[mirror_ubuntu-bionic-kernel.git] / arch / x86 / kernel / cpu / amd.c

#include <linux/init.h>
#include <linux/bitops.h>
#include <linux/mm.h>

#include <linux/io.h>
#include <asm/processor.h>
#include <asm/apic.h>
#include <asm/cpu.h>
#include <asm/pci-direct.h>

#ifdef CONFIG_X86_64
# include <asm/numa_64.h>
# include <asm/mmconfig.h>
# include <asm/cacheflush.h>
#endif

#include "cpu.h"

#ifdef CONFIG_X86_32
/*
 *	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 __cpuinit init_amd_k5(struct cpuinfo_x86 *c)
{
/*
 * General Systems BIOSen alias the cpu frequency registers
 * of the Elan at 0x000df000. Unfortuantly, one of the Linux
 * drivers subsequently pokes it, and changes the CPU speed.
 * Workaround : Remove the unneeded alias.
 */
#define CBAR		(0xfffc) /* Configuration Base Address  (32-bit) */
#define CBAR_ENB	(0x80000000)
#define CBAR_KEY	(0X000000CB)
	if (c->x86_model == 9 || c->x86_model == 10) {
		if (inl(CBAR) & CBAR_ENB)
			outl(0 | CBAR_KEY, CBAR);
	}
}


static void __cpuinit init_amd_k6(struct cpuinfo_x86 *c)
{
	u32 l, h;
	int mbytes = num_physpages >> (20-PAGE_SHIFT);

	if (c->x86_model < 6) {
		/* Based on AMD doc 20734R - June 2000 */
		if (c->x86_model == 0) {
			clear_cpu_cap(c, X86_FEATURE_APIC);
			set_cpu_cap(c, X86_FEATURE_PGE);
		}
		return;
	}

	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;

		if (d > 20*K6_BUG_LOOP)
			printk(KERN_CONT
				"system stability may be impaired when more than 32 MB are used.\n");
		else
			printk(KERN_CONT "probably OK (after B9730xxxx).\n");
		printk(KERN_INFO "Please see http://membres.lycos.fr/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);
		}
		return;
	}

	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);
		}

		return;
	}

	if (c->x86_model == 10) {
		/* AMD Geode LX is model 10 */
		/* placeholder for any needed mods */
		return;
	}
}

static void __cpuinit amd_k7_smp_check(struct cpuinfo_x86 *c)
{
#ifdef CONFIG_SMP
	/* calling is from identify_secondary_cpu() ? */
	if (!c->cpu_index)
		return;

	/*
	 * Certain Athlons might work (for various values of 'work') in SMP
	 * but they are not certified as MP capable.
	 */
	/* Athlon 660/661 is valid. */
	if ((c->x86_model == 6) && ((c->x86_mask == 0) ||
	    (c->x86_mask == 1)))
		goto valid_k7;

	/* Duron 670 is valid */
	if ((c->x86_model == 7) && (c->x86_mask == 0))
		goto valid_k7;

	/*
	 * Athlon 662, Duron 671, and Athlon >model 7 have capability
	 * bit. It's worth noting that the A5 stepping (662) of some
	 * Athlon XP's have the MP bit set.
	 * See http://www.heise.de/newsticker/data/jow-18.10.01-000 for
	 * more.
	 */
	if (((c->x86_model == 6) && (c->x86_mask >= 2)) ||
	    ((c->x86_model == 7) && (c->x86_mask >= 1)) ||
	     (c->x86_model > 7))
		if (cpu_has_mp)
			goto valid_k7;

	/* If we get here, not a certified SMP capable AMD system. */

	/*
	 * Don't taint if we are running SMP kernel on a single non-MP
	 * approved Athlon
	 */
	WARN_ONCE(1, "WARNING: This combination of AMD"
		" processors is not suitable for SMP.\n");
	if (!test_taint(TAINT_UNSAFE_SMP))
		add_taint(TAINT_UNSAFE_SMP);

valid_k7:
	;
#endif
}

static void __cpuinit init_amd_k7(struct cpuinfo_x86 *c)
{
	u32 l, h;

	/*
	 * Bit 15 of Athlon specific MSR 15, needs to be 0
	 * to enable SSE on Palomino/Morgan/Barton CPU's.
	 * If the BIOS didn't enable it already, enable it here.
	 */
	if (c->x86_model >= 6 && c->x86_model <= 10) {
		if (!cpu_has(c, X86_FEATURE_XMM)) {
			printk(KERN_INFO "Enabling disabled K7/SSE Support.\n");
			rdmsr(MSR_K7_HWCR, l, h);
			l &= ~0x00008000;
			wrmsr(MSR_K7_HWCR, l, h);
			set_cpu_cap(c, X86_FEATURE_XMM);
		}
	}

	/*
	 * It's been determined by AMD that Athlons since model 8 stepping 1
	 * are more robust with CLK_CTL set to 200xxxxx instead of 600xxxxx
	 * As per AMD technical note 27212 0.2
	 */
	if ((c->x86_model == 8 && c->x86_mask >= 1) || (c->x86_model > 8)) {
		rdmsr(MSR_K7_CLK_CTL, l, h);
		if ((l & 0xfff00000) != 0x20000000) {
			printk(KERN_INFO
			    "CPU: CLK_CTL MSR was %x. Reprogramming to %x\n",
					l, ((l & 0x000fffff)|0x20000000));
			wrmsr(MSR_K7_CLK_CTL, (l & 0x000fffff)|0x20000000, h);
		}
	}

	set_cpu_cap(c, X86_FEATURE_K7);

	amd_k7_smp_check(c);
}
#endif

#ifdef CONFIG_NUMA
/*
 * To workaround broken NUMA config.  Read the comment in
 * srat_detect_node().
 */
static int __cpuinit nearby_node(int apicid)
{
	int i, node;

	for (i = apicid - 1; i >= 0; i--) {
		node = __apicid_to_node[i];
		if (node != NUMA_NO_NODE && node_online(node))
			return node;
	}
	for (i = apicid + 1; i < MAX_LOCAL_APIC; i++) {
		node = __apicid_to_node[i];
		if (node != NUMA_NO_NODE && node_online(node))
			return node;
	}
	return first_node(node_online_map); /* Shouldn't happen */
}
#endif

