[PATCH] large x86 setup cleanup.

Patrick Mochel did a great job here at splitting up some of the larger
messy parts of arch/i386/kernel/setup.c, and introduced a nice abstraction
which gives us a much nicer way to ensure we can add workarounds for vendor
specific bugs / features without polluting other vendor code paths.

Mark Haverkamp also brought this up to date for merging in my tree circa
2.5.14, and asides from 1-2 now fixed small thinkos, there haven't been
any problems.

This also features a workaround for an errata item on stepping C0 of
the Intel Pentium 4 Xeon, which isn't in your tree yet, where we must
disable the hardware prefetcher to ensure sane operation.

arch/i386/kernel/Makefile

@@ -2,7 +2,7 @@
 # Makefile for the linux kernel.
 #
 
-EXTRA_TARGETS := head.o init_task.o
+EXTRA_TARGETS := first_rule kernel.o head.o init_task.o
 
 O_TARGET := kernel.o
 
@@ -14,6 +14,7 @@ obj-y	:= process.o semaphore.o signal.o entry.o traps.o irq.o vm86.o \
 		bootflag.o
 
 obj-$(CONFIG_MCA)		+= mca.o
+obj-$(CONFIG_EISA)		+= eisa.o
 obj-$(CONFIG_MTRR)		+= mtrr.o
 obj-$(CONFIG_X86_MSR)		+= msr.o
 obj-$(CONFIG_X86_CPUID)		+= cpuid.o
@@ -33,4 +34,8 @@ endif
 
 EXTRA_AFLAGS   := -traditional
 
+kernel-subdir-y                 := cpu
+subdir-y                        := $(kernel-subdir-y)
+obj-y   +=      $(foreach dir,$(kernel-subdir-y),$(dir)/$(dir).o)
+
 include $(TOPDIR)/Rules.make

arch/i386/kernel/cpu/Makefile

+O_TARGET	:= cpu.o
+
+obj-y	+=	amd.o
+obj-y	+=	cyrix.o
+obj-y	+=	centaur.o
+obj-y	+=	transmeta.o
+obj-y	+=	intel.o
+obj-y	+=	rise.o
+obj-y	+=	nexgen.o
+obj-y	+=	umc.o
+
+obj-y	+=	common.o
+obj-y	+=	proc.o
+
+include $(TOPDIR)/Rules.make

arch/i386/kernel/cpu/amd.c

+#include <linux/init.h>
+#include <linux/bitops.h>
+#include <asm/io.h>
+#include <asm/processor.h>
+
+#include "cpu.h"
+
+/*
+ *	B step AMD K6 before B 9730xxxx have hardware bugs that can cause
+ *	misexecution of code under Linux. Owners of such processors should
+ *	contact AMD for precise details and a CPU swap.
+ *
+ *	See	http://www.multimania.com/poulot/k6bug.html
+ *		http://www.amd.com/K6/k6docs/revgd.html
+ *
+ *	The following test is erm.. interesting. AMD neglected to up
+ *	the chip setting when fixing the bug but they also tweaked some
+ *	performance at the same time..
+ */
+ 
+extern void vide(void);
+__asm__(".align 4\nvide: ret");
+
+static void __init init_amd(struct cpuinfo_x86 *c)
+{
+	u32 l, h;
+	int mbytes = max_mapnr >> (20-PAGE_SHIFT);
+	int r;
+
+	/*
+	 *	FIXME: We should handle the K5 here. Set up the write
+	 *	range and also turn on MSR 83 bits 4 and 31 (write alloc,
+	 *	no bus pipeline)
+	 */
+
+	/* Bit 31 in normal CPUID used for nonstandard 3DNow ID;
+	   3DNow is IDd by bit 31 in extended CPUID (1*32+31) anyway */
+	clear_bit(0*32+31, c->x86_capability);
+	
+	r = get_model_name(c);
+
+	switch(c->x86)
+	{
+		case 5:
+			if( c->x86_model < 6 )
+			{
+				/* Based on AMD doc 20734R - June 2000 */
+				if ( c->x86_model == 0 ) {
+					clear_bit(X86_FEATURE_APIC, c->x86_capability);
+					set_bit(X86_FEATURE_PGE, c->x86_capability);
+				}
+				break;
+			}
+			
+			if ( c->x86_model == 6 && c->x86_mask == 1 ) {
+				const int K6_BUG_LOOP = 1000000;
+				int n;
+				void (*f_vide)(void);
+				unsigned long d, d2;
+				
+				printk(KERN_INFO "AMD K6 stepping B detected - ");
+				
+				/*
+				 * It looks like AMD fixed the 2.6.2 bug and improved indirect 
+				 * calls at the same time.
+				 */
+
+				n = K6_BUG_LOOP;
+				f_vide = vide;
+				rdtscl(d);
+				while (n--) 
+					f_vide();
+				rdtscl(d2);
+				d = d2-d;
+				
+				/* Knock these two lines out if it debugs out ok */
+				printk(KERN_INFO "K6 BUG %ld %d (Report these if test report is incorrect)\n", d, 20*K6_BUG_LOOP);
+				printk(KERN_INFO "AMD K6 stepping B detected - ");
+				/* -- cut here -- */
+				if (d > 20*K6_BUG_LOOP) 
+					printk("system stability may be impaired when more than 32 MB are used.\n");
+				else 
+					printk("probably OK (after B9730xxxx).\n");
+				printk(KERN_INFO "Please see http://www.mygale.com/~poulot/k6bug.html\n");
+			}
+
+			/* K6 with old style WHCR */
+			if (c->x86_model < 8 ||
+			   (c->x86_model== 8 && c->x86_mask < 8)) {
+				/* We can only write allocate on the low 508Mb */
+				if(mbytes>508)
+					mbytes=508;
+
+				rdmsr(MSR_K6_WHCR, l, h);
+				if ((l&0x0000FFFF)==0) {
+					unsigned long flags;
+					l=(1<<0)|((mbytes/4)<<1);
+					local_irq_save(flags);
+					wbinvd();
+					wrmsr(MSR_K6_WHCR, l, h);
+					local_irq_restore(flags);
+					printk(KERN_INFO "Enabling old style K6 write allocation for %d Mb\n",
+						mbytes);
+				}
+				break;
+			}
+
+			if ((c->x86_model == 8 && c->x86_mask >7) ||
+			     c->x86_model == 9 || c->x86_model == 13) {
+				/* The more serious chips .. */
+
+				if(mbytes>4092)
+					mbytes=4092;
+
+				rdmsr(MSR_K6_WHCR, l, h);
+				if ((l&0xFFFF0000)==0) {
+					unsigned long flags;
+					l=((mbytes>>2)<<22)|(1<<16);
+					local_irq_save(flags);
+					wbinvd();
+					wrmsr(MSR_K6_WHCR, l, h);
+					local_irq_restore(flags);
+					printk(KERN_INFO "Enabling new style K6 write allocation for %d Mb\n",
+						mbytes);
+				}
+
+				/*  Set MTRR capability flag if appropriate */
+				if (c->x86_model == 13 || c->x86_model == 9 ||
+				   (c->x86_model == 8 && c->x86_mask >= 8))
+					set_bit(X86_FEATURE_K6_MTRR, c->x86_capability);
+				break;
+			}
+			break;
+
+		case 6: /* An Athlon/Duron */
+ 
+			/* Bit 15 of Athlon specific MSR 15, needs to be 0
+ 			 * to enable SSE on Palomino/Morgan CPU's.
+			 * If the BIOS didn't enable it already, enable it
+			 * here.
+			 */
+			if (c->x86_model == 6 || c->x86_model == 7) {
+				if (!test_bit(X86_FEATURE_XMM, c->x86_capability)) {
+					printk(KERN_INFO "Enabling disabled K7/SSE Support.\n");
+					rdmsr(MSR_K7_HWCR, l, h);
+					l &= ~0x00008000;
+					wrmsr(MSR_K7_HWCR, l, h);
+					set_bit(X86_FEATURE_XMM, c->x86_capability);
+				}
+			}
+			break;
+
+	}
+
+	display_cacheinfo(c);
+//	return r;
+}
+
+static void amd_identify(struct cpuinfo_x86 * c)
+{
+	u32 xlvl;
+
+	if (have_cpuid_p()) {
+		generic_identify(c);
+
+		/* AMD-defined flags: level 0x80000001 */
+		xlvl = cpuid_eax(0x80000000);
+		if ( (xlvl & 0xffff0000) == 0x80000000 ) {
+			if ( xlvl >= 0x80000001 )
+				c->x86_capability[1] = cpuid_edx(0x80000001);
+			if ( xlvl >= 0x80000004 )
+				get_model_name(c); /* Default name */
+		}
+	}
+}
+
+static unsigned int amd_size_cache(struct cpuinfo_x86 * c, unsigned int size)
+{
+	/* AMD errata T13 (order #21922) */
+	if ((c->x86 == 6)) {
+		if (c->x86_model == 3 && c->x86_mask == 0)	/* Duron Rev A0 */
+			size = 64;
+		if (c->x86_model == 4 &&
+		    (c->x86_mask==0 || c->x86_mask==1))	/* Tbird rev A1/A2 */
+			size = 256;
+	}
+	return size;
+}
+
+static struct cpu_dev amd_cpu_dev __initdata = {
+	c_vendor:	"AMD",
+	c_ident: 	{ "AuthenticAMD" },
+	c_models: {
+		{ X86_VENDOR_AMD,	4,
+		  {
+			  [3] "486 DX/2",
+			  [7] "486 DX/2-WB",
+			  [8] "486 DX/4", 
+			  [9] "486 DX/4-WB", 
+			  [14] "Am5x86-WT",
+			  [15] "Am5x86-WB" 
+		  }
+		},
+	},
+	c_init:		init_amd,
+	c_identify:	amd_identify,
+	c_size_cache:	amd_size_cache,
+};
+
+int __init amd_init_cpu(void)
+{
+	cpu_devs[X86_VENDOR_AMD] = &amd_cpu_dev;
+	return 0;
+}
+
+//early_arch_initcall(amd_init_cpu);

