Documentation/ABI/testing/sysfs-devices-memory

   1 What:           /sys/devices/system/memory
   2 Date:           June 2008
   3 Contact:        Badari Pulavarty <pbadari@us.ibm.com>
   4 Description:
   5                 The /sys/devices/system/memory contains a snapshot of the
   6                 internal state of the kernel memory blocks. Files could be
   7                 added or removed dynamically to represent hot-add/remove
   8                 operations.
   9 Users:          hotplug memory add/remove tools
  10                 http://www.ibm.com/developerworks/wikis/display/LinuxP/powerpc-utils
  11
  12 What:           /sys/devices/system/memory/memoryX/removable
  13 Date:           June 2008
  14 Contact:        Badari Pulavarty <pbadari@us.ibm.com>
  15 Description:
  16                 The file /sys/devices/system/memory/memoryX/removable is a
  17                 legacy interface used to indicated whether a memory block is
  18                 likely to be offlineable or not.  Newer kernel versions return
  19                 "1" if and only if the kernel supports memory offlining.
  20 Users:          hotplug memory remove tools
  21                 http://www.ibm.com/developerworks/wikis/display/LinuxP/powerpc-utils
  22                 lsmem/chmem part of util-linux
  23
  24 What:           /sys/devices/system/memory/memoryX/phys_device
  25 Date:           September 2008
  26 Contact:        Badari Pulavarty <pbadari@us.ibm.com>
  27 Description:
  28                 The file /sys/devices/system/memory/memoryX/phys_device
  29                 is read-only;  it is a legacy interface only ever used on s390x
  30                 to expose the covered storage increment.
  31 Users:          Legacy s390-tools lsmem/chmem
  32
  33 What:           /sys/devices/system/memory/memoryX/phys_index
  34 Date:           September 2008
  35 Contact:        Badari Pulavarty <pbadari@us.ibm.com>
  36 Description:
  37                 The file /sys/devices/system/memory/memoryX/phys_index
  38                 is read-only and contains the section ID in hexadecimal
  39                 which is equivalent to decimal X contained in the
  40                 memory section directory name.
  41
  42 What:           /sys/devices/system/memory/memoryX/state
  43 Date:           September 2008
  44 Contact:        Badari Pulavarty <pbadari@us.ibm.com>
  45 Description:
  46                 The file /sys/devices/system/memory/memoryX/state
  47                 is read-write.  When read, it returns the online/offline
  48                 state of the memory block.  When written, root can toggle
  49                 the online/offline state of a memory block using the following
  50                 commands::
  51
  52                   # echo online > /sys/devices/system/memory/memoryX/state
  53                   # echo offline > /sys/devices/system/memory/memoryX/state
  54
  55                 On newer kernel versions, advanced states can be specified
  56                 when onlining to select a target zone: "online_movable"
  57                 selects the movable zone.  "online_kernel" selects the
  58                 applicable kernel zone (DMA, DMA32, or Normal).  However,
  59                 after successfully setting one of the advanced states,
  60                 reading the file will return "online"; the zone information
  61                 can be obtained via "valid_zones" instead.
  62
  63                 While onlining is unlikely to fail, there are no guarantees
  64                 that offlining will succeed.  Offlining is more likely to
  65                 succeed if "valid_zones" indicates "Movable".
  66 Users:          hotplug memory remove tools
  67                 http://www.ibm.com/developerworks/wikis/display/LinuxP/powerpc-utils
  68
  69
  70 What:           /sys/devices/system/memory/memoryX/valid_zones
  71 Date:           July 2014
  72 Contact:        Zhang Zhen <zhenzhang.zhang@huawei.com>
  73 Description:
  74                 The file /sys/devices/system/memory/memoryX/valid_zones is
  75                 read-only.
  76
  77                 For online memory blocks, it returns in which zone memory
  78                 provided by a memory block is managed.  If multiple zones
  79                 apply (not applicable for hotplugged memory), "None" is returned
  80                 and the memory block cannot be offlined.
  81
  82                 For offline memory blocks, it returns by which zone memory
  83                 provided by a memory block can be managed when onlining.
  84                 The first returned zone ("default") will be used when setting
  85                 the state of an offline memory block to "online".  Only one of
  86                 the kernel zones (DMA, DMA32, Normal) is applicable for a single
  87                 memory block.
  88
  89 What:           /sys/devices/system/memoryX/nodeY
  90 Date:           October 2009
  91 Contact:        Linux Memory Management list <linux-mm@kvack.org>
  92 Description:
  93                 When CONFIG_NUMA is enabled, a symbolic link that
  94                 points to the corresponding NUMA node directory.
  95
  96                 For example, the following symbolic link is created for
  97                 memory section 9 on node0:
  98
  99                 /sys/devices/system/memory/memory9/node0 -> ../../node/node0
 100
 101
 102 What:           /sys/devices/system/node/nodeX/memoryY
 103 Date:           September 2008
 104 Contact:        Gary Hade <garyhade@us.ibm.com>
 105 Description:
 106                 When CONFIG_NUMA is enabled
 107                 /sys/devices/system/node/nodeX/memoryY is a symbolic link that
 108                 points to the corresponding /sys/devices/system/memory/memoryY
 109                 memory section directory.  For example, the following symbolic
 110                 link is created for memory section 9 on node0.
 111
 112                 /sys/devices/system/node/node0/memory9 -> ../../memory/memory9
 113
 114 What:           /sys/devices/system/memory/crash_hotplug
 115 Date:           Aug 2023
 116 Contact:        Linux kernel mailing list <linux-kernel@vger.kernel.org>
 117 Description:
 118                 (RO) indicates whether or not the kernel updates relevant kexec
 119                 segments on memory hot un/plug and/or on/offline events, avoiding the
 120                 need to reload kdump kernel.