| @@ -1,90 +1,173 @@ | | | @@ -1,90 +1,173 @@ |
1 | /* $NetBSD: cpu_topology.c,v 1.1 2009/04/30 00:07:23 rmind Exp $ */ | | 1 | /* $NetBSD: cpu_topology.c,v 1.2 2009/05/26 01:42:02 rmind Exp $ */ |
2 | | | 2 | |
3 | /*- | | 3 | /*- |
4 | * Copyright (c)2008 YAMAMOTO Takashi, | | 4 | * Copyright (c) 2009 Mindaugas Rasiukevicius <rmind at NetBSD org>, |
| | | 5 | * Copyright (c) 2008 YAMAMOTO Takashi, |
5 | * All rights reserved. | | 6 | * All rights reserved. |
6 | * | | 7 | * |
7 | * Redistribution and use in source and binary forms, with or without | | 8 | * Redistribution and use in source and binary forms, with or without |
8 | * modification, are permitted provided that the following conditions | | 9 | * modification, are permitted provided that the following conditions |
9 | * are met: | | 10 | * are met: |
10 | * 1. Redistributions of source code must retain the above copyright | | 11 | * 1. Redistributions of source code must retain the above copyright |
11 | * notice, this list of conditions and the following disclaimer. | | 12 | * notice, this list of conditions and the following disclaimer. |
12 | * 2. Redistributions in binary form must reproduce the above copyright | | 13 | * 2. Redistributions in binary form must reproduce the above copyright |
13 | * notice, this list of conditions and the following disclaimer in the | | 14 | * notice, this list of conditions and the following disclaimer in the |
14 | * documentation and/or other materials provided with the distribution. | | 15 | * documentation and/or other materials provided with the distribution. |
15 | * | | 16 | * |
16 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND | | 17 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND |
17 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE | | 18 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
18 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE | | 19 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
19 | * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE | | 20 | * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE |
20 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL | | 21 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
21 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS | | 22 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
22 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) | | 23 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
23 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT | | 24 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
24 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY | | 25 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
25 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF | | 26 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
26 | * SUCH DAMAGE. | | 27 | * SUCH DAMAGE. |
27 | */ | | 28 | */ |
28 | | | 29 | |
| | | 30 | /* |
| | | 31 | * x86 CPU topology detection. |
| | | 32 | * |
| | | 33 | * References: |
| | | 34 | * - 53668.pdf (7.10.2), 276613.pdf |
| | | 35 | * - 31116.pdf, 41256.pdf, 25481.pdf |
| | | 36 | */ |
| | | 37 | |
29 | #include <sys/cdefs.h> | | 38 | #include <sys/cdefs.h> |
30 | __KERNEL_RCSID(0, "$NetBSD: cpu_topology.c,v 1.1 2009/04/30 00:07:23 rmind Exp $"); | | 39 | __KERNEL_RCSID(0, "$NetBSD: cpu_topology.c,v 1.2 2009/05/26 01:42:02 rmind Exp $"); |