arch/i386/kernel/cpu/centaur.c

+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/bitops.h>
+#include <asm/processor.h>
+#include <asm/msr.h>
+#include "cpu.h"
+
+#ifdef CONFIG_X86_OOSTORE
+
+static u32 __init power2(u32 x)
+{
+	u32 s=1;
+	while(s<=x)
+		s<<=1;
+	return s>>=1;
+}
+
+
+/*
+ *	Set up an actual MCR
+ */
+ 
+static void __init centaur_mcr_insert(int reg, u32 base, u32 size, int key)
+{
+	u32 lo, hi;
+	
+	hi = base & ~0xFFF;
+	lo = ~(size-1);		/* Size is a power of 2 so this makes a mask */
+	lo &= ~0xFFF;		/* Remove the ctrl value bits */
+	lo |= key;		/* Attribute we wish to set */
+	wrmsr(reg+MSR_IDT_MCR0, lo, hi);
+	mtrr_centaur_report_mcr(reg, lo, hi);	/* Tell the mtrr driver */
+}
+
+/*
+ *	Figure what we can cover with MCR's
+ *
+ *	Shortcut: We know you can't put 4Gig of RAM on a winchip
+ */
+
+static u32 __init ramtop(void)		/* 16388 */
+{
+	int i;
+	u32 top = 0;
+	u32 clip = 0xFFFFFFFFUL;
+	
+	for (i = 0; i < e820.nr_map; i++) {
+		unsigned long start, end;
+
+		if (e820.map[i].addr > 0xFFFFFFFFUL)
+			continue;
+		/*
+		 *	Don't MCR over reserved space. Ignore the ISA hole
+		 *	we frob around that catastrophy already
+		 */
+		 			
+		if (e820.map[i].type == E820_RESERVED)
+		{
+			if(e820.map[i].addr >= 0x100000UL && e820.map[i].addr < clip)
+				clip = e820.map[i].addr;
+			continue;
+		}
+		start = e820.map[i].addr;
+		end = e820.map[i].addr + e820.map[i].size;
+		if (start >= end)
+			continue;
+		if (end > top)
+			top = end;
+	}
+	/* Everything below 'top' should be RAM except for the ISA hole.
+	   Because of the limited MCR's we want to map NV/ACPI into our
+	   MCR range for gunk in RAM 
+	   
+	   Clip might cause us to MCR insufficient RAM but that is an
+	   acceptable failure mode and should only bite obscure boxes with
+	   a VESA hole at 15Mb
+	   
+	   The second case Clip sometimes kicks in is when the EBDA is marked
+	   as reserved. Again we fail safe with reasonable results
+	*/
+	
+	if(top>clip)
+		top=clip;
+		
+	return top;
+}
+
+/*
+ *	Compute a set of MCR's to give maximum coverage
+ */
+
+static int __init centaur_mcr_compute(int nr, int key)
+{
+	u32 mem = ramtop();
+	u32 root = power2(mem);
+	u32 base = root;
+	u32 top = root;
+	u32 floor = 0;
+	int ct = 0;
+	
+	while(ct<nr)
+	{
+		u32 fspace = 0;
+
+		/*
+		 *	Find the largest block we will fill going upwards
+		 */
+
+		u32 high = power2(mem-top);	
+
+		/*
+		 *	Find the largest block we will fill going downwards
+		 */
+
+		u32 low = base/2;
+
+		/*
+		 *	Don't fill below 1Mb going downwards as there
+		 *	is an ISA hole in the way.
+		 */		
+		 
+		if(base <= 1024*1024)
+			low = 0;
+			
+		/*
+		 *	See how much space we could cover by filling below
+		 *	the ISA hole
+		 */
+		 
+		if(floor == 0)
+			fspace = 512*1024;
+		else if(floor ==512*1024)
+			fspace = 128*1024;
+
+		/* And forget ROM space */
+		
+		/*
+		 *	Now install the largest coverage we get
+		 */
+		 
+		if(fspace > high && fspace > low)
+		{
+			centaur_mcr_insert(ct, floor, fspace, key);
+			floor += fspace;
+		}
+		else if(high > low)
+		{
+			centaur_mcr_insert(ct, top, high, key);
+			top += high;
+		}
+		else if(low > 0)
+		{
+			base -= low;
+			centaur_mcr_insert(ct, base, low, key);
+		}
+		else break;
+		ct++;
+	}
+	/*
+	 *	We loaded ct values. We now need to set the mask. The caller
+	 *	must do this bit.
+	 */
+	 
+	return ct;
+}
+
+static void __init centaur_create_optimal_mcr(void)
+{
+	int i;
+	/*
+	 *	Allocate up to 6 mcrs to mark as much of ram as possible
+	 *	as write combining and weak write ordered.
+	 *
+	 *	To experiment with: Linux never uses stack operations for 
+	 *	mmio spaces so we could globally enable stack operation wc
+	 *
+	 *	Load the registers with type 31 - full write combining, all
+	 *	writes weakly ordered.
+	 */
+	int used = centaur_mcr_compute(6, 31);
+
+	/*
+	 *	Wipe unused MCRs
+	 */
+	 
+	for(i=used;i<8;i++)
+		wrmsr(MSR_IDT_MCR0+i, 0, 0);
+}
+
+static void __init winchip2_create_optimal_mcr(void)
+{
+	u32 lo, hi;
+	int i;
+
+	/*
+	 *	Allocate up to 6 mcrs to mark as much of ram as possible
+	 *	as write combining, weak store ordered.
+	 *
+	 *	Load the registers with type 25
+	 *		8	-	weak write ordering
+	 *		16	-	weak read ordering
+	 *		1	-	write combining
+	 */
+
+	int used = centaur_mcr_compute(6, 25);
+	
+	/*
+	 *	Mark the registers we are using.
+	 */
+	 
+	rdmsr(MSR_IDT_MCR_CTRL, lo, hi);
+	for(i=0;i<used;i++)
+		lo|=1<<(9+i);
+	wrmsr(MSR_IDT_MCR_CTRL, lo, hi);
+	
+	/*
+	 *	Wipe unused MCRs
+	 */
+	 
+	for(i=used;i<8;i++)
+		wrmsr(MSR_IDT_MCR0+i, 0, 0);
+}
+
+/*
+ *	Handle the MCR key on the Winchip 2.
+ */
+
+static void __init winchip2_unprotect_mcr(void)
+{
+	u32 lo, hi;
+	u32 key;
+	
+	rdmsr(MSR_IDT_MCR_CTRL, lo, hi);
+	lo&=~0x1C0;	/* blank bits 8-6 */
+	key = (lo>>17) & 7;
+	lo |= key<<6;	/* replace with unlock key */
+	wrmsr(MSR_IDT_MCR_CTRL, lo, hi);
+}
+
+static void __init winchip2_protect_mcr(void)
+{
+	u32 lo, hi;
+	
+	rdmsr(MSR_IDT_MCR_CTRL, lo, hi);
+	lo&=~0x1C0;	/* blank bits 8-6 */
+	wrmsr(MSR_IDT_MCR_CTRL, lo, hi);
+}
+#endif
+
+static void __init init_centaur(struct cpuinfo_x86 *c)
+{
+	enum {
+		ECX8=1<<1,
+		EIERRINT=1<<2,
+		DPM=1<<3,
+		DMCE=1<<4,
+		DSTPCLK=1<<5,
+		ELINEAR=1<<6,
+		DSMC=1<<7,
+		DTLOCK=1<<8,
+		EDCTLB=1<<8,
+		EMMX=1<<9,
+		DPDC=1<<11,
+		EBRPRED=1<<12,
+		DIC=1<<13,
+		DDC=1<<14,
+		DNA=1<<15,
+		ERETSTK=1<<16,
+		E2MMX=1<<19,
+		EAMD3D=1<<20,
+	};
+
+	char *name;
+	u32  fcr_set=0;
+	u32  fcr_clr=0;
+	u32  lo,hi,newlo;
+	u32  aa,bb,cc,dd;
+
+	/* Bit 31 in normal CPUID used for nonstandard 3DNow ID;
+	   3DNow is IDd by bit 31 in extended CPUID (1*32+31) anyway */
+	clear_bit(0*32+31, c->x86_capability);
+
+	switch (c->x86) {
+
+		case 5:
+			switch(c->x86_model) {
+			case 4:
+				name="C6";
+				fcr_set=ECX8|DSMC|EDCTLB|EMMX|ERETSTK;
+				fcr_clr=DPDC;
+				printk(KERN_NOTICE "Disabling bugged TSC.\n");
+				clear_bit(X86_FEATURE_TSC, c->x86_capability);
+#ifdef CONFIG_X86_OOSTORE
+				centaur_create_optimal_mcr();
+				/* Enable
+					write combining on non-stack, non-string
+					write combining on string, all types
+					weak write ordering 
+					
+				   The C6 original lacks weak read order 
+				   
+				   Note 0x120 is write only on Winchip 1 */
+				   
+				wrmsr(MSR_IDT_MCR_CTRL, 0x01F0001F, 0);
+#endif				
+				break;
+			case 8:
+				switch(c->x86_mask) {
+				default:
+					name="2";
+					break;
+				case 7 ... 9:
+					name="2A";
+					break;
+				case 10 ... 15:
+					name="2B";
+					break;
+				}
+				fcr_set=ECX8|DSMC|DTLOCK|EMMX|EBRPRED|ERETSTK|E2MMX|EAMD3D;
+				fcr_clr=DPDC;
+#ifdef CONFIG_X86_OOSTORE
+				winchip2_unprotect_mcr();
+				winchip2_create_optimal_mcr();
+				rdmsr(MSR_IDT_MCR_CTRL, lo, hi);
+				/* Enable
+					write combining on non-stack, non-string
+					write combining on string, all types
+					weak write ordering 
+				*/
+				lo|=31;				
+				wrmsr(MSR_IDT_MCR_CTRL, lo, hi);
+				winchip2_protect_mcr();
+#endif
+				break;
+			case 9:
+				name="3";
+				fcr_set=ECX8|DSMC|DTLOCK|EMMX|EBRPRED|ERETSTK|E2MMX|EAMD3D;
+				fcr_clr=DPDC;
+#ifdef CONFIG_X86_OOSTORE
+				winchip2_unprotect_mcr();
+				winchip2_create_optimal_mcr();
+				rdmsr(MSR_IDT_MCR_CTRL, lo, hi);
+				/* Enable
+					write combining on non-stack, non-string
+					write combining on string, all types
+					weak write ordering 
+				*/
+				lo|=31;				
+				wrmsr(MSR_IDT_MCR_CTRL, lo, hi);
+				winchip2_protect_mcr();
+#endif
+				break;
+			case 10:
+				name="4";
+				/* no info on the WC4 yet */
+				break;
+			default:
+				name="??";
+			}
+
+			rdmsr(MSR_IDT_FCR1, lo, hi);
+			newlo=(lo|fcr_set) & (~fcr_clr);
+
+			if (newlo!=lo) {
+				printk(KERN_INFO "Centaur FCR was 0x%X now 0x%X\n", lo, newlo );
+				wrmsr(MSR_IDT_FCR1, newlo, hi );
+			} else {
+				printk(KERN_INFO "Centaur FCR is 0x%X\n",lo);
+			}
+			/* Emulate MTRRs using Centaur's MCR. */
+			set_bit(X86_FEATURE_CENTAUR_MCR, c->x86_capability);
+			/* Report CX8 */
+			set_bit(X86_FEATURE_CX8, c->x86_capability);
+			/* Set 3DNow! on Winchip 2 and above. */
+			if (c->x86_model >=8)
+				set_bit(X86_FEATURE_3DNOW, c->x86_capability);
+			/* See if we can find out some more. */
+			if ( cpuid_eax(0x80000000) >= 0x80000005 ) {
+				/* Yes, we can. */
+				cpuid(0x80000005,&aa,&bb,&cc,&dd);
+				/* Add L1 data and code cache sizes. */
+				c->x86_cache_size = (cc>>24)+(dd>>24);
+			}
+			sprintf( c->x86_model_id, "WinChip %s", name );
+			break;
+
+		case 6:
+			switch (c->x86_model) {
+				case 6 ... 8:		/* Cyrix III family */
+					rdmsr (MSR_VIA_FCR, lo, hi);
+					lo |= (1<<1 | 1<<7);	/* Report CX8 & enable PGE */
+					wrmsr (MSR_VIA_FCR, lo, hi);
+
+					set_bit(X86_FEATURE_CX8, c->x86_capability);
+					set_bit(X86_FEATURE_3DNOW, c->x86_capability);
+
+					get_model_name(c);
+					display_cacheinfo(c);
+					break;
+			}
+			break;
+	}
+}
+
+static unsigned int centaur_size_cache(struct cpuinfo_x86 * c, unsigned int size)
+{
+	/* VIA C3 CPUs (670-68F) need further shifting. */
+	if ((c->x86 == 6) && ((c->x86_model == 7) || (c->x86_model == 8)))
+		size >>= 8;
+	return size;
+}
+
+static struct cpu_dev centaur_cpu_dev __initdata = {
+	c_vendor:	"Centaur",
+	c_ident:	{ "CentaurHauls" },
+	c_init:		init_centaur,
+	c_size_cache:	centaur_size_cache,
+};
+
+int __init centaur_init_cpu(void)
+{
+	cpu_devs[X86_VENDOR_CENTAUR] = &centaur_cpu_dev;
+	return 0;
+}
+
+//early_arch_initcall(centaur_init_cpu);

