1 .. SPDX-License-Identifier: GPL-2.0
2
3 =================
4 x86 Feature Flags
5 =================
6
7 Introduction
8 ============
9
10 The list of feature flags in /proc/cpuinfo is not complete and
11 represents an ill-fated attempt from long time ago to put feature flags
12 in an easy to find place for userspace.
13
14 However, the amount of feature flags is growing by the CPU generation,
15 leading to unparseable and unwieldy /proc/cpuinfo.
16
17 What is more, those feature flags do not even need to be in that file
18 because userspace doesn't care about them - glibc et al already use
19 CPUID to find out what the target machine supports and what not.
20
21 And even if it doesn't show a particular feature flag - although the CPU
22 still does have support for the respective hardware functionality and
23 said CPU supports CPUID faulting - userspace can simply probe for the
24 feature and figure out if it is supported or not, regardless of whether
25 it is being advertised somewhere.
26
27 Furthermore, those flag strings become an ABI the moment they appear
28 there and maintaining them forever when nothing even uses them is a lot
29 of wasted effort.
30
31 So, the current use of /proc/cpuinfo is to show features which the
32 kernel has *enabled* and *supports*. As in: the CPUID feature flag is
33 there, there's an additional setup which the kernel has done while
34 booting and the functionality is ready to use. A perfect example for
35 that is "user_shstk" where additional code enablement is present in the
36 kernel to support shadow stack for user programs.
37
38 So, if users want to know if a feature is available on a given system,
39 they try to find the flag in /proc/cpuinfo. If a given flag is present,
40 it means that
41
42 * the kernel knows about the feature enough to have an X86_FEATURE bit
43
44 * the kernel supports it and is currently making it available either to
45 userspace or some other part of the kernel
46
47 * if the flag represents a hardware feature the hardware supports it.
48
49 The absence of a flag in /proc/cpuinfo by itself means almost nothing to
50 an end user.
51
52 On the one hand, a feature like "vaes" might be fully available to user
53 applications on a kernel that has not defined X86_FEATURE_VAES and thus
54 there is no "vaes" in /proc/cpuinfo.
55
56 On the other hand, a new kernel running on non-VAES hardware would also
57 have no "vaes" in /proc/cpuinfo. There's no way for an application or
58 user to tell the difference.
59
60 The end result is that the flags field in /proc/cpuinfo is marginally
61 useful for kernel debugging, but not really for anything else.
62 Applications should instead use things like the glibc facilities for
63 querying CPU support. Users should rely on tools like
64 tools/arch/x86/kcpuid and cpuid(1).
65
66 Regarding implementation, flags appearing in /proc/cpuinfo have an
67 X86_FEATURE definition in arch/x86/include/asm/cpufeatures.h. These flags
68 represent hardware features as well as software features.
69
70 If the kernel cares about a feature or KVM want to expose the feature to
71 a KVM guest, it should only then expose it to the guest when the guest
72 needs to parse /proc/cpuinfo. Which, as mentioned above, is highly
73 unlikely. KVM can synthesize the CPUID bit and the KVM guest can simply
74 query CPUID and figure out what the hypervisor supports and what not. As
75 already stated, /proc/cpuinfo is not a dumping ground for useless
76 feature flags.
77
78
79 How are feature flags created?
80 ==============================
81
82 a: Feature flags can be derived from the contents of CPUID leaves.
83 ------------------------------------------------------------------
84 These feature definitions are organized mirroring the layout of CPUID
85 leaves and grouped in words with offsets as mapped in enum cpuid_leafs
86 in cpufeatures.h (see arch/x86/include/asm/cpufeatures.h for details).
87 If a feature is defined with a X86_FEATURE_<name> definition in
88 cpufeatures.h, and if it is detected at run time, the flags will be
89 displayed accordingly in /proc/cpuinfo. For example, the flag "avx2"
90 comes from X86_FEATURE_AVX2 in cpufeatures.h.
91
92 b: Flags can be from scattered CPUID-based features.
93 ----------------------------------------------------
94 Hardware features enumerated in sparsely populated CPUID leaves get
95 software-defined values. Still, CPUID needs to be queried to determine
96 if a given feature is present. This is done in init_scattered_cpuid_features().
97 For instance, X86_FEATURE_CQM_LLC is defined as 11*32 + 0 and its presence is
98 checked at runtime in the respective CPUID leaf [EAX=f, ECX=0] bit EDX[1].
99
100 The intent of scattering CPUID leaves is to not bloat struct
101 cpuinfo_x86.x86_capability[] unnecessarily. For instance, the CPUID leaf
102 [EAX=7, ECX=0] has 30 features and is dense, but the CPUID leaf [EAX=7, EAX=1]
103 has only one feature and would waste 31 bits of space in the x86_capability[]
104 array. Since there is a struct cpuinfo_x86 for each possible CPU, the wasted
105 memory is not trivial.
106
107 c: Flags can be created synthetically under certain conditions for hardware features.