/*
 * Fixup core topology information for
 * (1) AMD multi-node processors
 *     Assumption: Number of cores in each internal node is the same.
 * (2) AMD processors supporting compute units
 */
#ifdef CONFIG_X86_HT
static void __cpuinit amd_get_topology(struct cpuinfo_x86 *c)
{
	u32 nodes, cores_per_cu = 1;
	u8 node_id;
	int cpu = smp_processor_id();

	/* get information required for multi-node processors */
	if (cpu_has(c, X86_FEATURE_TOPOEXT)) {
		u32 eax, ebx, ecx, edx;

		cpuid(0x8000001e, &eax, &ebx, &ecx, &edx);
		nodes = ((ecx >> 8) & 7) + 1;
		node_id = ecx & 7;

		/* get compute unit information */
		smp_num_siblings = ((ebx >> 8) & 3) + 1;
		c->compute_unit_id = ebx & 0xff;
		cores_per_cu += ((ebx >> 8) & 3);
	} else if (cpu_has(c, X86_FEATURE_NODEID_MSR)) {
		u64 value;

		rdmsrl(MSR_FAM10H_NODE_ID, value);
		nodes = ((value >> 3) & 7) + 1;
		node_id = value & 7;
	} else
		return;

	/* fixup multi-node processor information */
	if (nodes > 1) {
		u32 cores_per_node;
		u32 cus_per_node;

		set_cpu_cap(c, X86_FEATURE_AMD_DCM);
		cores_per_node = c->x86_max_cores / nodes;
		cus_per_node = cores_per_node / cores_per_cu;

		/* store NodeID, use llc_shared_map to store sibling info */
		per_cpu(cpu_llc_id, cpu) = node_id;

		/* core id has to be in the [0 .. cores_per_node - 1] range */
		c->cpu_core_id %= cores_per_node;
		c->compute_unit_id %= cus_per_node;
	}
}
#endif

/*
 * On a AMD dual core setup the lower bits of the APIC id distingush the cores.
 * Assumes number of cores is a power of two.
 */
static void __cpuinit amd_detect_cmp(struct cpuinfo_x86 *c)
{
#ifdef CONFIG_X86_HT
	unsigned bits;
	int cpu = smp_processor_id();

	bits = c->x86_coreid_bits;
	/* Low order bits define the core id (index of core in socket) */
	c->cpu_core_id = c->initial_apicid & ((1 << bits)-1);
	/* Convert the initial APIC ID into the socket ID */
	c->phys_proc_id = c->initial_apicid >> bits;
	/* use socket ID also for last level cache */
	per_cpu(cpu_llc_id, cpu) = c->phys_proc_id;
	amd_get_topology(c);
#endif
}

int amd_get_nb_id(int cpu)
{
	int id = 0;
#ifdef CONFIG_SMP
	id = per_cpu(cpu_llc_id, cpu);
#endif
	return id;
}
EXPORT_SYMBOL_GPL(amd_get_nb_id);

static void __cpuinit srat_detect_node(struct cpuinfo_x86 *c)
{
#ifdef CONFIG_NUMA
	int cpu = smp_processor_id();
	int node;
	unsigned apicid = c->apicid;

	node = numa_cpu_node(cpu);
	if (node == NUMA_NO_NODE)
		node = per_cpu(cpu_llc_id, cpu);

	if (!node_online(node)) {
		/*
		 * Two possibilities here:
		 *
		 * - The CPU is missing memory and no node was created.  In
		 *   that case try picking one from a nearby CPU.
		 *
		 * - The APIC IDs differ from the HyperTransport node IDs
		 *   which the K8 northbridge parsing fills in.  Assume
		 *   they are all increased by a constant offset, but in
		 *   the same order as the HT nodeids.  If that doesn't
		 *   result in a usable node fall back to the path for the
		 *   previous case.
		 *
		 * This workaround operates directly on the mapping between
		 * APIC ID and NUMA node, assuming certain relationship
		 * between APIC ID, HT node ID and NUMA topology.  As going
		 * through CPU mapping may alter the outcome, directly
		 * access __apicid_to_node[].
		 */
		int ht_nodeid = c->initial_apicid;

		if (ht_nodeid >= 0 &&
		    __apicid_to_node[ht_nodeid] != NUMA_NO_NODE)
			node = __apicid_to_node[ht_nodeid];
		/* Pick a nearby node */
		if (!node_online(node))
			node = nearby_node(apicid);
	}
	numa_set_node(cpu, node);
#endif
}

static void __cpuinit early_init_amd_mc(struct cpuinfo_x86 *c)
{
#ifdef CONFIG_X86_HT
	unsigned bits, ecx;

	/* Multi core CPU? */
	if (c->extended_cpuid_level < 0x80000008)
		return;

	ecx = cpuid_ecx(0x80000008);

	c->x86_max_cores = (ecx & 0xff) + 1;

	/* CPU telling us the core id bits shift? */
	bits = (ecx >> 12) & 0xF;

	/* Otherwise recompute */
	if (bits == 0) {
		while ((1 << bits) < c->x86_max_cores)
			bits++;
	}

	c->x86_coreid_bits = bits;
#endif
}

static void __cpuinit early_init_amd(struct cpuinfo_x86 *c)
{
	early_init_amd_mc(c);

	/*
	 * c->x86_power is 8000_0007 edx. Bit 8 is TSC runs at constant rate
	 * with P/T states and does not stop in deep C-states
	 */
	if (c->x86_power & (1 << 8)) {
		set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
		set_cpu_cap(c, X86_FEATURE_NONSTOP_TSC);
	}

#ifdef CONFIG_X86_64
	set_cpu_cap(c, X86_FEATURE_SYSCALL32);
#else
	/*  Set MTRR capability flag if appropriate */
	if (c->x86 == 5)
		if (c->x86_model == 13 || c->x86_model == 9 ||
		    (c->x86_model == 8 && c->x86_mask >= 8))
			set_cpu_cap(c, X86_FEATURE_K6_MTRR);
#endif
#if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_PCI)
	/* check CPU config space for extended APIC ID */
	if (cpu_has_apic && c->x86 >= 0xf) {
		unsigned int val;
		val = read_pci_config(0, 24, 0, 0x68);
		if ((val & ((1 << 17) | (1 << 18))) == ((1 << 17) | (1 << 18)))
			set_cpu_cap(c, X86_FEATURE_EXTD_APICID);
	}
#endif