arch/i386/kernel/cpu/changelog

+/*
+ *  Enhanced CPU type detection by Mike Jagdis, Patrick St. Jean
+ *  and Martin Mares, November 1997.
+ *
+ *  Force Cyrix 6x86(MX) and M II processors to report MTRR capability
+ *  and Cyrix "coma bug" recognition by
+ *  Zoltn Bszrmnyi <zboszor@mail.externet.hu> February 1999.
+ * 
+ *  Force Centaur C6 processors to report MTRR capability.
+ *  Bart Hartgers <bart@etpmod.phys.tue.nl>, May 1999.
+ *
+ *  Intel Mobile Pentium II detection fix. Sean Gilley, June 1999.
+ *
+ *  IDT Winchip tweaks, misc clean ups.
+ *  Dave Jones <davej@suse.de>, August 1999
+ *
+ *  Better detection of Centaur/IDT WinChip models.
+ *  Bart Hartgers <bart@etpmod.phys.tue.nl>, August 1999.
+ *
+ *  Cleaned up cache-detection code
+ *  Dave Jones <davej@suse.de>, October 1999
+ *
+ *  Added proper L2 cache detection for Coppermine
+ *  Dragan Stancevic <visitor@valinux.com>, October 1999
+ *
+ *  Added the original array for capability flags but forgot to credit 
+ *  myself :) (~1998) Fixed/cleaned up some cpu_model_info and other stuff
+ *  Jauder Ho <jauderho@carumba.com>, January 2000
+ *
+ *  Detection for Celeron coppermine, identify_cpu() overhauled,
+ *  and a few other clean ups.
+ *  Dave Jones <davej@suse.de>, April 2000
+ *
+ *  Pentium III FXSR, SSE support
+ *  General FPU state handling cleanups
+ *  Gareth Hughes <gareth@valinux.com>, May 2000
+ *
+ *  Added proper Cascades CPU and L2 cache detection for Cascades
+ *  and 8-way type cache happy bunch from Intel:^)
+ *  Dragan Stancevic <visitor@valinux.com>, May 2000 
+ *
+ *  Forward port AMD Duron errata T13 from 2.2.17pre
+ *  Dave Jones <davej@suse.de>, August 2000
+ *
+ *  Forward port lots of fixes/improvements from 2.2.18pre
+ *  Cyrix III, Pentium IV support.
+ *  Dave Jones <davej@suse.de>, October 2000
+ *
+ *  Massive cleanup of CPU detection and bug handling;
+ *  Transmeta CPU detection,
+ *  H. Peter Anvin <hpa@zytor.com>, November 2000
+ *
+ *  VIA C3 Support.
+ *  Dave Jones <davej@suse.de>, March 2001
+ *
+ *  AMD Athlon/Duron/Thunderbird bluesmoke support.
+ *  Dave Jones <davej@suse.de>, April 2001.
+ *
+ *  CacheSize bug workaround updates for AMD, Intel & VIA Cyrix.
+ *  Dave Jones <davej@suse.de>, September, October 2001.
+ *
+ */
+

