);
+
+	total_trim_size = 0;
+	/* check the head */
+	if (range[0].start)
+		total_trim_size += real_trim_memory(0, range[0].start);
+	/* check the holes */
+	for (i = 0; i < nr_range - 1; i++) {
+		if (range[i].end + 1 < range[i+1].start)
+			total_trim_size += real_trim_memory(range[i].end + 1,
+							    range[i+1].start);
+	}
+	/* check the top */
+	i = nr_range - 1;
+	if (range[i].end + 1 < end_pfn)
+		total_trim_size += real_trim_memory(range[i].end + 1,
+							 end_pfn);
+
+	if (total_trim_size) {
 		printk(KERN_WARNING "WARNING: BIOS bug: CPU MTRRs don't cover"
-			" all of memory, losing %luMB of RAM.\n",
-			(end_pfn - highest_pfn) >> (20 - PAGE_SHIFT));
+			" all of memory, losing %lluMB of RAM.\n",
+			total_trim_size >> 20);
 
-		WARN_ON(1);
+		if (!changed_by_mtrr_cleanup)
+			WARN_ON(1);
 
 		printk(KERN_INFO "update e820 for mtrr\n");
-		trim_start = highest_pfn;
-		trim_start <<= PAGE_SHIFT;
-		trim_size = end_pfn;
-		trim_size <<= PAGE_SHIFT;
-		trim_size -= trim_start;
-		update_memory_range(trim_start, trim_size, E820_RAM,
-					E820_RESERVED);
 		update_e820();
+
 		return 1;
 	}
 
@@ -729,18 +1578,21 @@ int __init mtrr_trim_uncached_memory(unsigned long end_pfn)
  */
 void __init mtrr_bp_init(void)
 {
+	u32 phys_addr;
 	init_ifs();
 
+	phys_addr = 32;
+
 	if (cpu_has_mtrr) {
 		mtrr_if = &generic_mtrr_ops;
 		size_or_mask = 0xff000000;	/* 36 bits */
 		size_and_mask = 0x00f00000;
+		phys_addr = 36;
 
 		/* This is an AMD specific MSR, but we assume(hope?) that
 		   Intel will implement it to when they extend the address
 		   bus of the Xeon. */
 		if (cpuid_eax(0x80000000) >= 0x80000008) {
-			u32 phys_addr;
 			phys_addr = cpuid_eax(0x80000008) & 0xff;
 			/* CPUID workaround for Intel 0F33/0F34 CPU */
 			if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL &&
@@ -758,6 +1610,7 @@ void __init mtrr_bp_init(void)
 			   don't support PAE */
 			size_or_mask = 0xfff00000;	/* 32 bits */
 			size_and_mask = 0;
+			phys_addr = 32;
 		}
 	} else {
 		switch (boot_cpu_data.x86_vendor) {
@@ -791,8 +1644,15 @@ void __init mtrr_bp_init(void)
 	if (mtrr_if) {
 		set_num_var_ranges();
 		init_table();
-		if (use_intel())
+		if (use_intel()) {
 			get_mtrr_state();
+
+			if (mtrr_cleanup(phys_addr)) {
+				changed_by_mtrr_cleanup = 1;
+				mtrr_if->set_all();
+			}
+
+		}
 	}
 }
 
@@ -829,9 +1689,10 @@ static int __init mtrr_init_finialize(void)
 {
 	if (!mtrr_if)
 		return 0;
-	if (use_intel())
-		mtrr_state_warn();
-	else {
+	if (use_intel()) {
+		if (!changed_by_mtrr_cleanup)
+			mtrr_state_warn();
+	} else {
 		/* The CPUs haven't MTRR and seem to not support SMP. They have
 		 * specific drivers, we use a tricky method to support
 		 * suspend/resume for them.
diff --git a/arch/x86/kernel/cpu/mtrr/mtrr.h b/arch/x86/kernel/cpu/mtrr/mtrr.h
index 2cc77eb6fea..2dc4ec656b2 100644
--- a/arch/x86/kernel/cpu/mtrr/mtrr.h
+++ b/arch/x86/kernel/cpu/mtrr/mtrr.h
@@ -81,6 +81,8 @@ void set_mtrr_done(struct set_mtrr_context *ctxt);
 void set_mtrr_cache_disable(struct set_mtrr_context *ctxt);
 void set_mtrr_prepare_save(struct set_mtrr_context *ctxt);
 
+void fill_mtrr_var_range(unsigned int index,
+		u32 base_lo, u32 base_hi, u32 mask_lo, u32 mask_hi);
 void get_mtrr_state(void);
 
 extern void set_mtrr_ops(struct mtrr_ops * ops);
@@ -92,6 +94,7 @@ extern struct mtrr_ops * mtrr_if;
 #define use_intel()	(mtrr_if && mtrr_if->use_intel_if == 1)
 
 extern unsigned int num_var_ranges;
+extern u64 mtrr_tom2;
 
 void mtrr_state_warn(void);
 const char *mtrr_attrib_to_str(int x);