	/* We need to do the following only once */
	if (c != &boot_cpu_data)
		return;

	if (cpu_has(c, X86_FEATURE_CONSTANT_TSC)) {

		if (c->x86 > 0x10 ||
		    (c->x86 == 0x10 && c->x86_model >= 0x2)) {
			u64 val;

			rdmsrl(MSR_K7_HWCR, val);
			if (!(val & BIT(24)))
				printk(KERN_WARNING FW_BUG "TSC doesn't count "
					"with P0 frequency!\n");
		}
	}
}

static void __cpuinit init_amd(struct cpuinfo_x86 *c)
{
#ifdef CONFIG_SMP
	unsigned long long value;

	/*
	 * Disable TLB flush filter by setting HWCR.FFDIS on K8
	 * bit 6 of msr C001_0015
	 *
	 * Errata 63 for SH-B3 steppings
	 * Errata 122 for all steppings (F+ have it disabled by default)
	 */
	if (c->x86 == 0xf) {
		rdmsrl(MSR_K7_HWCR, value);
		value |= 1 << 6;
		wrmsrl(MSR_K7_HWCR, value);
	}
#endif

	early_init_amd(c);

	/*
	 * Bit 31 in normal CPUID used for nonstandard 3DNow ID;
	 * 3DNow is IDd by bit 31 in extended CPUID (1*32+31) anyway
	 */
	clear_cpu_cap(c, 0*32+31);

#ifdef CONFIG_X86_64
	/* On C+ stepping K8 rep microcode works well for copy/memset */
	if (c->x86 == 0xf) {
		u32 level;

		level = cpuid_eax(1);
		if ((level >= 0x0f48 && level < 0x0f50) || level >= 0x0f58)
			set_cpu_cap(c, X86_FEATURE_REP_GOOD);

		/*
		 * Some BIOSes incorrectly force this feature, but only K8
		 * revision D (model = 0x14) and later actually support it.
		 * (AMD Erratum #110, docId: 25759).
		 */
		if (c->x86_model < 0x14 && cpu_has(c, X86_FEATURE_LAHF_LM)) {
			u64 val;

			clear_cpu_cap(c, X86_FEATURE_LAHF_LM);
			if (!rdmsrl_amd_safe(0xc001100d, &val)) {
				val &= ~(1ULL << 32);
				wrmsrl_amd_safe(0xc001100d, val);
			}
		}

	}
	if (c->x86 >= 0x10)
		set_cpu_cap(c, X86_FEATURE_REP_GOOD);

	/* get apicid instead of initial apic id from cpuid */
	c->apicid = hard_smp_processor_id();
#else

	/*
	 *	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)
	 */

	switch (c->x86) {
	case 4:
		init_amd_k5(c);
		break;
	case 5:
		init_amd_k6(c);
		break;
	case 6: /* An Athlon/Duron */
		init_amd_k7(c);
		break;
	}

	/* K6s reports MCEs but don't actually have all the MSRs */
	if (c->x86 < 6)
		clear_cpu_cap(c, X86_FEATURE_MCE);
#endif

	/* Enable workaround for FXSAVE leak */
	if (c->x86 >= 6)
		set_cpu_cap(c, X86_FEATURE_FXSAVE_LEAK);

	if (!c->x86_model_id[0]) {
		switch (c->x86) {
		case 0xf:
			/* Should distinguish Models here, but this is only
			   a fallback anyways. */
			strcpy(c->x86_model_id, "Hammer");
			break;
		}
	}

	cpu_detect_cache_sizes(c);

	/* Multi core CPU? */
	if (c->extended_cpuid_level >= 0x80000008) {
		amd_detect_cmp(c);
		srat_detect_node(c);
	}

#ifdef CONFIG_X86_32
	detect_ht(c);
#endif

	if (c->extended_cpuid_level >= 0x80000006) {
		if (cpuid_edx(0x80000006) & 0xf000)
			num_cache_leaves = 4;
		else
			num_cache_leaves = 3;
	}

	if (c->x86 >= 0xf)
		set_cpu_cap(c, X86_FEATURE_K8);

	if (cpu_has_xmm2) {
		/* MFENCE stops RDTSC speculation */
		set_cpu_cap(c, X86_FEATURE_MFENCE_RDTSC);
	}

#ifdef CONFIG_X86_64
	if (c->x86 == 0x10) {
		/* do this for boot cpu */
		if (c == &boot_cpu_data)
			check_enable_amd_mmconf_dmi();

		fam10h_check_enable_mmcfg();
	}

	if (c == &boot_cpu_data && c->x86 >= 0xf) {
		unsigned long long tseg;

		/*
		 * Split up direct mapping around the TSEG SMM area.
		 * Don't do it for gbpages because there seems very little
		 * benefit in doing so.
		 */
		if (!rdmsrl_safe(MSR_K8_TSEG_ADDR, &tseg)) {
			printk(KERN_DEBUG "tseg: %010llx\n", tseg);
			if ((tseg>>PMD_SHIFT) <
				(max_low_pfn_mapped>>(PMD_SHIFT-PAGE_SHIFT)) ||
				((tseg>>PMD_SHIFT) <
				(max_pfn_mapped>>(PMD_SHIFT-PAGE_SHIFT)) &&
				(tseg>>PMD_SHIFT) >= (1ULL<<(32 - PMD_SHIFT))))
				set_memory_4k((unsigned long)__va(tseg), 1);
		}
	}
#endif
}

#ifdef CONFIG_X86_32
static unsigned int __cpuinit amd_size_cache(struct cpuinfo_x86 *c,
							unsigned int size)
{
	/* AMD errata T13 (order #21922) */
	if ((c->x86 == 6)) {
		/* Duron Rev A0 */
		if (c->x86_model == 3 && c->x86_mask == 0)
			size = 64;
		/* Tbird rev A1/A2 */
		if (c->x86_model == 4 &&
			(c->x86_mask == 0 || c->x86_mask == 1))
			size = 256;
	}
	return size;
}
#endif

static const struct cpu_dev __cpuinitconst amd_cpu_dev = {
	.c_vendor	= "AMD",
	.c_ident	= { "AuthenticAMD" },
#ifdef CONFIG_X86_32
	.c_models = {
		{ .vendor = X86_VENDOR_AMD, .family = 4, .model_names =
		  {
			  [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_size_cache	= amd_size_cache,
#endif
	.c_early_init   = early_init_amd,
	.c_init		= init_amd,
	.c_x86_vendor	= X86_VENDOR_AMD,
};

cpu_dev_register(amd_cpu_dev);