arch/i386/kernel/cpu/cpu.h

+
+struct cpu_model_info {
+	int vendor;
+	int family;
+	char *model_names[16];
+};
+
+/* attempt to consolidate cpu attributes */
+struct cpu_dev {
+	char	* c_vendor;
+
+	/* some have two possibilities for cpuid string */
+	char	* c_ident[2];	
+
+	struct		cpu_model_info c_models[4];
+
+	void		(*c_init)(struct cpuinfo_x86 * c);
+	void		(*c_identify)(struct cpuinfo_x86 * c);
+	unsigned int	(*c_size_cache)(struct cpuinfo_x86 * c, unsigned int size);
+};
+
+extern struct cpu_dev * cpu_devs [X86_VENDOR_NUM];
+
+extern int get_model_name(struct cpuinfo_x86 *c);
+extern void display_cacheinfo(struct cpuinfo_x86 *c);
+
+extern void generic_identify(struct cpuinfo_x86 * c);
+extern int have_cpuid_p(void);

arch/i386/kernel/cpu/cyrix.c

+#include <linux/init.h>
+#include <linux/bitops.h>
+#include <linux/delay.h>
+#include <linux/pci_ids.h>
+#include <asm/dma.h>
+#include <asm/io.h>
+#include <asm/processor.h>
+
+#include "cpu.h"
+
+/*
+ * Read NSC/Cyrix DEVID registers (DIR) to get more detailed info. about the CPU
+ */
+void __init do_cyrix_devid(unsigned char *dir0, unsigned char *dir1)
+{
+	unsigned char ccr2, ccr3;
+	unsigned long flags;
+	
+	/* we test for DEVID by checking whether CCR3 is writable */
+	local_irq_save(flags);
+	ccr3 = getCx86(CX86_CCR3);
+	setCx86(CX86_CCR3, ccr3 ^ 0x80);
+	getCx86(0xc0);   /* dummy to change bus */
+
+	if (getCx86(CX86_CCR3) == ccr3) {       /* no DEVID regs. */
+		ccr2 = getCx86(CX86_CCR2);
+		setCx86(CX86_CCR2, ccr2 ^ 0x04);
+		getCx86(0xc0);  /* dummy */
+
+		if (getCx86(CX86_CCR2) == ccr2) /* old Cx486SLC/DLC */
+			*dir0 = 0xfd;
+		else {                          /* Cx486S A step */
+			setCx86(CX86_CCR2, ccr2);
+			*dir0 = 0xfe;
+		}
+	}
+	else {
+		setCx86(CX86_CCR3, ccr3);  /* restore CCR3 */
+
+		/* read DIR0 and DIR1 CPU registers */
+		*dir0 = getCx86(CX86_DIR0);
+		*dir1 = getCx86(CX86_DIR1);
+	}
+	local_irq_restore(flags);
+}
+
+/*
+ * Cx86_dir0_msb is a HACK needed by check_cx686_cpuid/slop in bugs.h in
+ * order to identify the Cyrix CPU model after we're out of setup.c
+ *
+ * Actually since bugs.h doesnt even reference this perhaps someone should
+ * fix the documentation ???
+ */
+static unsigned char Cx86_dir0_msb __initdata = 0;
+
+static char Cx86_model[][9] __initdata = {
+	"Cx486", "Cx486", "5x86 ", "6x86", "MediaGX ", "6x86MX ",
+	"M II ", "Unknown"
+};
+static char Cx486_name[][5] __initdata = {
+	"SLC", "DLC", "SLC2", "DLC2", "SRx", "DRx",
+	"SRx2", "DRx2"
+};
+static char Cx486S_name[][4] __initdata = {
+	"S", "S2", "Se", "S2e"
+};
+static char Cx486D_name[][4] __initdata = {
+	"DX", "DX2", "?", "?", "?", "DX4"
+};
+static char Cx86_cb[] __initdata = "?.5x Core/Bus Clock";
+static char cyrix_model_mult1[] __initdata = "12??43";
+static char cyrix_model_mult2[] __initdata = "12233445";
+
+/*
+ * Reset the slow-loop (SLOP) bit on the 686(L) which is set by some old
+ * BIOSes for compatability with DOS games.  This makes the udelay loop
+ * work correctly, and improves performance.
+ *
+ * FIXME: our newer udelay uses the tsc. We dont need to frob with SLOP
+ */
+
+extern void calibrate_delay(void) __init;
+
+static void __init check_cx686_slop(struct cpuinfo_x86 *c)
+{
+	unsigned long flags;
+	
+	if (Cx86_dir0_msb == 3) {
+		unsigned char ccr3, ccr5;
+
+		local_irq_save(flags);
+		ccr3 = getCx86(CX86_CCR3);
+		setCx86(CX86_CCR3, (ccr3 & 0x0f) | 0x10); /* enable MAPEN  */
+		ccr5 = getCx86(CX86_CCR5);
+		if (ccr5 & 2)
+			setCx86(CX86_CCR5, ccr5 & 0xfd);  /* reset SLOP */
+		setCx86(CX86_CCR3, ccr3);                 /* disable MAPEN */
+		local_irq_restore(flags);
+
+		if (ccr5 & 2) { /* possible wrong calibration done */
+			printk(KERN_INFO "Recalibrating delay loop with SLOP bit reset\n");
+			calibrate_delay();
+			c->loops_per_jiffy = loops_per_jiffy;
+		}
+	}
+}
+
+static void __init init_cyrix(struct cpuinfo_x86 *c)
+{
+	unsigned char dir0, dir0_msn, dir0_lsn, dir1 = 0;
+	char *buf = c->x86_model_id;
+	const char *p = NULL;
+
+	/* Bit 31 in normal CPUID used for nonstandard 3DNow ID;
+	   3DNow is IDd by bit 31 in extended CPUID (1*32+31) anyway */
+	clear_bit(0*32+31, c->x86_capability);
+
+	/* Cyrix used bit 24 in extended (AMD) CPUID for Cyrix MMX extensions */
+	if ( test_bit(1*32+24, c->x86_capability) ) {
+		clear_bit(1*32+24, c->x86_capability);
+		set_bit(X86_FEATURE_CXMMX, c->x86_capability);
+	}
+
+	do_cyrix_devid(&dir0, &dir1);
+
+	check_cx686_slop(c);
+
+	Cx86_dir0_msb = dir0_msn = dir0 >> 4; /* identifies CPU "family"   */
+	dir0_lsn = dir0 & 0xf;                /* model or clock multiplier */
+
+	/* common case step number/rev -- exceptions handled below */
+	c->x86_model = (dir1 >> 4) + 1;
+	c->x86_mask = dir1 & 0xf;
+
+	/* Now cook; the original recipe is by Channing Corn, from Cyrix.
+	 * We do the same thing for each generation: we work out
+	 * the model, multiplier and stepping.  Black magic included,
+	 * to make the silicon step/rev numbers match the printed ones.
+	 */
+	 
+	switch (dir0_msn) {
+		unsigned char tmp;
+
+	case 0: /* Cx486SLC/DLC/SRx/DRx */
+		p = Cx486_name[dir0_lsn & 7];
+		break;
+
+	case 1: /* Cx486S/DX/DX2/DX4 */
+		p = (dir0_lsn & 8) ? Cx486D_name[dir0_lsn & 5]
+			: Cx486S_name[dir0_lsn & 3];
+		break;
+
+	case 2: /* 5x86 */
+		Cx86_cb[2] = cyrix_model_mult1[dir0_lsn & 5];
+		p = Cx86_cb+2;
+		break;
+
+	case 3: /* 6x86/6x86L */