108 -------------------------------------------------------------------------------------
109 Examples of conditions include whether certain features are present in
110 MSR_IA32_CORE_CAPS or specific CPU models are identified. If the needed
111 conditions are met, the features are enabled by the set_cpu_cap or
112 setup_force_cpu_cap macros. For example, if bit 5 is set in MSR_IA32_CORE_CAPS,
113 the feature X86_FEATURE_SPLIT_LOCK_DETECT will be enabled and
114 "split_lock_detect" will be displayed. The flag "ring3mwait" will be
115 displayed only when running on INTEL_XEON_PHI_[KNL|KNM] processors.
116
117 d: Flags can represent purely software features.
118 ------------------------------------------------
119 These flags do not represent hardware features. Instead, they represent a
120 software feature implemented in the kernel. For example, Kernel Page Table
121 Isolation is purely software feature and its feature flag X86_FEATURE_PTI is
122 also defined in cpufeatures.h.
123
124 Naming of Flags
125 ===============
126
127 The script arch/x86/kernel/cpu/mkcapflags.sh processes the
128 #define X86_FEATURE_<name> from cpufeatures.h and generates the
129 x86_cap/bug_flags[] arrays in kernel/cpu/capflags.c. The names in the
130 resulting x86_cap/bug_flags[] are used to populate /proc/cpuinfo. The naming
131 of flags in the x86_cap/bug_flags[] are as follows:
132
133 a: The name of the flag is from the string in X86_FEATURE_<name> by default.
134 ----------------------------------------------------------------------------
135 By default, the flag <name> in /proc/cpuinfo is extracted from the respective
136 X86_FEATURE_<name> in cpufeatures.h. For example, the flag "avx2" is from
137 X86_FEATURE_AVX2.
138
139 b: The naming can be overridden.
140 --------------------------------
141 If the comment on the line for the #define X86_FEATURE_* starts with a
142 double-quote character (""), the string inside the double-quote characters
143 will be the name of the flags. For example, the flag "sse4_1" comes from
144 the comment "sse4_1" following the X86_FEATURE_XMM4_1 definition.
145
146 There are situations in which overriding the displayed name of the flag is
147 needed. For instance, /proc/cpuinfo is a userspace interface and must remain
148 constant. If, for some reason, the naming of X86_FEATURE_<name> changes, one
149 shall override the new naming with the name already used in /proc/cpuinfo.
150
151 c: The naming override can be "", which means it will not appear in /proc/cpuinfo.
152 ----------------------------------------------------------------------------------
153 The feature shall be omitted from /proc/cpuinfo if it does not make sense for
154 the feature to be exposed to userspace. For example, X86_FEATURE_ALWAYS is
155 defined in cpufeatures.h but that flag is an internal kernel feature used
156 in the alternative runtime patching functionality. So, its name is overridden
157 with "". Its flag will not appear in /proc/cpuinfo.
158
159 Flags are missing when one or more of these happen
160 ==================================================
161
162 a: The hardware does not enumerate support for it.
163 --------------------------------------------------
164 For example, when a new kernel is running on old hardware or the feature is
165 not enabled by boot firmware. Even if the hardware is new, there might be a
166 problem enabling the feature at run time, the flag will not be displayed.
167
168 b: The kernel does not know about the flag.
169 -------------------------------------------
170 For example, when an old kernel is running on new hardware.
171
172 c: The kernel disabled support for it at compile-time.
173 ------------------------------------------------------
174 For example, if 5-level-paging is not enabled when building (i.e.,
175 CONFIG_X86_5LEVEL is not selected) the flag "la57" will not show up [#f1]_.
176 Even though the feature will still be detected via CPUID, the kernel disables
177 it by clearing via setup_clear_cpu_cap(X86_FEATURE_LA57).
178
179 d: The feature is disabled at boot-time.
180 ----------------------------------------
181 A feature can be disabled either using a command-line parameter or because
182 it failed to be enabled. The command-line parameter clearcpuid= can be used
183 to disable features using the feature number as defined in
184 /arch/x86/include/asm/cpufeatures.h. For instance, User Mode Instruction
185 Protection can be disabled using clearcpuid=514. The number 514 is calculated
186 from #define X86_FEATURE_UMIP (16*32 + 2).
187
188 In addition, there exists a variety of custom command-line parameters that
189 disable specific features. The list of parameters includes, but is not limited
190 to, nofsgsbase, nosgx, noxsave, etc. 5-level paging can also be disabled using
191 "no5lvl".
192
193 e: The feature was known to be non-functional.
194 ----------------------------------------------
195 The feature was known to be non-functional because a dependency was
196 missing at runtime. For example, AVX flags will not show up if XSAVE feature
197 is disabled since they depend on XSAVE feature. Another example would be broken
198 CPUs and them missing microcode patches. Due to that, the kernel decides not to
199 enable a feature.
200
201 .. [#f1] 5-level paging uses linear address of 57 bits.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.