/*
 * AMD errata checking
 *
 * Errata are defined as arrays of ints using the AMD_LEGACY_ERRATUM() or
 * AMD_OSVW_ERRATUM() macros. The latter is intended for newer errata that
 * have an OSVW id assigned, which it takes as first argument. Both take a
 * variable number of family-specific model-stepping ranges created by
 * AMD_MODEL_RANGE(). Each erratum also has to be declared as extern const
 * int[] in arch/x86/include/asm/processor.h.
 *
 * Example:
 *
 * const int amd_erratum_319[] =
 *	AMD_LEGACY_ERRATUM(AMD_MODEL_RANGE(0x10, 0x2, 0x1, 0x4, 0x2),
 *			   AMD_MODEL_RANGE(0x10, 0x8, 0x0, 0x8, 0x0),
 *			   AMD_MODEL_RANGE(0x10, 0x9, 0x0, 0x9, 0x0));
 */

const int amd_erratum_400[] =
	AMD_OSVW_ERRATUM(1, AMD_MODEL_RANGE(0xf, 0x41, 0x2, 0xff, 0xf),
			    AMD_MODEL_RANGE(0x10, 0x2, 0x1, 0xff, 0xf));
EXPORT_SYMBOL_GPL(amd_erratum_400);

const int amd_erratum_383[] =
	AMD_OSVW_ERRATUM(3, AMD_MODEL_RANGE(0x10, 0, 0, 0xff, 0xf));
EXPORT_SYMBOL_GPL(amd_erratum_383);

bool cpu_has_amd_erratum(const int *erratum)
{
	struct cpuinfo_x86 *cpu = __this_cpu_ptr(&cpu_info);
	int osvw_id = *erratum++;
	u32 range;
	u32 ms;

	/*
	 * If called early enough that current_cpu_data hasn't been initialized
	 * yet, fall back to boot_cpu_data.
	 */
	if (cpu->x86 == 0)
		cpu = &boot_cpu_data;

	if (cpu->x86_vendor != X86_VENDOR_AMD)
		return false;

	if (osvw_id >= 0 && osvw_id < 65536 &&
	    cpu_has(cpu, X86_FEATURE_OSVW)) {
		u64 osvw_len;

		rdmsrl(MSR_AMD64_OSVW_ID_LENGTH, osvw_len);
		if (osvw_id < osvw_len) {
			u64 osvw_bits;

			rdmsrl(MSR_AMD64_OSVW_STATUS + (osvw_id >> 6),
			    osvw_bits);
			return osvw_bits & (1ULL << (osvw_id & 0x3f));
		}
	}

	/* OSVW unavailable or ID unknown, match family-model-stepping range */
	ms = (cpu->x86_model << 4) | cpu->x86_mask;
	while ((range = *erratum++))
		if ((cpu->x86 == AMD_MODEL_RANGE_FAMILY(range)) &&
		    (ms >= AMD_MODEL_RANGE_START(range)) &&
		    (ms <= AMD_MODEL_RANGE_END(range)))
			return true;