+		Cx86_cb[1] = ' ';
+		Cx86_cb[2] = cyrix_model_mult1[dir0_lsn & 5];
+		if (dir1 > 0x21) { /* 686L */
+			Cx86_cb[0] = 'L';
+			p = Cx86_cb;
+			(c->x86_model)++;
+		} else             /* 686 */
+			p = Cx86_cb+1;
+		/* Emulate MTRRs using Cyrix's ARRs. */
+		set_bit(X86_FEATURE_CYRIX_ARR, c->x86_capability);
+		/* 6x86's contain this bug */
+		c->coma_bug = 1;
+		break;
+
+	case 4: /* MediaGX/GXm */
+#ifdef CONFIG_PCI
+		/* It isn't really a PCI quirk directly, but the cure is the
+		   same. The MediaGX has deep magic SMM stuff that handles the
+		   SB emulation. It thows away the fifo on disable_dma() which
+		   is wrong and ruins the audio. 
+
+		   Bug2: VSA1 has a wrap bug so that using maximum sized DMA 
+		   causes bad things. According to NatSemi VSA2 has another
+		   bug to do with 'hlt'. I've not seen any boards using VSA2
+		   and X doesn't seem to support it either so who cares 8).
+		   VSA1 we work around however.
+		*/
+
+		printk(KERN_INFO "Working around Cyrix MediaGX virtual DMA bugs.\n");
+		isa_dma_bridge_buggy = 2;
+#endif		
+		c->x86_cache_size=16;	/* Yep 16K integrated cache thats it */
+
+		/* GXm supports extended cpuid levels 'ala' AMD */
+		if (c->cpuid_level == 2) {
+			/* Enable Natsemi MMX extensions */
+			setCx86(CX86_CCR7, getCx86(CX86_CCR7) | 1);
+
+			get_model_name(c);  /* get CPU marketing name */
+			/*
+			 *  The 5510/5520 companion chips have a funky PIT
+			 *  that breaks the TSC synchronizing, so turn it off
+			 */
+			if(pci_find_device(PCI_VENDOR_ID_CYRIX, PCI_DEVICE_ID_CYRIX_5510, NULL) ||
+				pci_find_device(PCI_VENDOR_ID_CYRIX, PCI_DEVICE_ID_CYRIX_5520, NULL))
+				clear_bit(X86_FEATURE_TSC, c->x86_capability);
+			return;
+		}
+		else {  /* MediaGX */
+			Cx86_cb[2] = (dir0_lsn & 1) ? '3' : '4';
+			p = Cx86_cb+2;
+			c->x86_model = (dir1 & 0x20) ? 1 : 2;
+#ifndef CONFIG_CS5520
+			clear_bit(X86_FEATURE_TSC, c->x86_capability);
+#endif
+		}
+		break;
+
+        case 5: /* 6x86MX/M II */
+		if (dir1 > 7)
+		{
+			dir0_msn++;  /* M II */
+			/* Enable MMX extensions (App note 108) */
+			setCx86(CX86_CCR7, getCx86(CX86_CCR7)|1);
+		}
+		else
+		{
+			c->coma_bug = 1;      /* 6x86MX, it has the bug. */
+		}
+		tmp = (!(dir0_lsn & 7) || dir0_lsn & 1) ? 2 : 0;
+		Cx86_cb[tmp] = cyrix_model_mult2[dir0_lsn & 7];
+		p = Cx86_cb+tmp;
+        	if (((dir1 & 0x0f) > 4) || ((dir1 & 0xf0) == 0x20))
+			(c->x86_model)++;
+		/* Emulate MTRRs using Cyrix's ARRs. */
+		set_bit(X86_FEATURE_CYRIX_ARR, c->x86_capability);
+		break;
+
+	case 0xf:  /* Cyrix 486 without DEVID registers */
+		switch (dir0_lsn) {
+		case 0xd:  /* either a 486SLC or DLC w/o DEVID */
+			dir0_msn = 0;
+			p = Cx486_name[(c->hard_math) ? 1 : 0];
+			break;
+
+		case 0xe:  /* a 486S A step */
+			dir0_msn = 0;
+			p = Cx486S_name[0];
+			break;
+		}
+		break;
+
+	default:  /* unknown (shouldn't happen, we know everyone ;-) */
+		dir0_msn = 7;
+		break;
+	}
+	strcpy(buf, Cx86_model[dir0_msn & 7]);
+	if (p) strcat(buf, p);
+	return;
+}
+
+/*
+ * Cyrix CPUs without cpuid or with cpuid not yet enabled can be detected
+ * by the fact that they preserve the flags across the division of 5/2.
+ * PII and PPro exhibit this behavior too, but they have cpuid available.
+ */
+ 
+/*
+ * Perform the Cyrix 5/2 test. A Cyrix won't change
+ * the flags, while other 486 chips will.
+ */
+static inline int test_cyrix_52div(void)
+{
+	unsigned int test;
+
+	__asm__ __volatile__(
+	     "sahf\n\t"		/* clear flags (%eax = 0x0005) */
+	     "div %b2\n\t"	/* divide 5 by 2 */
+	     "lahf"		/* store flags into %ah */
+	     : "=a" (test)
+	     : "0" (5), "q" (2)
+	     : "cc");
+
+	/* AH is 0x02 on Cyrix after the divide.. */
+	return (unsigned char) (test >> 8) == 0x02;
+}
+
+static void cyrix_identify(struct cpuinfo_x86 * c)
+{
+	/* Detect Cyrix with disabled CPUID */
+	if ( c->x86 == 4 && test_cyrix_52div() ) {
+		unsigned char dir0, dir1;
+		
+		strcpy(c->x86_vendor_id, "CyrixInstead");
+	        c->x86_vendor = X86_VENDOR_CYRIX;
+	        
+	        /* Actually enable cpuid on the older cyrix */
+	    
+	    	/* Retrieve CPU revisions */
+	    	
+		do_cyrix_devid(&dir0, &dir1);
+
+		dir0>>=4;		
+		
+		/* Check it is an affected model */
+		
+   	        if (dir0 == 5 || dir0 == 3)
+   	        {
+			unsigned char ccr3, ccr4;
+			unsigned long flags;
+			printk(KERN_INFO "Enabling CPUID on Cyrix processor.\n");
+			local_irq_save(flags);
+			ccr3 = getCx86(CX86_CCR3);
+			setCx86(CX86_CCR3, (ccr3 & 0x0f) | 0x10); /* enable MAPEN  */
+			ccr4 = getCx86(CX86_CCR4);
+			setCx86(CX86_CCR4, ccr4 | 0x80);          /* enable cpuid  */
+			setCx86(CX86_CCR3, ccr3);                 /* disable MAPEN */
+			local_irq_restore(flags);
+		}
+	}
+	generic_identify(c);
+}
+
+static struct cpu_dev cyrix_cpu_dev __initdata = {
+	c_vendor:	"Cyrix",
+	c_ident: 	{ "CyrixInstead" },
+	c_init:		init_cyrix,
+	c_identify:	cyrix_identify,
+};
+
+int __init cyrix_init_cpu(void)
+{
+	cpu_devs[X86_VENDOR_CYRIX] = &cyrix_cpu_dev;
+	return 0;
+}
+
+//early_arch_initcall(cyrix_init_cpu);
+
+static struct cpu_dev nsc_cpu_dev __initdata = {
+	c_vendor:	"NSC",
+	c_ident: 	{ "Geode by NSC" },
+	c_init:		init_cyrix,
+	c_identify:	generic_identify,
+};
+
+int __init nsc_init_cpu(void)
+{
+	cpu_devs[X86_VENDOR_NSC] = &nsc_cpu_dev;
+	return 0;
+}
+
+//early_arch_initcall(nsc_init_cpu);

