xref: /linux/Documentation/admin-guide/mm/memory-hotplug.rst (revision 8639a0c790add196bdd457c98437fec299cde60e)
1.. _admin_guide_memory_hotplug:
2
3==============
4Memory Hotplug
5==============
6
7:Created:							Jul 28 2007
8:Updated: Add description of notifier of memory hotplug:	Oct 11 2007
9
10This document is about memory hotplug including how-to-use and current status.
11Because Memory Hotplug is still under development, contents of this text will
12be changed often.
13
14.. contents:: :local:
15
16.. CONTENTS
17
18  1. Introduction
19    1.1 Purpose of memory hotplug
20    1.2. Phases of memory hotplug
21    1.3. Unit of Memory online/offline operation
22  2. Kernel Configuration
23  3. sysfs files for memory hotplug
24  4. Physical memory hot-add phase
25    4.1 Hardware(Firmware) Support
26    4.2 Notify memory hot-add event by hand
27  5. Logical Memory hot-add phase
28    5.1. State of memory
29    5.2. How to online memory
30  6. Logical memory remove
31    6.1 Memory offline and ZONE_MOVABLE
32    6.2. How to offline memory
33  7. Physical memory remove
34  9. Future Work List
35
36
37.. note::
38
39    (1) x86_64's has special implementation for memory hotplug.
40        This text does not describe it.
41    (2) This text assumes that sysfs is mounted at ``/sys``.
42
43
44Introduction
45============
46
47Purpose of memory hotplug
48-------------------------
49
50Memory Hotplug allows users to increase/decrease the amount of memory.
51Generally, there are two purposes.
52
53(A) For changing the amount of memory.
54    This is to allow a feature like capacity on demand.
55(B) For installing/removing DIMMs or NUMA-nodes physically.
56    This is to exchange DIMMs/NUMA-nodes, reduce power consumption, etc.
57
58(A) is required by highly virtualized environments and (B) is required by
59hardware which supports memory power management.
60
61Linux memory hotplug is designed for both purpose.
62
63Phases of memory hotplug
64------------------------
65
66There are 2 phases in Memory Hotplug:
67
68  1) Physical Memory Hotplug phase
69  2) Logical Memory Hotplug phase.
70
71The First phase is to communicate hardware/firmware and make/erase
72environment for hotplugged memory. Basically, this phase is necessary
73for the purpose (B), but this is good phase for communication between
74highly virtualized environments too.
75
76When memory is hotplugged, the kernel recognizes new memory, makes new memory
77management tables, and makes sysfs files for new memory's operation.
78
79If firmware supports notification of connection of new memory to OS,
80this phase is triggered automatically. ACPI can notify this event. If not,
81"probe" operation by system administration is used instead.
82(see :ref:`memory_hotplug_physical_mem`).
83
84Logical Memory Hotplug phase is to change memory state into
85available/unavailable for users. Amount of memory from user's view is
86changed by this phase. The kernel makes all memory in it as free pages
87when a memory range is available.
88
89In this document, this phase is described as online/offline.
90
91Logical Memory Hotplug phase is triggered by write of sysfs file by system
92administrator. For the hot-add case, it must be executed after Physical Hotplug
93phase by hand.
94(However, if you writes udev's hotplug scripts for memory hotplug, these
95phases can be execute in seamless way.)
96
97Unit of Memory online/offline operation
98---------------------------------------
99
100Memory hotplug uses SPARSEMEM memory model which allows memory to be divided
101into chunks of the same size. These chunks are called "sections". The size of
102a memory section is architecture dependent. For example, power uses 16MiB, ia64
103uses 1GiB.
104
105Memory sections are combined into chunks referred to as "memory blocks". The
106size of a memory block is architecture dependent and represents the logical
107unit upon which memory online/offline operations are to be performed. The
108default size of a memory block is the same as memory section size unless an
109architecture specifies otherwise. (see :ref:`memory_hotplug_sysfs_files`.)
