Documentation/cputopology.txt

   1 ===========================================
   2 How CPU topology info is exported via sysfs
   3 ===========================================
   4
   5 Export CPU topology info via sysfs. Items (attributes) are similar
   6 to /proc/cpuinfo output of some architectures:
   7
   8 1) /sys/devices/system/cpu/cpuX/topology/physical_package_id:
   9
  10         physical package id of cpuX. Typically corresponds to a physical
  11         socket number, but the actual value is architecture and platform
  12         dependent.
  13
  14 2) /sys/devices/system/cpu/cpuX/topology/core_id:
  15
  16         the CPU core ID of cpuX. Typically it is the hardware platform's
  17         identifier (rather than the kernel's).  The actual value is
  18         architecture and platform dependent.
  19
  20 3) /sys/devices/system/cpu/cpuX/topology/book_id:
  21
  22         the book ID of cpuX. Typically it is the hardware platform's
  23         identifier (rather than the kernel's).  The actual value is
  24         architecture and platform dependent.
  25
  26 4) /sys/devices/system/cpu/cpuX/topology/drawer_id:
  27
  28         the drawer ID of cpuX. Typically it is the hardware platform's
  29         identifier (rather than the kernel's).  The actual value is
  30         architecture and platform dependent.
  31
  32 5) /sys/devices/system/cpu/cpuX/topology/thread_siblings:
  33
  34         internal kernel map of cpuX's hardware threads within the same
  35         core as cpuX.
  36
  37 6) /sys/devices/system/cpu/cpuX/topology/thread_siblings_list:
  38
  39         human-readable list of cpuX's hardware threads within the same
  40         core as cpuX.
  41
  42 7) /sys/devices/system/cpu/cpuX/topology/core_siblings:
  43
  44         internal kernel map of cpuX's hardware threads within the same
  45         physical_package_id.
  46
  47 8) /sys/devices/system/cpu/cpuX/topology/core_siblings_list:
  48
  49         human-readable list of cpuX's hardware threads within the same
  50         physical_package_id.
  51
  52 9) /sys/devices/system/cpu/cpuX/topology/book_siblings:
  53
  54         internal kernel map of cpuX's hardware threads within the same
  55         book_id.
  56
  57 10) /sys/devices/system/cpu/cpuX/topology/book_siblings_list:
  58
  59         human-readable list of cpuX's hardware threads within the same
  60         book_id.
  61
  62 11) /sys/devices/system/cpu/cpuX/topology/drawer_siblings:
  63
  64         internal kernel map of cpuX's hardware threads within the same
  65         drawer_id.
  66
  67 12) /sys/devices/system/cpu/cpuX/topology/drawer_siblings_list:
  68
  69         human-readable list of cpuX's hardware threads within the same
  70         drawer_id.
  71
  72 To implement it in an architecture-neutral way, a new source file,
  73 drivers/base/topology.c, is to export the 6 to 12 attributes. The book
  74 and drawer related sysfs files will only be created if CONFIG_SCHED_BOOK
  75 and CONFIG_SCHED_DRAWER are selected.
  76
  77 CONFIG_SCHED_BOOK and CONFIG_DRAWER are currently only used on s390, where
  78 they reflect the cpu and cache hierarchy.
  79
  80 For an architecture to support this feature, it must define some of
  81 these macros in include/asm-XXX/topology.h::
  82
  83         #define topology_physical_package_id(cpu)
  84         #define topology_core_id(cpu)
  85         #define topology_book_id(cpu)
  86         #define topology_drawer_id(cpu)
  87         #define topology_sibling_cpumask(cpu)
  88         #define topology_core_cpumask(cpu)
  89         #define topology_book_cpumask(cpu)
  90         #define topology_drawer_cpumask(cpu)
  91
  92 The type of ``**_id macros`` is int.
  93 The type of ``**_cpumask macros`` is ``(const) struct cpumask *``. The latter
  94 correspond with appropriate ``**_siblings`` sysfs attributes (except for
  95 topology_sibling_cpumask() which corresponds with thread_siblings).
  96
  97 To be consistent on all architectures, include/linux/topology.h
  98 provides default definitions for any of the above macros that are
  99 not defined by include/asm-XXX/topology.h:
 100
 101 1) physical_package_id: -1
 102 2) core_id: 0
 103 3) sibling_cpumask: just the given CPU
 104 4) core_cpumask: just the given CPU
 105
 106 For architectures that don't support books (CONFIG_SCHED_BOOK) there are no
 107 default definitions for topology_book_id() and topology_book_cpumask().
 108 For architectures that don't support drawers (CONFIG_SCHED_DRAWER) there are
 109 no default definitions for topology_drawer_id() and topology_drawer_cpumask().
 110
 111 Additionally, CPU topology information is provided under
 112 /sys/devices/system/cpu and includes these files.  The internal
 113 source for the output is in brackets ("[]").
 114
 115     =========== ==========================================================
 116     kernel_max: the maximum CPU index allowed by the kernel configuration.
 117                 [NR_CPUS-1]
 118
 119     offline:    CPUs that are not online because they have been
 120                 HOTPLUGGED off (see cpu-hotplug.txt) or exceed the limit
 121                 of CPUs allowed by the kernel configuration (kernel_max
 122                 above). [~cpu_online_mask + cpus >= NR_CPUS]
 123
 124     online:     CPUs that are online and being scheduled [cpu_online_mask]
 125
 126     possible:   CPUs that have been allocated resources and can be
 127                 brought online if they are present. [cpu_possible_mask]
 128
 129     present:    CPUs that have been identified as being present in the
 130                 system. [cpu_present_mask]
 131     =========== ==========================================================
 132
 133 The format for the above output is compatible with cpulist_parse()
 134 [see <linux/cpumask.h>].  Some examples follow.
 135
 136 In this example, there are 64 CPUs in the system but cpus 32-63 exceed
 137 the kernel max which is limited to 0..31 by the NR_CPUS config option
 138 being 32.  Note also that CPUs 2 and 4-31 are not online but could be
 139 brought online as they are both present and possible::
 140
 141      kernel_max: 31
 142         offline: 2,4-31,32-63
 143          online: 0-1,3
 144        possible: 0-31
 145         present: 0-31
 146
 147 In this example, the NR_CPUS config option is 128, but the kernel was
 148 started with possible_cpus=144.  There are 4 CPUs in the system and cpu2
 149 was manually taken offline (and is the only CPU that can be brought
 150 online.)::
 151
 152      kernel_max: 127
 153         offline: 2,4-127,128-143
 154          online: 0-1,3
 155        possible: 0-127
 156         present: 0-3
 157
 158 See cpu-hotplug.txt for the possible_cpus=NUM kernel start parameter
 159 as well as more information on the various cpumasks.