arch/i386/kernel/cpu/intel.c

+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/string.h>
+#include <linux/bitops.h>
+#include <linux/smp.h>
+#include <asm/processor.h>
+#include <asm/thread_info.h>
+#include <asm/msr.h>
+
+#include "cpu.h"
+
+static int disable_x86_serial_nr __initdata = 1;
+static int disable_P4_HT __initdata = 0;
+extern int trap_init_f00f_bug(void);
+
+/*
+ *	Early probe support logic for ppro memory erratum #50
+ *
+ *	This is called before we do cpu ident work
+ */
+ 
+int __init ppro_with_ram_bug(void)
+{
+	char vendor_id[16];
+	int ident;
+
+	/* Must have CPUID */
+	if(!have_cpuid_p())
+		return 0;
+	if(cpuid_eax(0)<1)
+		return 0;
+	
+	/* Must be Intel */
+	cpuid(0, &ident, 
+		(int *)&vendor_id[0],
+		(int *)&vendor_id[8],
+		(int *)&vendor_id[4]);
+	
+	if(memcmp(vendor_id, "IntelInside", 12))
+		return 0;
+	
+	ident = cpuid_eax(1);
+
+	/* Model 6 */
+
+	if(((ident>>8)&15)!=6)
+		return 0;
+	
+	/* Pentium Pro */
+
+	if(((ident>>4)&15)!=1)
+		return 0;
+	
+	if((ident&15) < 8)
+	{
+		printk(KERN_INFO "Pentium Pro with Errata#50 detected. Taking evasive action.\n");
+		return 1;
+	}
+	printk(KERN_INFO "Your Pentium Pro seems ok.\n");
+	return 0;
+}
+	
+static void __init squash_the_stupid_serial_number(struct cpuinfo_x86 *c)
+{
+	if( test_bit(X86_FEATURE_PN, c->x86_capability) &&
+	    disable_x86_serial_nr ) {
+		/* Disable processor serial number */
+		unsigned long lo,hi;
+		rdmsr(MSR_IA32_BBL_CR_CTL,lo,hi);
+		lo |= 0x200000;
+		wrmsr(MSR_IA32_BBL_CR_CTL,lo,hi);
+		printk(KERN_NOTICE "CPU serial number disabled.\n");
+		clear_bit(X86_FEATURE_PN, c->x86_capability);
+
+		/* Disabling the serial number may affect the cpuid level */
+		c->cpuid_level = cpuid_eax(0);
+	}
+}
+
+static int __init x86_serial_nr_setup(char *s)
+{
+	disable_x86_serial_nr = 0;
+	return 1;
+}
+__setup("serialnumber", x86_serial_nr_setup);
+
+static int __init P4_disable_ht(char *s)
+{
+	disable_P4_HT = 1;
+	return 1;
+}
+__setup("noht", P4_disable_ht);
+
+
+static void __init init_intel(struct cpuinfo_x86 *c)
+{
+	char *p = NULL;
+	unsigned int l1i = 0, l1d = 0, l2 = 0, l3 = 0; /* Cache sizes */
+
+#ifdef CONFIG_X86_F00F_BUG
+	/*
+	 * All current models of Pentium and Pentium with MMX technology CPUs
+	 * have the F0 0F bug, which lets nonpriviledged users lock up the system.
+	 * Note that the workaround only should be initialized once...
+	 */
+	c->f00f_bug = 0;
+	if ( c->x86 == 5 ) {
+		static int f00f_workaround_enabled = 0;
+
+		c->f00f_bug = 1;
+		if ( !f00f_workaround_enabled ) {
+			trap_init_f00f_bug();
+			printk(KERN_NOTICE "Intel Pentium with F0 0F bug - workaround enabled.\n");
+			f00f_workaround_enabled = 1;
+		}
+	}
+#endif
+
+
+	if (c->cpuid_level > 1) {
+		/* supports eax=2  call */
+		int i, j, n;
+		int regs[4];
+		unsigned char *dp = (unsigned char *)regs;
+
+		/* Number of times to iterate */
+		n = cpuid_eax(2) & 0xFF;
+
+		for ( i = 0 ; i < n ; i++ ) {
+			cpuid(2, &regs[0], &regs[1], &regs[2], &regs[3]);
+			
+			/* If bit 31 is set, this is an unknown format */
+			for ( j = 0 ; j < 3 ; j++ ) {
+				if ( regs[j] < 0 ) regs[j] = 0;
+			}
+
+			/* Byte 0 is level count, not a descriptor */
+			for ( j = 1 ; j < 16 ; j++ ) {
+				unsigned char des = dp[j];
+				unsigned char dl, dh;
+				unsigned int cs;
+
+				dh = des >> 4;
+				dl = des & 0x0F;
+
+				/* Black magic... */
+
+				switch ( dh )
+				{
+				case 0:
+					switch ( dl ) {
+					case 6:
+						/* L1 I cache */
+						l1i += 8;
+						break;
+					case 8:
+						/* L1 I cache */
+						l1i += 16;
+						break;
+					case 10:
+						/* L1 D cache */
+						l1d += 8;
+						break;
+					case 12:
+						/* L1 D cache */
+						l1d += 16;
+						break;
+					default:;
+						/* TLB, or unknown */
+					}
+					break;
+				case 2:
+					if ( dl ) {
+						/* L3 cache */
+						cs = (dl-1) << 9;
+						l3 += cs;
+					}
+					break;
+				case 4:
+					if ( c->x86 > 6 && dl ) {
+						/* P4 family */
+						/* L3 cache */
+						cs = 128 << (dl-1);
+						l3 += cs;
+						break;
+					}
+					/* else same as 8 - fall through */
+				case 8:
+					if ( dl ) {
+						/* L2 cache */
+						cs = 128 << (dl-1);
+						l2 += cs;
+					}
+					break;
+				case 6:
+					if (dl > 5) {
+						/* L1 D cache */
+						cs = 8<<(dl-6);
+						l1d += cs;
+					}
+					break;
+				case 7:
+					if ( dl >= 8 ) 
+					{
+						/* L2 cache */
+						cs = 64<<(dl-8);
+						l2 += cs;
+					} else {
+						/* L0 I cache, count as L1 */
+						cs = dl ? (16 << (dl-1)) : 12;
+						l1i += cs;
+					}
+					break;
+				default:
+					/* TLB, or something else we don't know about */
+					break;
+				}
+			}
+		}
+		if ( l1i || l1d )
+			printk(KERN_INFO "CPU: L1 I cache: %dK, L1 D cache: %dK\n",
+			      l1i, l1d);
+		if ( l2 )
+			printk(KERN_INFO "CPU: L2 cache: %dK\n", l2);
+		if ( l3 )
+			printk(KERN_INFO "CPU: L3 cache: %dK\n", l3);
+
+		/*
+		 * This assumes the L3 cache is shared; it typically lives in
+		 * the northbridge.  The L1 caches are included by the L2
+		 * cache, and so should not be included for the purpose of
+		 * SMP switching weights.
+		 */
+		c->x86_cache_size = l2 ? l2 : (l1i+l1d);
+	}
+
+	/* SEP CPUID bug: Pentium Pro reports SEP but doesn't have it */
+	if ( c->x86 == 6 && c->x86_model < 3 && c->x86_mask < 3 )
+		clear_bit(X86_FEATURE_SEP, c->x86_capability);
+	
+	/* Names for the Pentium II/Celeron processors 
+	   detectable only by also checking the cache size.
+	   Dixon is NOT a Celeron. */
+	if (c->x86 == 6) {
+		switch (c->x86_model) {
+		case 5:
+			if (l2 == 0)
+				p = "Celeron (Covington)";
+			if (l2 == 256)
+				p = "Mobile Pentium II (Dixon)";
+			break;
+			
+		case 6:
+			if (l2 == 128)
+				p = "Celeron (Mendocino)";
+			break;
+			
+		case 8:
+			if (l2 == 128)
+				p = "Celeron (Coppermine)";
+			break;
+		}
+	}
+
+	if ( p )
+		strcpy(c->x86_model_id, p);
+	
+#ifdef CONFIG_SMP
+	if (test_bit(X86_FEATURE_HT, c->x86_capability) && !disable_P4_HT) {
+		extern	int phys_proc_id[NR_CPUS];
+		
+		u32 	eax, ebx, ecx, edx;
+		int 	index_lsb, index_msb, tmp;
+		int	initial_apic_id;
+		int 	cpu = smp_processor_id();
+
+		cpuid(1, &eax, &ebx, &ecx, &edx);
+		smp_num_siblings = (ebx & 0xff0000) >> 16;
+
+		if (smp_num_siblings == 1) {
+			printk(KERN_INFO  "CPU: Hyper-Threading is disabled\n");
+		} else if (smp_num_siblings > 1 ) {
+			index_lsb = 0;
+			index_msb = 31;
+			/*
+			 * At this point we only support two siblings per
+			 * processor package.
+			 */
+#define NR_SIBLINGS	2
+			if (smp_num_siblings != NR_SIBLINGS) {
+				printk(KERN_WARNING "CPU: Unsupported number of the siblings %d", smp_num_siblings);
+				smp_num_siblings = 1;
+				goto too_many_siblings;
+			}
+			tmp = smp_num_siblings;
+			while ((tmp & 1) == 0) {
+				tmp >>=1 ;
+				index_lsb++;
+			}
+			tmp = smp_num_siblings;
+			while ((tmp & 0x80000000 ) == 0) {
+				tmp <<=1 ;
+				index_msb--;
+			}
+			if (index_lsb != index_msb )
+				index_msb++;
+			initial_apic_id = ebx >> 24 & 0xff;
+			phys_proc_id[cpu] = initial_apic_id >> index_msb;
+
+			printk(KERN_INFO  "CPU: Physical Processor ID: %d\n",
+                               phys_proc_id[cpu]);
+		}
+
+	}
+too_many_siblings:
+
+	if (disable_P4_HT)
+		clear_bit(X86_FEATURE_HT, &c->x86_capability);
+#endif
+
+	/* Disable the PN if appropriate */
+	squash_the_stupid_serial_number(c);
+}
+
+static unsigned int intel_size_cache(struct cpuinfo_x86 * c, unsigned int size)
+{
+	/* Intel PIII Tualatin. This comes in two flavours.
+	 * One has 256kb of cache, the other 512. We have no way
+	 * to determine which, so we use a boottime override
+	 * for the 512kb model, and assume 256 otherwise.
+	 */
+	if ((c->x86 == 6) && (c->x86_model == 11) && (size == 0))
+		size = 256;
+	return size;
+}
+
+static struct cpu_dev intel_cpu_dev __initdata = {
+	c_vendor:	"Intel",
+	c_ident: 	{ "GenuineIntel" },
+	c_models: {
+		{ X86_VENDOR_INTEL,	4,
+		  { 
+			  [0] "486 DX-25/33", 
+			  [1] "486 DX-50", 
+			  [2] "486 SX", 
+			  [3] "486 DX/2", 
+			  [4] "486 SL", 
+			  [5] "486 SX/2", 
+			  [7] "486 DX/2-WB", 
+			  [8] "486 DX/4", 
+			  [9] "486 DX/4-WB"
+		  }
+		},
+		{ X86_VENDOR_INTEL,	5,
+		  { 
+			  [0] "Pentium 60/66 A-step", 
+			  [1] "Pentium 60/66", 
+			  [2] "Pentium 75 - 200",
+			  [3] "OverDrive PODP5V83", 
+			  [4] "Pentium MMX",
+			  [7] "Mobile Pentium 75 - 200", 
+			  [8] "Mobile Pentium MMX"
+		  }
+		},
+	},
+	c_init:		init_intel,
+	c_identify:	generic_identify,
+	c_size_cache:	intel_size_cache,
+};
+
+__init int intel_cpu_init(void)
+{
+	cpu_devs[X86_VENDOR_INTEL] = &intel_cpu_dev;
+	return 0;
+}
+
+// arch_initcall(intel_cpu_init);
+