31 | | | 40 | |
32 | #include <sys/param.h> | | 41 | #include <sys/param.h> |
33 | #include <sys/bitops.h> | | 42 | #include <sys/bitops.h> |
34 | | | 43 | |
35 | #include <machine/specialreg.h> | | 44 | #include <machine/specialreg.h> |
36 | #include <machine/cpu.h> | | 45 | #include <machine/cpu.h> |
37 | | | 46 | |
38 | #include <x86/cpufunc.h> | | 47 | #include <x86/cpufunc.h> |
39 | #include <x86/cputypes.h> | | 48 | #include <x86/cputypes.h> |
40 | #include <x86/cpuvar.h> | | 49 | #include <x86/cpuvar.h> |
41 | | | 50 | |
42 | void | | 51 | void |
43 | x86_cpu_toplogy(struct cpu_info *ci) | | 52 | x86_cpu_toplogy(struct cpu_info *ci) |
44 | { | | 53 | { |
45 | u_int lp_max = 1; /* logical processors per package */ | | 54 | u_int lp_max; /* Logical processors per package */ |
46 | u_int smt_max; /* smt per core */ | | 55 | u_int core_max; /* Core per package */ |
47 | u_int core_max = 1; /* core per package */ | | 56 | int n, cpu_family, apic_id, smt_bits, core_bits = 0; |
48 | int smt_bits, core_bits; | | 57 | uint32_t descs[4], lextmode; |
49 | uint32_t descs[4]; | | 58 | |
| | | 59 | apic_id = ci->ci_initapicid; |
| | | 60 | cpu_family = CPUID2FAMILY(ci->ci_signature); |
| | | 61 | |
| | | 62 | /* Initial values. */ |
| | | 63 | ci->ci_packageid = apic_id; |
| | | 64 | ci->ci_coreid = 0; |
| | | 65 | ci->ci_smtid = 0; |
50 | | | 66 | |
51 | if (cpu_vendor != CPUVENDOR_INTEL || | | 67 | switch (cpu_vendor) { |
52 | CPUID2FAMILY(ci->ci_signature) < 6) | | 68 | case CPUVENDOR_INTEL: |
| | | 69 | if (cpu_family < 6) |
| | | 70 | return; |
| | | 71 | break; |
| | | 72 | case CPUVENDOR_AMD: |
| | | 73 | if (cpu_family < 0xf) |
| | | 74 | return; |
| | | 75 | break; |
| | | 76 | default: |
53 | return; | | 77 | return; |
| | | 78 | } |
54 | | | 79 | |
55 | /* Determine the extended feature flags. */ | | 80 | /* Determine the extended feature flags. */ |
56 | x86_cpuid(0x80000000, descs); | | 81 | x86_cpuid(0x80000000, descs); |
57 | if (descs[0] >= 0x80000001) { | | 82 | lextmode = descs[0]; |
| | | 83 | if (lextmode >= 0x80000001) { |
58 | x86_cpuid(0x80000001, descs); | | 84 | x86_cpuid(0x80000001, descs); |
59 | ci->ci_feature3_flags |= descs[3]; /* %edx */ | | 85 | ci->ci_feature3_flags |= descs[3]; /* edx */ |
60 | } | | 86 | } |
61 | | | 87 | |
62 | /* Determine topology. 253668.pdf 7.10.2. */ | | 88 | /* Check for HTT support. See notes below regarding AMD. */ |
63 | ci->ci_packageid = ci->ci_initapicid; | | | |
64 | ci->ci_coreid = 0; | | | |
65 | ci->ci_smtid = 0; | | | |
66 | if ((ci->ci_feature_flags & CPUID_HTT) != 0) { | | 89 | if ((ci->ci_feature_flags & CPUID_HTT) != 0) { |
| | | 90 | /* Maximum number of LPs sharing a cache (ebx[23:16]). */ |
67 | x86_cpuid(1, descs); | | 91 | x86_cpuid(1, descs); |
68 | lp_max = (descs[1] >> 16) & 0xff; | | 92 | lp_max = (descs[1] >> 16) & 0xff; |
| | | 93 | } else { |
| | | 94 | lp_max = 1; |
69 | } | | 95 | } |
70 | x86_cpuid(0, descs); | | 96 | |
71 | if (descs[0] >= 4) { | | 97 | switch (cpu_vendor) { |
72 | x86_cpuid2(4, 0, descs); | | 98 | case CPUVENDOR_INTEL: |
73 | core_max = (descs[0] >> 26) + 1; | | 99 | /* Check for leaf 4 support. */ |
| | | 100 | x86_cpuid(0, descs); |
| | | 101 | if (descs[0] >= 4) { |