	return false;
}

EXPORT_SYMBOL_GPL(cpu_has_amd_erratum);
Commit	Line	Data
	1	#include <linux/init.h>
	2	#include <linux/bitops.h>
	3	#include <linux/mm.h>
	4
	5	#include <linux/io.h>
	6	#include <asm/processor.h>
	7	#include <asm/apic.h>
	8	#include <asm/cpu.h>
	9	#include <asm/pci-direct.h>
	10
	11	#ifdef CONFIG_X86_64
	12	# include <asm/numa_64.h>
	13	# include <asm/mmconfig.h>
	14	# include <asm/cacheflush.h>
	15	#endif
	16
	17	#include "cpu.h"
	18
	19	#ifdef CONFIG_X86_32
	20	/*
	21	* B step AMD K6 before B 9730xxxx have hardware bugs that can cause
	22	* misexecution of code under Linux. Owners of such processors should
	23	* contact AMD for precise details and a CPU swap.
	24	*
	25	* See http://www.multimania.com/poulot/k6bug.html
	26	* http://www.amd.com/K6/k6docs/revgd.html
	27	*
	28	* The following test is erm.. interesting. AMD neglected to up
	29	* the chip setting when fixing the bug but they also tweaked some
	30	* performance at the same time..
	31	*/
	32
	33	extern void vide(void);
	34	__asm__(".align 4\nvide: ret");
	35
	36	static void __cpuinit init_amd_k5(struct cpuinfo_x86 *c)
	37	{
	38	/*
	39	* General Systems BIOSen alias the cpu frequency registers
	40	* of the Elan at 0x000df000. Unfortuantly, one of the Linux
	41	* drivers subsequently pokes it, and changes the CPU speed.
	42	* Workaround : Remove the unneeded alias.
	43	*/
	44	#define CBAR (0xfffc) /* Configuration Base Address (32-bit) */
	45	#define CBAR_ENB (0x80000000)
	46	#define CBAR_KEY (0X000000CB)
	47	if (c->x86_model == 9 \|\| c->x86_model == 10) {
	48	if (inl(CBAR) & CBAR_ENB)
	49	outl(0 \| CBAR_KEY, CBAR);
	50	}
	51	}
	52
	53
	54	static void __cpuinit init_amd_k6(struct cpuinfo_x86 *c)
	55	{
	56	u32 l, h;
	57	int mbytes = num_physpages >> (20-PAGE_SHIFT);
	58
	59	if (c->x86_model < 6) {
	60	/* Based on AMD doc 20734R - June 2000 */
	61	if (c->x86_model == 0) {
	62	clear_cpu_cap(c, X86_FEATURE_APIC);
	63	set_cpu_cap(c, X86_FEATURE_PGE);
	64	}
	65	return;
	66	}
	67
	68	if (c->x86_model == 6 && c->x86_mask == 1) {
	69	const int K6_BUG_LOOP = 1000000;
	70	int n;
	71	void (*f_vide)(void);
	72	unsigned long d, d2;
	73
	74	printk(KERN_INFO "AMD K6 stepping B detected - ");
	75
	76	/*
	77	* It looks like AMD fixed the 2.6.2 bug and improved indirect
	78	* calls at the same time.
	79	*/
	80
	81	n = K6_BUG_LOOP;
	82	f_vide = vide;
	83	rdtscl(d);
	84	while (n--)
	85	f_vide();
	86	rdtscl(d2);
	87	d = d2-d;
	88
	89	if (d > 20*K6_BUG_LOOP)
	90	printk(KERN_CONT
	91	"system stability may be impaired when more than 32 MB are used.\n");
	92	else
	93	printk(KERN_CONT "probably OK (after B9730xxxx).\n");
	94	printk(KERN_INFO "Please see http://membres.lycos.fr/poulot/k6bug.html\n");
	95	}
	96
	97	/* K6 with old style WHCR */
	98	if (c->x86_model < 8 \|\|
	99	(c->x86_model == 8 && c->x86_mask < 8)) {
	100	/* We can only write allocate on the low 508Mb */
	101	if (mbytes > 508)
	102	mbytes = 508;
	103
	104	rdmsr(MSR_K6_WHCR, l, h);
	105	if ((l&0x0000FFFF) == 0) {
	106	unsigned long flags;
	107	l = (1<<0)\|((mbytes/4)<<1);
	108	local_irq_save(flags);
	109	wbinvd();
	110	wrmsr(MSR_K6_WHCR, l, h);
	111	local_irq_restore(flags);
	112	printk(KERN_INFO "Enabling old style K6 write allocation for %d Mb\n",
	113	mbytes);
	114	}
	115	return;
	116	}
	117
	118	if ((c->x86_model == 8 && c->x86_mask > 7) \|\|
	119	c->x86_model == 9 \|\| c->x86_model == 13) {
	120	/* The more serious chips .. */
	121
	122	if (mbytes > 4092)
	123	mbytes = 4092;
	124
	125	rdmsr(MSR_K6_WHCR, l, h);
	126	if ((l&0xFFFF0000) == 0) {
	127	unsigned long flags;
	128	l = ((mbytes>>2)<<22)\|(1<<16);
	129	local_irq_save(flags);
	130	wbinvd();
	131	wrmsr(MSR_K6_WHCR, l, h);
	132	local_irq_restore(flags);
	133	printk(KERN_INFO "Enabling new style K6 write allocation for %d Mb\n",
	134	mbytes);
	135	}
	136
	137	return;
	138	}
	139
	140	if (c->x86_model == 10) {
	141	/* AMD Geode LX is model 10 */
	142	/* placeholder for any needed mods */
	143	return;
	144	}
	145	}
	146
	147	static void __cpuinit amd_k7_smp_check(struct cpuinfo_x86 *c)
	148	{
	149	#ifdef CONFIG_SMP
	150	/* calling is from identify_secondary_cpu() ? */
	151	if (!c->cpu_index)
	152	return;
	153
	154	/*
	155	* Certain Athlons might work (for various values of 'work') in SMP
	156	* but they are not certified as MP capable.
	157	*/
	158	/* Athlon 660/661 is valid. */
	159	if ((c->x86_model == 6) && ((c->x86_mask == 0) \|\|
	160	(c->x86_mask == 1)))
	161	goto valid_k7;
	162
	163	/* Duron 670 is valid */
	164	if ((c->x86_model == 7) && (c->x86_mask == 0))
	165	goto valid_k7;
	166
	167	/*
	168	* Athlon 662, Duron 671, and Athlon >model 7 have capability
	169	* bit. It's worth noting that the A5 stepping (662) of some
	170	* Athlon XP's have the MP bit set.
	171	* See http://www.heise.de/newsticker/data/jow-18.10.01-000 for
	172	* more.
	173	*/
	174	if (((c->x86_model == 6) && (c->x86_mask >= 2)) \|\|
	175	((c->x86_model == 7) && (c->x86_mask >= 1)) \|\|
	176	(c->x86_model > 7))
	177	if (cpu_has_mp)
	178	goto valid_k7;
	179
	180	/* If we get here, not a certified SMP capable AMD system. */
	181
	182	/*
	183	* Don't taint if we are running SMP kernel on a single non-MP