arch/i386/kernel/cpu/nexgen.c

+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/string.h>
+#include <asm/processor.h>
+
+#include "cpu.h"
+
+/*
+ *	Detect a NexGen CPU running without BIOS hypercode new enough
+ *	to have CPUID. (Thanks to Herbert Oppmann)
+ */
+ 
+static int __init deep_magic_nexgen_probe(void)
+{
+	int ret;
+	
+	__asm__ __volatile__ (
+		"	movw	$0x5555, %%ax\n"
+		"	xorw	%%dx,%%dx\n"
+		"	movw	$2, %%cx\n"
+		"	divw	%%cx\n"
+		"	movl	$0, %%eax\n"
+		"	jnz	1f\n"
+		"	movl	$1, %%eax\n"
+		"1:\n" 
+		: "=a" (ret) : : "cx", "dx" );
+	return  ret;
+}
+
+static void __init init_nexgen(struct cpuinfo_x86 * c)
+{
+	c->x86_cache_size = 256; /* A few had 1 MB... */
+}
+
+static void nexgen_identify(struct cpuinfo_x86 * c)
+{
+	/* Detect NexGen with old hypercode */
+	if ( deep_magic_nexgen_probe() ) {
+		strcpy(c->x86_vendor_id, "NexGenDriven");
+	}
+	generic_identify(c);
+}
+
+static struct cpu_dev nexgen_cpu_dev __initdata = {
+	c_vendor:	"Nexgen",
+	c_ident:	{ "NexGenDriven" },
+	c_models: {
+		{ X86_VENDOR_NEXGEN,5, { [1] "Nx586" } },
+	},
+	c_init:		init_nexgen,
+	c_identify:	nexgen_identify,
+};
+
+int __init nexgen_init_cpu(void)
+{
+	cpu_devs[X86_VENDOR_NEXGEN] = &nexgen_cpu_dev;
+	return 0;
+}
+
+//early_arch_initcall(nexgen_init_cpu);

arch/i386/kernel/cpu/proc.c

+#include <linux/smp.h>
+#include <linux/timex.h>
+#include <linux/string.h>
+#include <asm/semaphore.h>
+#include <linux/seq_file.h>
+
+/*
+ *	Get CPU information for use by the procfs.
+ */
+static int show_cpuinfo(struct seq_file *m, void *v)
+{
+	/* 
+	 * These flag bits must match the definitions in <asm/cpufeature.h>.
+	 * NULL means this bit is undefined or reserved; either way it doesn't
+	 * have meaning as far as Linux is concerned.  Note that it's important
+	 * to realize there is a difference between this table and CPUID -- if
+	 * applications want to get the raw CPUID data, they should access
+	 * /dev/cpu/<cpu_nr>/cpuid instead.
+	 */
+	static char *x86_cap_flags[] = {
+		/* Intel-defined */
+	        "fpu", "vme", "de", "pse", "tsc", "msr", "pae", "mce",
+	        "cx8", "apic", NULL, "sep", "mtrr", "pge", "mca", "cmov",
+	        "pat", "pse36", "pn", "clflush", NULL, "dts", "acpi", "mmx",
+	        "fxsr", "sse", "sse2", "ss", "ht", "tm", "ia64", NULL,
+
+		/* AMD-defined */
+		NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
+		NULL, NULL, NULL, "syscall", NULL, NULL, NULL, NULL,
+		NULL, NULL, NULL, "mp", NULL, NULL, "mmxext", NULL,
+		NULL, NULL, NULL, NULL, NULL, "lm", "3dnowext", "3dnow",
+
+		/* Transmeta-defined */
+		"recovery", "longrun", NULL, "lrti", NULL, NULL, NULL, NULL,
+		NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
+		NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
+		NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
+
+		/* Other (Linux-defined) */
+		"cxmmx", "k6_mtrr", "cyrix_arr", "centaur_mcr", NULL, NULL, NULL, NULL,
+		NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
+		NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
+		NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
+	};
+	struct cpuinfo_x86 *c = v;
+	int i, n = c - cpu_data;
+	int fpu_exception;
+
+#ifdef CONFIG_SMP
+	if (!(cpu_online_map & (1<<n)))
+		return 0;
+#endif
+	seq_printf(m, "processor\t: %d\n"
+		"vendor_id\t: %s\n"
+		"cpu family\t: %d\n"
+		"model\t\t: %d\n"
+		"model name\t: %s\n",
+		n,
+		c->x86_vendor_id[0] ? c->x86_vendor_id : "unknown",
+		c->x86,
+		c->x86_model,
+		c->x86_model_id[0] ? c->x86_model_id : "unknown");
+
+	if (c->x86_mask || c->cpuid_level >= 0)
+		seq_printf(m, "stepping\t: %d\n", c->x86_mask);
+	else
+		seq_printf(m, "stepping\t: unknown\n");
+
+	if ( test_bit(X86_FEATURE_TSC, &c->x86_capability) ) {
+		seq_printf(m, "cpu MHz\t\t: %lu.%03lu\n",
+			cpu_khz / 1000, (cpu_khz % 1000));
+	}
+
+	/* Cache size */
+	if (c->x86_cache_size >= 0)
+		seq_printf(m, "cache size\t: %d KB\n", c->x86_cache_size);
+	
+	/* We use exception 16 if we have hardware math and we've either seen it or the CPU claims it is internal */
+	fpu_exception = c->hard_math && (ignore_irq13 || cpu_has_fpu);
+	seq_printf(m, "fdiv_bug\t: %s\n"
+			"hlt_bug\t\t: %s\n"
+			"f00f_bug\t: %s\n"
+			"coma_bug\t: %s\n"
+			"fpu\t\t: %s\n"
+			"fpu_exception\t: %s\n"
+			"cpuid level\t: %d\n"
+			"wp\t\t: %s\n"
+			"flags\t\t:",
+		     c->fdiv_bug ? "yes" : "no",
+		     c->hlt_works_ok ? "no" : "yes",
+		     c->f00f_bug ? "yes" : "no",
+		     c->coma_bug ? "yes" : "no",
+		     c->hard_math ? "yes" : "no",
+		     fpu_exception ? "yes" : "no",
+		     c->cpuid_level,
+		     c->wp_works_ok ? "yes" : "no");
+
+	for ( i = 0 ; i < 32*NCAPINTS ; i++ )
+		if ( test_bit(i, &c->x86_capability) &&
+		     x86_cap_flags[i] != NULL )
+			seq_printf(m, " %s", x86_cap_flags[i]);
+
+	seq_printf(m, "\nbogomips\t: %lu.%02lu\n\n",
+		     c->loops_per_jiffy/(500000/HZ),
+		     (c->loops_per_jiffy/(5000/HZ)) % 100);
+	return 0;
+}
+
+static void *c_start(struct seq_file *m, loff_t *pos)
+{
+	return *pos < NR_CPUS ? cpu_data + *pos : NULL;
+}
+static void *c_next(struct seq_file *m, void *v, loff_t *pos)
+{
+	++*pos;
+	return c_start(m, pos);
+}
+static void c_stop(struct seq_file *m, void *v)
+{
+}
+struct seq_operations cpuinfo_op = {
+	start:	c_start,
+	next:	c_next,
+	stop:	c_stop,
+	show:	show_cpuinfo,
+};