| | | 102 | /* Maximum number of Cores per package (eax[31:26]). */ |
| | | 103 | x86_cpuid2(4, 0, descs); |
| | | 104 | core_max = (descs[0] >> 26) + 1; |
| | | 105 | } else { |
| | | 106 | core_max = 1; |
| | | 107 | } |
| | | 108 | break; |
| | | 109 | case CPUVENDOR_AMD: |
| | | 110 | /* In a case of AMD, HTT flag means CMP support. */ |
| | | 111 | if ((ci->ci_feature_flags & CPUID_HTT) == 0) { |
| | | 112 | core_max = 1; |
| | | 113 | break; |
| | | 114 | } |
| | | 115 | /* Legacy Method, LPs represent Cores. */ |
| | | 116 | if (cpu_family < 0x10 || lextmode < 0x80000008) { |
| | | 117 | core_max = lp_max; |
| | | 118 | break; |
| | | 119 | } |
| | | 120 | /* Number of Cores (NC) per package (ecx[7:0]). */ |
| | | 121 | x86_cpuid(0x80000008, descs); |
| | | 122 | core_max = (descs[2] & 0xff) + 1; |
| | | 123 | /* Amount of bits representing Core ID (ecx[15:12]). */ |
| | | 124 | n = (descs[2] >> 12) & 0x0f; |
| | | 125 | if (n != 0) { |
| | | 126 | /* |
| | | 127 | * Extended Method. |
| | | 128 | * core_bits = 2 ^ n (power of two) |
| | | 129 | */ |
| | | 130 | core_bits = 1 << n; |
| | | 131 | } |
| | | 132 | break; |
| | | 133 | default: |
| | | 134 | core_max = 1; |
74 | } | | 135 | } |
| | | 136 | |
75 | KASSERT(lp_max >= core_max); | | 137 | KASSERT(lp_max >= core_max); |
76 | smt_max = lp_max / core_max; | | 138 | smt_bits = ilog2((lp_max / core_max) - 1) + 1; |
77 | smt_bits = ilog2(smt_max - 1) + 1; | | 139 | if (core_bits == 0) { |
78 | core_bits = ilog2(core_max - 1) + 1; | | 140 | core_bits = ilog2(core_max - 1) + 1; |
| | | 141 | } |
| | | 142 | |
| | | 143 | /* |
| | | 144 | * Family 0xf and 0x10 processors may have different structure of |
| | | 145 | * APIC ID. Detect that via special MSR register and move the bits, |
| | | 146 | * if necessary (ref: InitApicIdCpuIdLo). |
| | | 147 | */ |
| | | 148 | if (cpu_vendor == CPUVENDOR_AMD && cpu_family < 0x11) { /* XXX */ |
| | | 149 | const uint64_t reg = rdmsr(MSR_NB_CFG); |
| | | 150 | if ((reg & NB_CFG_INITAPICCPUIDLO) == 0) { |
| | | 151 | /* |
| | | 152 | * 0xf: { CoreId, NodeId[2:0] } |
| | | 153 | * 0x10: { CoreId[1:0], 000b, NodeId[2:0] } |
| | | 154 | */ |
| | | 155 | const u_int node_id = apic_id & __BITS(0, 2); |
| | | 156 | apic_id = (cpu_family == 0xf) ? |
| | | 157 | (apic_id >> core_bits) | (node_id << core_bits) : |
| | | 158 | (apic_id >> 5) | (node_id << 2); |
| | | 159 | } |
| | | 160 | } |
| | | 161 | |
79 | if (smt_bits + core_bits) { | | 162 | if (smt_bits + core_bits) { |
80 | ci->ci_packageid = ci->ci_initapicid >> (smt_bits + core_bits); | | 163 | ci->ci_packageid = apic_id >> (smt_bits + core_bits); |
81 | } | | 164 | } |
82 | if (core_bits) { | | 165 | if (core_bits) { |
83 | u_int core_mask = __BITS(smt_bits, smt_bits + core_bits - 1); | | 166 | u_int core_mask = __BITS(smt_bits, smt_bits + core_bits - 1); |
84 | ci->ci_coreid = __SHIFTOUT(ci->ci_initapicid, core_mask); | | 167 | ci->ci_coreid = __SHIFTOUT(apic_id, core_mask); |
85 | } | | 168 | } |
86 | if (smt_bits) { | | 169 | if (smt_bits) { |
87 | u_int smt_mask = __BITS(0, smt_bits - 1); | | 170 | u_int smt_mask = __BITS(0, smt_bits - 1); |
88 | ci->ci_smtid = __SHIFTOUT(ci->ci_initapicid, smt_mask); | | 171 | ci->ci_smtid = __SHIFTOUT(apic_id, smt_mask); |
89 | } | | 172 | } |
90 | } | | 173 | } |