path: root/Documentation/cputopology.txt (plain)
blob: f722f227a73bbc2cc0bbbbd6e0e65ab0d2f8b23e

1
2	Export CPU topology info via sysfs. Items (attributes) are similar
3	to /proc/cpuinfo output of some architectures:
4
5	1) /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
11	2) /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
17	3) /sys/devices/system/cpu/cpuX/topology/book_id:
18
19	the book ID of cpuX. Typically it is the hardware platform's
20	identifier (rather than the kernel's). The actual value is
21	architecture and platform dependent.
22
23	4) /sys/devices/system/cpu/cpuX/topology/drawer_id:
24
25	the drawer ID of cpuX. Typically it is the hardware platform's
26	identifier (rather than the kernel's). The actual value is
27	architecture and platform dependent.
28
29	5) /sys/devices/system/cpu/cpuX/topology/thread_siblings:
30
31	internal kernel map of cpuX's hardware threads within the same
32	core as cpuX.
33
34	6) /sys/devices/system/cpu/cpuX/topology/thread_siblings_list:
35
36	human-readable list of cpuX's hardware threads within the same
37	core as cpuX.
38
39	7) /sys/devices/system/cpu/cpuX/topology/core_siblings:
40
41	internal kernel map of cpuX's hardware threads within the same
42	physical_package_id.
43
44	8) /sys/devices/system/cpu/cpuX/topology/core_siblings_list:
45
46	human-readable list of cpuX's hardware threads within the same
47	physical_package_id.
48
49	9) /sys/devices/system/cpu/cpuX/topology/book_siblings:
50
51	internal kernel map of cpuX's hardware threads within the same
52	book_id.
53
54	10) /sys/devices/system/cpu/cpuX/topology/book_siblings_list:
55
56	human-readable list of cpuX's hardware threads within the same
57	book_id.
58
59	11) /sys/devices/system/cpu/cpuX/topology/drawer_siblings:
60
61	internal kernel map of cpuX's hardware threads within the same
62	drawer_id.
63
64	12) /sys/devices/system/cpu/cpuX/topology/drawer_siblings_list:
65
66	human-readable list of cpuX's hardware threads within the same
67	drawer_id.
68
69	To implement it in an architecture-neutral way, a new source file,
70	drivers/base/topology.c, is to export the 6 to 12 attributes. The book
71	and drawer related sysfs files will only be created if CONFIG_SCHED_BOOK
72	and CONFIG_SCHED_DRAWER are selected.
73
74	CONFIG_SCHED_BOOK and CONFIG_DRAWER are currently only used on s390, where
75	they reflect the cpu and cache hierarchy.
76
77	For an architecture to support this feature, it must define some of
78	these macros in include/asm-XXX/topology.h:
79	#define topology_physical_package_id(cpu)
80	#define topology_core_id(cpu)
81	#define topology_book_id(cpu)
82	#define topology_drawer_id(cpu)
83	#define topology_sibling_cpumask(cpu)
84	#define topology_core_cpumask(cpu)
85	#define topology_book_cpumask(cpu)
86	#define topology_drawer_cpumask(cpu)
87
88	The type of **_id macros is int.
89	The type of **_cpumask macros is (const) struct cpumask *. The latter
90	correspond with appropriate **_siblings sysfs attributes (except for
91	topology_sibling_cpumask() which corresponds with thread_siblings).
92
93	To be consistent on all architectures, include/linux/topology.h
94	provides default definitions for any of the above macros that are
95	not defined by include/asm-XXX/topology.h:
96	1) physical_package_id: -1
97	2) core_id: 0
98	3) sibling_cpumask: just the given CPU
99	4) core_cpumask: just the given CPU
100
101	For architectures that don't support books (CONFIG_SCHED_BOOK) there are no
102	default definitions for topology_book_id() and topology_book_cpumask().
103	For architectures that don't support drawes (CONFIG_SCHED_DRAWER) there are
104	no default definitions for topology_drawer_id() and topology_drawer_cpumask().
105
106	Additionally, CPU topology information is provided under
107	/sys/devices/system/cpu and includes these files. The internal
108	source for the output is in brackets ("[]").
109
110	kernel_max: the maximum CPU index allowed by the kernel configuration.
111	[NR_CPUS-1]
112
113	offline: CPUs that are not online because they have been
114	HOTPLUGGED off (see cpu-hotplug.txt) or exceed the limit
115	of CPUs allowed by the kernel configuration (kernel_max
116	above). [~cpu_online_mask + cpus >= NR_CPUS]
117
118	online: CPUs that are online and being scheduled [cpu_online_mask]
119
120	possible: CPUs that have been allocated resources and can be
121	brought online if they are present. [cpu_possible_mask]
122
123	present: CPUs that have been identified as being present in the
124	system. [cpu_present_mask]
125
126	The format for the above output is compatible with cpulist_parse()
127	[see <linux/cpumask.h>]. Some examples follow.
128
129	In this example, there are 64 CPUs in the system but cpus 32-63 exceed
130	the kernel max which is limited to 0..31 by the NR_CPUS config option
131	being 32. Note also that CPUs 2 and 4-31 are not online but could be
132	brought online as they are both present and possible.
133
134	kernel_max: 31
135	offline: 2,4-31,32-63
136	online: 0-1,3
137	possible: 0-31
138	present: 0-31
139
140	In this example, the NR_CPUS config option is 128, but the kernel was
141	started with possible_cpus=144. There are 4 CPUs in the system and cpu2
142	was manually taken offline (and is the only CPU that can be brought
143	online.)
144
145	kernel_max: 127
146	offline: 2,4-127,128-143
147	online: 0-1,3
148	possible: 0-127
149	present: 0-3
150
151	See cpu-hotplug.txt for the possible_cpus=NUM kernel start parameter
152	as well as more information on the various cpumasks.
153