	184	* approved Athlon
	185	*/
	186	WARN_ONCE(1, "WARNING: This combination of AMD"
	187	" processors is not suitable for SMP.\n");
	188	if (!test_taint(TAINT_UNSAFE_SMP))
	189	add_taint(TAINT_UNSAFE_SMP);
	190
	191	valid_k7:
	192	;
	193	#endif
	194	}
	195
	196	static void __cpuinit init_amd_k7(struct cpuinfo_x86 *c)
	197	{
	198	u32 l, h;
	199
	200	/*
	201	* Bit 15 of Athlon specific MSR 15, needs to be 0
	202	* to enable SSE on Palomino/Morgan/Barton CPU's.
	203	* If the BIOS didn't enable it already, enable it here.
	204	*/
	205	if (c->x86_model >= 6 && c->x86_model <= 10) {
	206	if (!cpu_has(c, X86_FEATURE_XMM)) {
	207	printk(KERN_INFO "Enabling disabled K7/SSE Support.\n");
	208	rdmsr(MSR_K7_HWCR, l, h);
	209	l &= ~0x00008000;
	210	wrmsr(MSR_K7_HWCR, l, h);
	211	set_cpu_cap(c, X86_FEATURE_XMM);
	212	}
	213	}
	214
	215	/*
	216	* It's been determined by AMD that Athlons since model 8 stepping 1
	217	* are more robust with CLK_CTL set to 200xxxxx instead of 600xxxxx
	218	* As per AMD technical note 27212 0.2
	219	*/
	220	if ((c->x86_model == 8 && c->x86_mask >= 1) \|\| (c->x86_model > 8)) {
	221	rdmsr(MSR_K7_CLK_CTL, l, h);
	222	if ((l & 0xfff00000) != 0x20000000) {
	223	printk(KERN_INFO
	224	"CPU: CLK_CTL MSR was %x. Reprogramming to %x\n",
	225	l, ((l & 0x000fffff)\|0x20000000));
	226	wrmsr(MSR_K7_CLK_CTL, (l & 0x000fffff)\|0x20000000, h);
	227	}
	228	}
	229
	230	set_cpu_cap(c, X86_FEATURE_K7);
	231
	232	amd_k7_smp_check(c);
	233	}
	234	#endif
	235
	236	#ifdef CONFIG_NUMA
	237	/*
	238	* To workaround broken NUMA config. Read the comment in
	239	* srat_detect_node().
	240	*/
	241	static int __cpuinit nearby_node(int apicid)
	242	{
	243	int i, node;
	244
	245	for (i = apicid - 1; i >= 0; i--) {
	246	node = __apicid_to_node[i];
	247	if (node != NUMA_NO_NODE && node_online(node))
	248	return node;
	249	}
	250	for (i = apicid + 1; i < MAX_LOCAL_APIC; i++) {
	251	node = __apicid_to_node[i];
	252	if (node != NUMA_NO_NODE && node_online(node))
	253	return node;
	254	}
	255	return first_node(node_online_map); /* Shouldn't happen */
	256	}
	257	#endif
	258
	259	/*
	260	* Fixup core topology information for
	261	* (1) AMD multi-node processors
	262	* Assumption: Number of cores in each internal node is the same.
	263	* (2) AMD processors supporting compute units
	264	*/
	265	#ifdef CONFIG_X86_HT
	266	static void __cpuinit amd_get_topology(struct cpuinfo_x86 *c)
	267	{
	268	u32 nodes, cores_per_cu = 1;
	269	u8 node_id;
	270	int cpu = smp_processor_id();
	271
	272	/* get information required for multi-node processors */
	273	if (cpu_has(c, X86_FEATURE_TOPOEXT)) {
	274	u32 eax, ebx, ecx, edx;
	275
	276	cpuid(0x8000001e, &eax, &ebx, &ecx, &edx);
	277	nodes = ((ecx >> 8) & 7) + 1;
	278	node_id = ecx & 7;
	279
	280	/* get compute unit information */
	281	smp_num_siblings = ((ebx >> 8) & 3) + 1;
	282	c->compute_unit_id = ebx & 0xff;
	283	cores_per_cu += ((ebx >> 8) & 3);
	284	} else if (cpu_has(c, X86_FEATURE_NODEID_MSR)) {
	285	u64 value;
	286
	287	rdmsrl(MSR_FAM10H_NODE_ID, value);
	288	nodes = ((value >> 3) & 7) + 1;
	289	node_id = value & 7;
	290	} else
	291	return;
	292
	293	/* fixup multi-node processor information */
	294	if (nodes > 1) {
	295	u32 cores_per_node;
	296	u32 cus_per_node;
	297
	298	set_cpu_cap(c, X86_FEATURE_AMD_DCM);
	299	cores_per_node = c->x86_max_cores / nodes;
	300	cus_per_node = cores_per_node / cores_per_cu;
	301
	302	/* store NodeID, use llc_shared_map to store sibling info */
	303	per_cpu(cpu_llc_id, cpu) = node_id;
	304
	305	/* core id has to be in the [0 .. cores_per_node - 1] range */
	306	c->cpu_core_id %= cores_per_node;
	307	c->compute_unit_id %= cus_per_node;
	308	}
	309	}
	310	#endif
	311
	312	/*
	313	* On a AMD dual core setup the lower bits of the APIC id distingush the cores.
	314	* Assumes number of cores is a power of two.
	315	*/
	316	static void __cpuinit amd_detect_cmp(struct cpuinfo_x86 *c)
	317	{
	318	#ifdef CONFIG_X86_HT
	319	unsigned bits;
	320	int cpu = smp_processor_id();
	321
	322	bits = c->x86_coreid_bits;
	323	/* Low order bits define the core id (index of core in socket) */
	324	c->cpu_core_id = c->initial_apicid & ((1 << bits)-1);
	325	/* Convert the initial APIC ID into the socket ID */
	326	c->phys_proc_id = c->initial_apicid >> bits;
	327	/* use socket ID also for last level cache */
	328	per_cpu(cpu_llc_id, cpu) = c->phys_proc_id;
	329	amd_get_topology(c);
	330	#endif
	331	}
	332
	333	int amd_get_nb_id(int cpu)
	334	{
	335	int id = 0;
	336	#ifdef CONFIG_SMP
	337	id = per_cpu(cpu_llc_id, cpu);
	338	#endif
	339	return id;
	340	}
	341	EXPORT_SYMBOL_GPL(amd_get_nb_id);
	342
	343	static void __cpuinit srat_detect_node(struct cpuinfo_x86 *c)
	344	{
	345	#ifdef CONFIG_NUMA
	346	int cpu = smp_processor_id();
	347	int node;
	348	unsigned apicid = c->apicid;
	349
	350	node = numa_cpu_node(cpu);
	351	if (node == NUMA_NO_NODE)
	352	node = per_cpu(cpu_llc_id, cpu);
	353
	354	if (!node_online(node)) {
	355	/*
	356	* Two possibilities here:
	357	*
	358	* - The CPU is missing memory and no node was created. In
	359	* that case try picking one from a nearby CPU.
	360	*
