1 ===========================================
2 How CPU topology info is exported via sysfs
3 ===========================================
5 Export CPU topology info via sysfs. Items (attributes) are similar
6 to /proc/cpuinfo output of some architectures. They reside in
7 /sys/devices/system/cpu/cpuX/topology/:
11 physical package id of cpuX. Typically corresponds to a physical
12 socket number, but the actual value is architecture and platform
17 the CPU die ID of cpuX. Typically it is the hardware platform's
18 identifier (rather than the kernel's). The actual value is
19 architecture and platform dependent.
23 the CPU core ID of cpuX. Typically it is the hardware platform's
24 identifier (rather than the kernel's). The actual value is
25 architecture and platform dependent.
29 the book ID of cpuX. Typically it is the hardware platform's
30 identifier (rather than the kernel's). The actual value is
31 architecture and platform dependent.
35 the drawer ID of cpuX. Typically it is the hardware platform's
36 identifier (rather than the kernel's). The actual value is
37 architecture and platform dependent.
41 internal kernel map of CPUs within the same core.
42 (deprecated name: "thread_siblings")
46 human-readable list of CPUs within the same core.
47 (deprecated name: "thread_siblings_list");
51 internal kernel map of the CPUs sharing the same physical_package_id.
52 (deprecated name: "core_siblings")
56 human-readable list of CPUs sharing the same physical_package_id.
57 (deprecated name: "core_siblings_list")
61 internal kernel map of CPUs within the same die.
65 human-readable list of CPUs within the same die.
69 internal kernel map of cpuX's hardware threads within the same
74 human-readable list of cpuX's hardware threads within the same
79 internal kernel map of cpuX's hardware threads within the same
84 human-readable list of cpuX's hardware threads within the same
87 Architecture-neutral, drivers/base/topology.c, exports these attributes.
88 However, the book and drawer related sysfs files will only be created if
89 CONFIG_SCHED_BOOK and CONFIG_SCHED_DRAWER are selected, respectively.
91 CONFIG_SCHED_BOOK and CONFIG_SCHED_DRAWER are currently only used on s390,
92 where they reflect the cpu and cache hierarchy.
94 For an architecture to support this feature, it must define some of
95 these macros in include/asm-XXX/topology.h::
97 #define topology_physical_package_id(cpu)
98 #define topology_die_id(cpu)
99 #define topology_core_id(cpu)
100 #define topology_book_id(cpu)
101 #define topology_drawer_id(cpu)
102 #define topology_sibling_cpumask(cpu)
103 #define topology_core_cpumask(cpu)
104 #define topology_die_cpumask(cpu)
105 #define topology_book_cpumask(cpu)
106 #define topology_drawer_cpumask(cpu)
108 The type of ``**_id macros`` is int.
109 The type of ``**_cpumask macros`` is ``(const) struct cpumask *``. The latter
110 correspond with appropriate ``**_siblings`` sysfs attributes (except for
111 topology_sibling_cpumask() which corresponds with thread_siblings).
113 To be consistent on all architectures, include/linux/topology.h
114 provides default definitions for any of the above macros that are
115 not defined by include/asm-XXX/topology.h:
117 1) topology_physical_package_id: -1
118 2) topology_die_id: -1
119 3) topology_core_id: 0
120 4) topology_sibling_cpumask: just the given CPU
121 5) topology_core_cpumask: just the given CPU
122 6) topology_die_cpumask: just the given CPU
124 For architectures that don't support books (CONFIG_SCHED_BOOK) there are no
125 default definitions for topology_book_id() and topology_book_cpumask().
126 For architectures that don't support drawers (CONFIG_SCHED_DRAWER) there are
127 no default definitions for topology_drawer_id() and topology_drawer_cpumask().
129 Additionally, CPU topology information is provided under
130 /sys/devices/system/cpu and includes these files. The internal
131 source for the output is in brackets ("[]").
133 =========== ==========================================================
134 kernel_max: the maximum CPU index allowed by the kernel configuration.
137 offline: CPUs that are not online because they have been
138 HOTPLUGGED off (see cpu-hotplug.txt) or exceed the limit
139 of CPUs allowed by the kernel configuration (kernel_max
140 above). [~cpu_online_mask + cpus >= NR_CPUS]
142 online: CPUs that are online and being scheduled [cpu_online_mask]
144 possible: CPUs that have been allocated resources and can be
145 brought online if they are present. [cpu_possible_mask]
147 present: CPUs that have been identified as being present in the
148 system. [cpu_present_mask]
149 =========== ==========================================================
151 The format for the above output is compatible with cpulist_parse()
152 [see <linux/cpumask.h>]. Some examples follow.
154 In this example, there are 64 CPUs in the system but cpus 32-63 exceed
155 the kernel max which is limited to 0..31 by the NR_CPUS config option
156 being 32. Note also that CPUs 2 and 4-31 are not online but could be
157 brought online as they are both present and possible::
160 offline: 2,4-31,32-63
165 In this example, the NR_CPUS config option is 128, but the kernel was
166 started with possible_cpus=144. There are 4 CPUs in the system and cpu2
167 was manually taken offline (and is the only CPU that can be brought
171 offline: 2,4-127,128-143
176 See cpu-hotplug.txt for the possible_cpus=NUM kernel start parameter
177 as well as more information on the various cpumasks.