]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - Documentation/cputopology.txt
mm: add comment why mark_page_accessed() would be better than pte_mkyoung() in follow...
[mirror_ubuntu-jammy-kernel.git] / Documentation / cputopology.txt
1
2 Export cpu topology info via sysfs. Items (attributes) are similar
3 to /proc/cpuinfo.
4
5 1) /sys/devices/system/cpu/cpuX/topology/physical_package_id:
6 represent the physical package id of cpu X;
7 2) /sys/devices/system/cpu/cpuX/topology/core_id:
8 represent the cpu core id to cpu X;
9 3) /sys/devices/system/cpu/cpuX/topology/thread_siblings:
10 represent the thread siblings to cpu X in the same core;
11 4) /sys/devices/system/cpu/cpuX/topology/core_siblings:
12 represent the thread siblings to cpu X in the same physical package;
13
14 To implement it in an architecture-neutral way, a new source file,
15 drivers/base/topology.c, is to export the 4 attributes.
16
17 For an architecture to support this feature, it must define some of
18 these macros in include/asm-XXX/topology.h:
19 #define topology_physical_package_id(cpu)
20 #define topology_core_id(cpu)
21 #define topology_thread_cpumask(cpu)
22 #define topology_core_cpumask(cpu)
23
24 The type of **_id is int.
25 The type of siblings is (const) struct cpumask *.
26
27 To be consistent on all architectures, include/linux/topology.h
28 provides default definitions for any of the above macros that are
29 not defined by include/asm-XXX/topology.h:
30 1) physical_package_id: -1
31 2) core_id: 0
32 3) thread_siblings: just the given CPU
33 4) core_siblings: just the given CPU
34
35 Additionally, cpu topology information is provided under
36 /sys/devices/system/cpu and includes these files. The internal
37 source for the output is in brackets ("[]").
38
39 kernel_max: the maximum cpu index allowed by the kernel configuration.
40 [NR_CPUS-1]
41
42 offline: cpus that are not online because they have been
43 HOTPLUGGED off (see cpu-hotplug.txt) or exceed the limit
44 of cpus allowed by the kernel configuration (kernel_max
45 above). [~cpu_online_mask + cpus >= NR_CPUS]
46
47 online: cpus that are online and being scheduled [cpu_online_mask]
48
49 possible: cpus that have been allocated resources and can be
50 brought online if they are present. [cpu_possible_mask]
51
52 present: cpus that have been identified as being present in the
53 system. [cpu_present_mask]
54
55 The format for the above output is compatible with cpulist_parse()
56 [see <linux/cpumask.h>]. Some examples follow.
57
58 In this example, there are 64 cpus in the system but cpus 32-63 exceed
59 the kernel max which is limited to 0..31 by the NR_CPUS config option
60 being 32. Note also that cpus 2 and 4-31 are not online but could be
61 brought online as they are both present and possible.
62
63 kernel_max: 31
64 offline: 2,4-31,32-63
65 online: 0-1,3
66 possible: 0-31
67 present: 0-31
68
69 In this example, the NR_CPUS config option is 128, but the kernel was
70 started with possible_cpus=144. There are 4 cpus in the system and cpu2
71 was manually taken offline (and is the only cpu that can be brought
72 online.)
73
74 kernel_max: 127
75 offline: 2,4-127,128-143
76 online: 0-1,3
77 possible: 0-127
78 present: 0-3
79
80 See cpu-hotplug.txt for the possible_cpus=NUM kernel start parameter
81 as well as more information on the various cpumask's.