	361	* - The APIC IDs differ from the HyperTransport node IDs
	362	* which the K8 northbridge parsing fills in. Assume
	363	* they are all increased by a constant offset, but in
	364	* the same order as the HT nodeids. If that doesn't
	365	* result in a usable node fall back to the path for the
	366	* previous case.
	367	*
	368	* This workaround operates directly on the mapping between
	369	* APIC ID and NUMA node, assuming certain relationship
	370	* between APIC ID, HT node ID and NUMA topology. As going
	371	* through CPU mapping may alter the outcome, directly
	372	* access __apicid_to_node[].
	373	*/
	374	int ht_nodeid = c->initial_apicid;
	375
	376	if (ht_nodeid >= 0 &&
	377	__apicid_to_node[ht_nodeid] != NUMA_NO_NODE)
	378	node = __apicid_to_node[ht_nodeid];
	379	/* Pick a nearby node */
	380	if (!node_online(node))
	381	node = nearby_node(apicid);
	382	}
	383	numa_set_node(cpu, node);
	384	#endif
	385	}
	386
	387	static void __cpuinit early_init_amd_mc(struct cpuinfo_x86 *c)
	388	{
	389	#ifdef CONFIG_X86_HT
	390	unsigned bits, ecx;
	391
	392	/* Multi core CPU? */
	393	if (c->extended_cpuid_level < 0x80000008)
	394	return;
	395
	396	ecx = cpuid_ecx(0x80000008);
	397
	398	c->x86_max_cores = (ecx & 0xff) + 1;
	399
	400	/* CPU telling us the core id bits shift? */
	401	bits = (ecx >> 12) & 0xF;
	402
	403	/* Otherwise recompute */
	404	if (bits == 0) {
	405	while ((1 << bits) < c->x86_max_cores)
	406	bits++;
	407	}
	408
	409	c->x86_coreid_bits = bits;
	410	#endif
	411	}
	412
	413	static void __cpuinit early_init_amd(struct cpuinfo_x86 *c)
	414	{
	415	early_init_amd_mc(c);
	416
	417	/*
	418	* c->x86_power is 8000_0007 edx. Bit 8 is TSC runs at constant rate
	419	* with P/T states and does not stop in deep C-states
	420	*/
	421	if (c->x86_power & (1 << 8)) {
	422	set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
	423	set_cpu_cap(c, X86_FEATURE_NONSTOP_TSC);
	424	}
	425
	426	#ifdef CONFIG_X86_64
	427	set_cpu_cap(c, X86_FEATURE_SYSCALL32);
	428	#else
	429	/* Set MTRR capability flag if appropriate */
	430	if (c->x86 == 5)
	431	if (c->x86_model == 13 \|\| c->x86_model == 9 \|\|
	432	(c->x86_model == 8 && c->x86_mask >= 8))
	433	set_cpu_cap(c, X86_FEATURE_K6_MTRR);
	434	#endif
	435	#if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_PCI)
	436	/* check CPU config space for extended APIC ID */
	437	if (cpu_has_apic && c->x86 >= 0xf) {
	438	unsigned int val;
	439	val = read_pci_config(0, 24, 0, 0x68);
	440	if ((val & ((1 << 17) \| (1 << 18))) == ((1 << 17) \| (1 << 18)))
	441	set_cpu_cap(c, X86_FEATURE_EXTD_APICID);
	442	}
	443	#endif
	444
	445	/* We need to do the following only once */
	446	if (c != &boot_cpu_data)
	447	return;
	448
	449	if (cpu_has(c, X86_FEATURE_CONSTANT_TSC)) {
	450
	451	if (c->x86 > 0x10 \|\|
	452	(c->x86 == 0x10 && c->x86_model >= 0x2)) {
	453	u64 val;
	454
	455	rdmsrl(MSR_K7_HWCR, val);
	456	if (!(val & BIT(24)))
	457	printk(KERN_WARNING FW_BUG "TSC doesn't count "
	458	"with P0 frequency!\n");
	459	}
	460	}
	461	}
	462
	463	static void __cpuinit init_amd(struct cpuinfo_x86 *c)
	464	{
	465	#ifdef CONFIG_SMP
	466	unsigned long long value;
	467
	468	/*
	469	* Disable TLB flush filter by setting HWCR.FFDIS on K8
	470	* bit 6 of msr C001_0015
	471	*
	472	* Errata 63 for SH-B3 steppings
	473	* Errata 122 for all steppings (F+ have it disabled by default)
	474	*/
	475	if (c->x86 == 0xf) {
	476	rdmsrl(MSR_K7_HWCR, value);
	477	value \|= 1 << 6;
	478	wrmsrl(MSR_K7_HWCR, value);
	479	}
	480	#endif
	481
	482	early_init_amd(c);
	483
	484	/*
	485	* Bit 31 in normal CPUID used for nonstandard 3DNow ID;
	486	* 3DNow is IDd by bit 31 in extended CPUID (1*32+31) anyway
	487	*/
	488	clear_cpu_cap(c, 0*32+31);
	489
	490	#ifdef CONFIG_X86_64
	491	/* On C+ stepping K8 rep microcode works well for copy/memset */
	492	if (c->x86 == 0xf) {
	493	u32 level;
	494
	495	level = cpuid_eax(1);
	496	if ((level >= 0x0f48 && level < 0x0f50) \|\| level >= 0x0f58)
	497	set_cpu_cap(c, X86_FEATURE_REP_GOOD);
	498
	499	/*
	500	* Some BIOSes incorrectly force this feature, but only K8
	501	* revision D (model = 0x14) and later actually support it.
	502	* (AMD Erratum #110, docId: 25759).
	503	*/
	504	if (c->x86_model < 0x14 && cpu_has(c, X86_FEATURE_LAHF_LM)) {
	505	u64 val;
	506
	507	clear_cpu_cap(c, X86_FEATURE_LAHF_LM);
	508	if (!rdmsrl_amd_safe(0xc001100d, &val)) {
	509	val &= ~(1ULL << 32);
	510	wrmsrl_amd_safe(0xc001100d, val);
	511	}
	512	}
	513
	514	}
	515	if (c->x86 >= 0x10)
	516	set_cpu_cap(c, X86_FEATURE_REP_GOOD);
	517
	518	/* get apicid instead of initial apic id from cpuid */
	519	c->apicid = hard_smp_processor_id();
	520	#else
	521
	522	/*
	523	* FIXME: We should handle the K5 here. Set up the write
	524	* range and also turn on MSR 83 bits 4 and 31 (write alloc,
	525	* no bus pipeline)
	526	*/
	527
	528	switch (c->x86) {
	529	case 4:
	530	init_amd_k5(c);
	531	break;
	532	case 5:
	533	init_amd_k6(c);
	534	break;
	535	case 6: /* An Athlon/Duron */
	536	init_amd_k7(c);
	537	break;
	538	}
	539
	540	/* K6s reports MCEs but don't actually have all the MSRs */
	541	if (c->x86 < 6)
	542	clear_cpu_cap(c, X86_FEATURE_MCE);
	543	#endif
	544
