Root/Documentation/cputopology.txt

1
2Export CPU topology info via sysfs. Items (attributes) are similar
3to /proc/cpuinfo.
4
51) /sys/devices/system/cpu/cpuX/topology/physical_package_id:
6
7    physical package id of cpuX. Typically corresponds to a physical
8    socket number, but the actual value is architecture and platform
9    dependent.
10
112) /sys/devices/system/cpu/cpuX/topology/core_id:
12
13    the CPU core ID of cpuX. Typically it is the hardware platform's
14    identifier (rather than the kernel's). The actual value is
15    architecture and platform dependent.
16
173) /sys/devices/system/cpu/cpuX/topology/thread_siblings:
18
19    internel kernel map of cpuX's hardware threads within the same
20    core as cpuX
21
224) /sys/devices/system/cpu/cpuX/topology/core_siblings:
23
24    internal kernel map of cpuX's hardware threads within the same
25    physical_package_id.
26
27To implement it in an architecture-neutral way, a new source file,
28drivers/base/topology.c, is to export the 4 attributes.
29
30For an architecture to support this feature, it must define some of
31these macros in include/asm-XXX/topology.h:
32#define topology_physical_package_id(cpu)
33#define topology_core_id(cpu)
34#define topology_thread_cpumask(cpu)
35#define topology_core_cpumask(cpu)
36
37The type of **_id is int.
38The type of siblings is (const) struct cpumask *.
39
40To be consistent on all architectures, include/linux/topology.h
41provides default definitions for any of the above macros that are
42not defined by include/asm-XXX/topology.h:
431) physical_package_id: -1
442) core_id: 0
453) thread_siblings: just the given CPU
464) core_siblings: just the given CPU
47
48Additionally, CPU topology information is provided under
49/sys/devices/system/cpu and includes these files. The internal
50source for the output is in brackets ("[]").
51
52    kernel_max: the maximum CPU index allowed by the kernel configuration.
53        [NR_CPUS-1]
54
55    offline: CPUs that are not online because they have been
56        HOTPLUGGED off (see cpu-hotplug.txt) or exceed the limit
57        of CPUs allowed by the kernel configuration (kernel_max
58        above). [~cpu_online_mask + cpus >= NR_CPUS]
59
60    online: CPUs that are online and being scheduled [cpu_online_mask]
61
62    possible: CPUs that have been allocated resources and can be
63        brought online if they are present. [cpu_possible_mask]
64
65    present: CPUs that have been identified as being present in the
66        system. [cpu_present_mask]
67
68The format for the above output is compatible with cpulist_parse()
69[see <linux/cpumask.h>]. Some examples follow.
70
71In this example, there are 64 CPUs in the system but cpus 32-63 exceed
72the kernel max which is limited to 0..31 by the NR_CPUS config option
73being 32. Note also that CPUs 2 and 4-31 are not online but could be
74brought online as they are both present and possible.
75
76     kernel_max: 31
77        offline: 2,4-31,32-63
78         online: 0-1,3
79       possible: 0-31
80        present: 0-31
81
82In this example, the NR_CPUS config option is 128, but the kernel was
83started with possible_cpus=144. There are 4 CPUs in the system and cpu2
84was manually taken offline (and is the only CPU that can be brought
85online.)
86
87     kernel_max: 127
88        offline: 2,4-127,128-143
89         online: 0-1,3
90       possible: 0-127
91        present: 0-3
92
93See cpu-hotplug.txt for the possible_cpus=NUM kernel start parameter
94as well as more information on the various cpumasks.
95

Archive Download this file



interactive