1 Using XSTATE features in user space applications
2 ================================================
3
4 The x86 architecture supports floating-point extensions which are
5 enumerated via CPUID. Applications consult CPUID and use XGETBV to
6 evaluate which features have been enabled by the kernel XCR0.
7
8 Up to AVX-512 and PKRU states, these features are automatically enabled by
9 the kernel if available. Features like AMX TILE_DATA (XSTATE component 18)
10 are enabled by XCR0 as well, but the first use of related instruction is
11 trapped by the kernel because by default the required large XSTATE buffers
12 are not allocated automatically.
13
14 The purpose for dynamic features
15 --------------------------------
16
17 Legacy userspace libraries often have hard-coded, static sizes for
18 alternate signal stacks, often using MINSIGSTKSZ which is typically 2KB.
19 That stack must be able to store at *least* the signal frame that the
20 kernel sets up before jumping into the signal handler. That signal frame
21 must include an XSAVE buffer defined by the CPU.
22
23 However, that means that the size of signal stacks is dynamic, not static,
24 because different CPUs have differently-sized XSAVE buffers. A compiled-in
25 size of 2KB with existing applications is too small for new CPU features
26 like AMX. Instead of universally requiring larger stack, with the dynamic
27 enabling, the kernel can enforce userspace applications to have
28 properly-sized altstacks.
29
30 Using dynamically enabled XSTATE features in user space applications
31 --------------------------------------------------------------------
32
33 The kernel provides an arch_prctl(2) based mechanism for applications to
34 request the usage of such features. The arch_prctl(2) options related to
35 this are:
36
37 -ARCH_GET_XCOMP_SUPP
38
39 arch_prctl(ARCH_GET_XCOMP_SUPP, &features);
40
41 ARCH_GET_XCOMP_SUPP stores the supported features in userspace storage of
42 type uint64_t. The second argument is a pointer to that storage.
43
44 -ARCH_GET_XCOMP_PERM
45
46 arch_prctl(ARCH_GET_XCOMP_PERM, &features);
47
48 ARCH_GET_XCOMP_PERM stores the features for which the userspace process
49 has permission in userspace storage of type uint64_t. The second argument
50 is a pointer to that storage.
51
52 -ARCH_REQ_XCOMP_PERM
53
54 arch_prctl(ARCH_REQ_XCOMP_PERM, feature_nr);
55
56 ARCH_REQ_XCOMP_PERM allows to request permission for a dynamically enabled
57 feature or a feature set. A feature set can be mapped to a facility, e.g.
58 AMX, and can require one or more XSTATE components to be enabled.
59
60 The feature argument is the number of the highest XSTATE component which
61 is required for a facility to work.
62
63 When requesting permission for a feature, the kernel checks the
64 availability. The kernel ensures that sigaltstacks in the process's tasks
65 are large enough to accommodate the resulting large signal frame. It
66 enforces this both during ARCH_REQ_XCOMP_SUPP and during any subsequent
67 sigaltstack(2) calls. If an installed sigaltstack is smaller than the
68 resulting sigframe size, ARCH_REQ_XCOMP_SUPP results in -ENOSUPP. Also,
69 sigaltstack(2) results in -ENOMEM if the requested altstack is too small
70 for the permitted features.
71
72 Permission, when granted, is valid per process. Permissions are inherited
73 on fork(2) and cleared on exec(3).
74
75 The first use of an instruction related to a dynamically enabled feature is
76 trapped by the kernel. The trap handler checks whether the process has
77 permission to use the feature. If the process has no permission then the
78 kernel sends SIGILL to the application. If the process has permission then
79 the handler allocates a larger xstate buffer for the task so the large
80 state can be context switched. In the unlikely cases that the allocation
81 fails, the kernel sends SIGSEGV.
82
83 AMX TILE_DATA enabling example
84 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
85
86 Below is the example of how userspace applications enable
87 TILE_DATA dynamically:
88
89 1. The application first needs to query the kernel for AMX
90 support::
91
92 #include <asm/prctl.h>
93 #include <sys/syscall.h>
94 #include <stdio.h>
95 #include <unistd.h>
96
97 #ifndef ARCH_GET_XCOMP_SUPP
98 #define ARCH_GET_XCOMP_SUPP 0x1021
99 #endif
100
101 #ifndef ARCH_XCOMP_TILECFG
102 #define ARCH_XCOMP_TILECFG 17
103 #endif
104
105 #ifndef ARCH_XCOMP_TILEDATA
106 #define ARCH_XCOMP_TILEDATA 18
107 #endif
108
109 #define MASK_XCOMP_TILE ((1 << ARCH_XCOMP_TILECFG) | \
110 (1 << ARCH_XCOMP_TILEDATA))
111
112 unsigned long features;
113 long rc;
114
115 ...
116
117 rc = syscall(SYS_arch_prctl, ARCH_GET_XCOMP_SUPP, &features);
118
119 if (!rc && (features & MASK_XCOMP_TILE) == MASK_XCOMP_TILE)
120 printf("AMX is available.\n");
121
122 2. After that, determining support for AMX, an application must
123 explicitly ask permission to use it::
124
125 #ifndef ARCH_REQ_XCOMP_PERM
126 #define ARCH_REQ_XCOMP_PERM 0x1023
127 #endif
128
129 ...
130
131 rc = syscall(SYS_arch_prctl, ARCH_REQ_XCOMP_PERM, ARCH_XCOMP_TILEDATA);
132
133 if (!rc)
134 printf("AMX is ready for use.\n");
135
136 Note this example does not include the sigaltstack preparation.
137
138 Dynamic features in signal frames
139 ---------------------------------
140
141 Dynamically enabled features are not written to the signal frame upon signal
142 entry if the feature is in its initial configuration. This differs from
143 non-dynamic features which are always written regardless of their
144 configuration. Signal handlers can examine the XSAVE buffer's XSTATE_BV
145 field to determine if a features was written.
146
147 Dynamic features for virtual machines
148 -------------------------------------
149
150 The permission for the guest state component needs to be managed separately
151 from the host, as they are exclusive to each other. A coupled of options
152 are extended to control the guest permission:
153
154 -ARCH_GET_XCOMP_GUEST_PERM
155
156 arch_prctl(ARCH_GET_XCOMP_GUEST_PERM, &features);
157
158 ARCH_GET_XCOMP_GUEST_PERM is a variant of ARCH_GET_XCOMP_PERM. So it
159 provides the same semantics and functionality but for the guest
160 components.
161
162 -ARCH_REQ_XCOMP_GUEST_PERM
163
164 arch_prctl(ARCH_REQ_XCOMP_GUEST_PERM, feature_nr);
165
166 ARCH_REQ_XCOMP_GUEST_PERM is a variant of ARCH_REQ_XCOMP_PERM. It has the
167 same semantics for the guest permission. While providing a similar
168 functionality, this comes with a constraint. Permission is frozen when the
169 first VCPU is created. Any attempt to change permission after that point
170 is going to be rejected. So, the permission has to be requested before the
171 first VCPU creation.
172
173 Note that some VMMs may have already established a set of supported state
174 components. These options are not presumed to support any particular VMM.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.