	545	/* Enable workaround for FXSAVE leak */
	546	if (c->x86 >= 6)
	547	set_cpu_cap(c, X86_FEATURE_FXSAVE_LEAK);
	548
	549	if (!c->x86_model_id[0]) {
	550	switch (c->x86) {
	551	case 0xf:
	552	/* Should distinguish Models here, but this is only
	553	a fallback anyways. */
	554	strcpy(c->x86_model_id, "Hammer");
	555	break;
	556	}
	557	}
	558
	559	cpu_detect_cache_sizes(c);
	560
	561	/* Multi core CPU? */
	562	if (c->extended_cpuid_level >= 0x80000008) {
	563	amd_detect_cmp(c);
	564	srat_detect_node(c);
	565	}
	566
	567	#ifdef CONFIG_X86_32
	568	detect_ht(c);
	569	#endif
	570
	571	if (c->extended_cpuid_level >= 0x80000006) {
	572	if (cpuid_edx(0x80000006) & 0xf000)
	573	num_cache_leaves = 4;
	574	else
	575	num_cache_leaves = 3;
	576	}
	577
	578	if (c->x86 >= 0xf)
	579	set_cpu_cap(c, X86_FEATURE_K8);
	580
	581	if (cpu_has_xmm2) {
	582	/* MFENCE stops RDTSC speculation */
	583	set_cpu_cap(c, X86_FEATURE_MFENCE_RDTSC);
	584	}
	585
	586	#ifdef CONFIG_X86_64
	587	if (c->x86 == 0x10) {
	588	/* do this for boot cpu */
	589	if (c == &boot_cpu_data)
	590	check_enable_amd_mmconf_dmi();
	591
	592	fam10h_check_enable_mmcfg();
	593	}
	594
	595	if (c == &boot_cpu_data && c->x86 >= 0xf) {
	596	unsigned long long tseg;
	597
	598	/*
	599	* Split up direct mapping around the TSEG SMM area.
	600	* Don't do it for gbpages because there seems very little
	601	* benefit in doing so.
	602	*/
	603	if (!rdmsrl_safe(MSR_K8_TSEG_ADDR, &tseg)) {
	604	printk(KERN_DEBUG "tseg: %010llx\n", tseg);
	605	if ((tseg>>PMD_SHIFT) <
	606	(max_low_pfn_mapped>>(PMD_SHIFT-PAGE_SHIFT)) \|\|
	607	((tseg>>PMD_SHIFT) <
	608	(max_pfn_mapped>>(PMD_SHIFT-PAGE_SHIFT)) &&
	609	(tseg>>PMD_SHIFT) >= (1ULL<<(32 - PMD_SHIFT))))
	610	set_memory_4k((unsigned long)__va(tseg), 1);
	611	}
	612	}
	613	#endif
	614	}
	615
	616	#ifdef CONFIG_X86_32
	617	static unsigned int __cpuinit amd_size_cache(struct cpuinfo_x86 *c,
	618	unsigned int size)
	619	{
	620	/* AMD errata T13 (order #21922) */
	621	if ((c->x86 == 6)) {
	622	/* Duron Rev A0 */
	623	if (c->x86_model == 3 && c->x86_mask == 0)
	624	size = 64;
	625	/* Tbird rev A1/A2 */
	626	if (c->x86_model == 4 &&
	627	(c->x86_mask == 0 \|\| c->x86_mask == 1))
	628	size = 256;
	629	}
	630	return size;
	631	}
	632	#endif
	633
	634	static const struct cpu_dev __cpuinitconst amd_cpu_dev = {
	635	.c_vendor = "AMD",
	636	.c_ident = { "AuthenticAMD" },
	637	#ifdef CONFIG_X86_32
	638	.c_models = {
	639	{ .vendor = X86_VENDOR_AMD, .family = 4, .model_names =
	640	{
	641	[3] = "486 DX/2",
	642	[7] = "486 DX/2-WB",
	643	[8] = "486 DX/4",
	644	[9] = "486 DX/4-WB",
	645	[14] = "Am5x86-WT",
	646	[15] = "Am5x86-WB"
	647	}
	648	},
	649	},
	650	.c_size_cache = amd_size_cache,
	651	#endif
	652	.c_early_init = early_init_amd,
	653	.c_init = init_amd,
	654	.c_x86_vendor = X86_VENDOR_AMD,
	655	};
	656
	657	cpu_dev_register(amd_cpu_dev);
	658
	659	/*
	660	* AMD errata checking
	661	*
	662	* Errata are defined as arrays of ints using the AMD_LEGACY_ERRATUM() or
	663	* AMD_OSVW_ERRATUM() macros. The latter is intended for newer errata that
	664	* have an OSVW id assigned, which it takes as first argument. Both take a
	665	* variable number of family-specific model-stepping ranges created by
	666	* AMD_MODEL_RANGE(). Each erratum also has to be declared as extern const
	667	* int[] in arch/x86/include/asm/processor.h.
	668	*
	669	* Example:
	670	*
	671	* const int amd_erratum_319[] =
	672	* AMD_LEGACY_ERRATUM(AMD_MODEL_RANGE(0x10, 0x2, 0x1, 0x4, 0x2),
	673	* AMD_MODEL_RANGE(0x10, 0x8, 0x0, 0x8, 0x0),
	674	* AMD_MODEL_RANGE(0x10, 0x9, 0x0, 0x9, 0x0));
	675	*/
	676
	677	const int amd_erratum_400[] =
	678	AMD_OSVW_ERRATUM(1, AMD_MODEL_RANGE(0xf, 0x41, 0x2, 0xff, 0xf),
	679	AMD_MODEL_RANGE(0x10, 0x2, 0x1, 0xff, 0xf));
	680	EXPORT_SYMBOL_GPL(amd_erratum_400);
	681
	682	const int amd_erratum_383[] =
	683	AMD_OSVW_ERRATUM(3, AMD_MODEL_RANGE(0x10, 0, 0, 0xff, 0xf));
	684	EXPORT_SYMBOL_GPL(amd_erratum_383);
	685
	686	bool cpu_has_amd_erratum(const int *erratum)
	687	{
	688	struct cpuinfo_x86 *cpu = __this_cpu_ptr(&cpu_info);
	689	int osvw_id = *erratum++;
	690	u32 range;
	691	u32 ms;
	692
	693	/*
	694	* If called early enough that current_cpu_data hasn't been initialized
	695	* yet, fall back to boot_cpu_data.
	696	*/
	697	if (cpu->x86 == 0)
	698	cpu = &boot_cpu_data;
	699
	700	if (cpu->x86_vendor != X86_VENDOR_AMD)
	701	return false;
	702
	703	if (osvw_id >= 0 && osvw_id < 65536 &&
	704	cpu_has(cpu, X86_FEATURE_OSVW)) {
	705	u64 osvw_len;
	706
	707	rdmsrl(MSR_AMD64_OSVW_ID_LENGTH, osvw_len);
	708	if (osvw_id < osvw_len) {
	709	u64 osvw_bits;
	710
	711	rdmsrl(MSR_AMD64_OSVW_STATUS + (osvw_id >> 6),
	712	osvw_bits);
	713	return osvw_bits & (1ULL << (osvw_id & 0x3f));
	714	}
	715	}
	716
	717	/* OSVW unavailable or ID unknown, match family-model-stepping range */
	718	ms = (cpu->x86_model << 4) \| cpu->x86_mask;
	719	while ((range = *erratum++))
	720	if ((cpu->x86 == AMD_MODEL_RANGE_FAMILY(range)) &&
	721	(ms >= AMD_MODEL_RANGE_START(range)) &&
	722	(ms <= AMD_MODEL_RANGE_END(range)))
	723	return true;
	724
	725	return false;
	726	}
	727
	728	EXPORT_SYMBOL_GPL(cpu_has_amd_erratum);