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
   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/thread_siblings:
  18
  19         internel kernel map of cpuX's hardware threads within the same
  20         core as cpuX
  21
  22 4) /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
  27 To implement it in an architecture-neutral way, a new source file,
  28 drivers/base/topology.c, is to export the 4 attributes.
  29
  30 For an architecture to support this feature, it must define some of
  31 these 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
  37 The type of **_id is int.
  38 The type of siblings is (const) struct cpumask *.
  39
  40 To be consistent on all architectures, include/linux/topology.h
  41 provides default definitions for any of the above macros that are
  42 not defined by include/asm-XXX/topology.h:
  43 1) physical_package_id: -1
  44 2) core_id: 0
  45 3) thread_siblings: just the given CPU
  46 4) core_siblings: just the given CPU
  47
  48 Additionally, CPU topology information is provided under
  49 /sys/devices/system/cpu and includes these files.  The internal
  50 source 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
  68 The format for the above output is compatible with cpulist_parse()
  69 [see <linux/cpumask.h>].  Some examples follow.
  70
  71 In this example, there are 64 CPUs in the system but cpus 32-63 exceed
  72 the kernel max which is limited to 0..31 by the NR_CPUS config option
  73 being 32.  Note also that CPUs 2 and 4-31 are not online but could be
  74 brought 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
  82 In this example, the NR_CPUS config option is 128, but the kernel was
  83 started with possible_cpus=144.  There are 4 CPUs in the system and cpu2
  84 was manually taken offline (and is the only CPU that can be brought
  85 online.)
  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
  93 See cpu-hotplug.txt for the possible_cpus=NUM kernel start parameter
  94 as well as more information on the various cpumasks.