arch/i386/kernel/cpu/rise.c

+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/bitops.h>
+#include <asm/processor.h>
+
+#include "cpu.h"
+
+static void __init init_rise(struct cpuinfo_x86 *c)
+{
+	printk("CPU: Rise iDragon");
+	if (c->x86_model > 2)
+		printk(" II");
+	printk("\n");
+
+	/* Unhide possibly hidden capability flags
+	   The mp6 iDragon family don't have MSRs.
+	   We switch on extra features with this cpuid weirdness: */
+	__asm__ (
+		"movl $0x6363452a, %%eax\n\t"
+		"movl $0x3231206c, %%ecx\n\t"
+		"movl $0x2a32313a, %%edx\n\t"
+		"cpuid\n\t"
+		"movl $0x63634523, %%eax\n\t"
+		"movl $0x32315f6c, %%ecx\n\t"
+		"movl $0x2333313a, %%edx\n\t"
+		"cpuid\n\t" : : : "eax", "ebx", "ecx", "edx"
+	);
+	set_bit(X86_FEATURE_CX8, c->x86_capability);
+}
+
+static struct cpu_dev rise_cpu_dev __initdata = {
+	c_vendor:	"Rise",
+	c_ident:	{ "RiseRiseRise" },
+	c_models: {
+		{ X86_VENDOR_RISE,	5,
+		  { 
+			  [0] "iDragon", 
+			  [2] "iDragon", 
+			  [8] "iDragon II", 
+			  [9] "iDragon II"
+		  }
+		},
+	},
+	c_init:		init_rise,
+};
+
+int __init rise_init_cpu(void)
+{
+	cpu_devs[X86_VENDOR_RISE] = &rise_cpu_dev;
+	return 0;
+}
+
+//early_arch_initcall(rise_init_cpu);

arch/i386/kernel/cpu/transmeta.c

+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <asm/processor.h>
+#include <asm/msr.h>
+#include "cpu.h"
+
+static void __init init_transmeta(struct cpuinfo_x86 *c)
+{
+	unsigned int cap_mask, uk, max, dummy;
+	unsigned int cms_rev1, cms_rev2;
+	unsigned int cpu_rev, cpu_freq, cpu_flags;
+	char cpu_info[65];
+
+	get_model_name(c);	/* Same as AMD/Cyrix */
+	display_cacheinfo(c);
+
+	/* Print CMS and CPU revision */
+	max = cpuid_eax(0x80860000);
+	if ( max >= 0x80860001 ) {
+		cpuid(0x80860001, &dummy, &cpu_rev, &cpu_freq, &cpu_flags); 
+		printk(KERN_INFO "CPU: Processor revision %u.%u.%u.%u, %u MHz\n",
+		       (cpu_rev >> 24) & 0xff,
+		       (cpu_rev >> 16) & 0xff,
+		       (cpu_rev >> 8) & 0xff,
+		       cpu_rev & 0xff,
+		       cpu_freq);
+	}
+	if ( max >= 0x80860002 ) {
+		cpuid(0x80860002, &dummy, &cms_rev1, &cms_rev2, &dummy);
+		printk(KERN_INFO "CPU: Code Morphing Software revision %u.%u.%u-%u-%u\n",
+		       (cms_rev1 >> 24) & 0xff,
+		       (cms_rev1 >> 16) & 0xff,
+		       (cms_rev1 >> 8) & 0xff,
+		       cms_rev1 & 0xff,
+		       cms_rev2);
+	}
+	if ( max >= 0x80860006 ) {
+		cpuid(0x80860003,
+		      (void *)&cpu_info[0],
+		      (void *)&cpu_info[4],
+		      (void *)&cpu_info[8],
+		      (void *)&cpu_info[12]);
+		cpuid(0x80860004,
+		      (void *)&cpu_info[16],
+		      (void *)&cpu_info[20],
+		      (void *)&cpu_info[24],
+		      (void *)&cpu_info[28]);
+		cpuid(0x80860005,
+		      (void *)&cpu_info[32],
+		      (void *)&cpu_info[36],
+		      (void *)&cpu_info[40],
+		      (void *)&cpu_info[44]);
+		cpuid(0x80860006,
+		      (void *)&cpu_info[48],
+		      (void *)&cpu_info[52],
+		      (void *)&cpu_info[56],
+		      (void *)&cpu_info[60]);
+		cpu_info[64] = '\0';
+		printk(KERN_INFO "CPU: %s\n", cpu_info);
+	}
+
+	/* Unhide possibly hidden capability flags */
+	rdmsr(0x80860004, cap_mask, uk);
+	wrmsr(0x80860004, ~0, uk);
+	c->x86_capability[0] = cpuid_edx(0x00000001);
+	wrmsr(0x80860004, cap_mask, uk);
+}
+
+static void transmeta_identify(struct cpuinfo_x86 * c)
+{
+	u32 xlvl;
+	generic_identify(c);
+
+	/* Transmeta-defined flags: level 0x80860001 */
+	xlvl = cpuid_eax(0x80860000);
+	if ( (xlvl & 0xffff0000) == 0x80860000 ) {
+		if (  xlvl >= 0x80860001 )
+			c->x86_capability[2] = cpuid_edx(0x80860001);
+	}
+}
+
+static struct cpu_dev transmeta_cpu_dev __initdata = {
+	c_vendor:	"Transmeta",
+	c_ident:	{ "GenuineTMx86", "TransmetaCPU" },
+	c_init:		init_transmeta,
+	c_identify:	transmeta_identify,
+};
+
+int __init transmeta_init_cpu(void)
+{
+	cpu_devs[X86_VENDOR_TRANSMETA] = &transmeta_cpu_dev;
+	return 0;
+}
+
+//early_arch_initcall(transmeta_init_cpu);

arch/i386/kernel/cpu/umc.c

+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <asm/processor.h>
+#include "cpu.h"
+
+/* UMC chips appear to be only either 386 or 486, so no special init takes place.
+ */
+static void __init init_umc(struct cpuinfo_x86 * c)
+{
+
+}
+
+static struct cpu_dev umc_cpu_dev __initdata = {
+	c_vendor:	"UMC",
+	c_ident: 	{ "UMC UMC UMC" },
+	c_models: {
+		{ X86_VENDOR_UMC,	4,
+		  { 
+			  [1] "U5D", 
+			  [2] "U5S", 
+		  }
+		},
+	},
+	c_init:		init_umc,
+};
+
+int __init umc_init_cpu(void)
+{
+	cpu_devs[X86_VENDOR_UMC] = &umc_cpu_dev;
+	return 0;
+}
+
+//early_arch_initcall(umc_init_cpu);

include/asm-i386/processor.h

@@ -60,6 +60,7 @@ struct cpuinfo_x86 {
 #define X86_VENDOR_RISE 6
 #define X86_VENDOR_TRANSMETA 7
 #define X86_VENDOR_NSC 8
+#define X86_VENDOR_NUM 9
 #define X86_VENDOR_UNKNOWN 0xff
 
 /*