110
111To determine the size (in bytes) of a memory block please read this file::
112
113  /sys/devices/system/memory/block_size_bytes
114
115Kernel Configuration
116====================
117
118To use memory hotplug feature, kernel must be compiled with following
119config options.
120
121- For all memory hotplug:
122    - Memory model -> Sparse Memory  (``CONFIG_SPARSEMEM``)
123    - Allow for memory hot-add       (``CONFIG_MEMORY_HOTPLUG``)
124
125- To enable memory removal, the following are also necessary:
126    - Allow for memory hot remove    (``CONFIG_MEMORY_HOTREMOVE``)
127    - Page Migration                 (``CONFIG_MIGRATION``)
128
129- For ACPI memory hotplug, the following are also necessary:
130    - Memory hotplug (under ACPI Support menu) (``CONFIG_ACPI_HOTPLUG_MEMORY``)
131    - This option can be kernel module.
132
133- As a related configuration, if your box has a feature of NUMA-node hotplug
134  via ACPI, then this option is necessary too.
135
136    - ACPI0004,PNP0A05 and PNP0A06 Container Driver (under ACPI Support menu)
137      (``CONFIG_ACPI_CONTAINER``).
138
139     This option can be kernel module too.
140
141
142.. _memory_hotplug_sysfs_files:
143
144sysfs files for memory hotplug
145==============================
146
147All memory blocks have their device information in sysfs.  Each memory block
148is described under ``/sys/devices/system/memory`` as::
149
150	/sys/devices/system/memory/memoryXXX
151
152where XXX is the memory block id.
153
154For the memory block covered by the sysfs directory.  It is expected that all
155memory sections in this range are present and no memory holes exist in the
156range. Currently there is no way to determine if there is a memory hole, but
157the existence of one should not affect the hotplug capabilities of the memory
158block.
159
160For example, assume 1GiB memory block size. A device for a memory starting at
1610x100000000 is ``/sys/device/system/memory/memory4``::
162
163	(0x100000000 / 1Gib = 4)
164
165This device covers address range [0x100000000 ... 0x140000000)
166
167Under each memory block, you can see 5 files:
168
169- ``/sys/devices/system/memory/memoryXXX/phys_index``
170- ``/sys/devices/system/memory/memoryXXX/phys_device``
171- ``/sys/devices/system/memory/memoryXXX/state``
172- ``/sys/devices/system/memory/memoryXXX/removable``
173- ``/sys/devices/system/memory/memoryXXX/valid_zones``
174
175=================== ============================================================
176``phys_index``      read-only and contains memory block id, same as XXX.
177``state``           read-write
178
179                    - at read:  contains online/offline state of memory.
180                    - at write: user can specify "online_kernel",
181
182                    "online_movable", "online", "offline" command
183                    which will be performed on all sections in the block.
184``phys_device``     read-only: designed to show the name of physical memory
185                    device.  This is not well implemented now.
186``removable``       read-only: contains an integer value indicating
187                    whether the memory block is removable or not
188                    removable.  A value of 1 indicates that the memory
189                    block is removable and a value of 0 indicates that
190                    it is not removable. A memory block is removable only if
191                    every section in the block is removable.
192``valid_zones``     read-only: designed to show which zones this memory block
193		    can be onlined to.
194
195		    The first column shows it`s default zone.
196
197		    "memory6/valid_zones: Normal Movable" shows this memoryblock
198		    can be onlined to ZONE_NORMAL by default and to ZONE_MOVABLE
199		    by online_movable.
200
201		    "memory7/valid_zones: Movable Normal" shows this memoryblock
202		    can be onlined to ZONE_MOVABLE by default and to ZONE_NORMAL
203		    by online_kernel.
204=================== ============================================================
205
206.. note::
207
208  These directories/files appear after physical memory hotplug phase.
209
210If CONFIG_NUMA is enabled the memoryXXX/ directories can also be accessed
211via symbolic links located in the ``/sys/devices/system/node/node*`` directories.
212
213For example::
214
215	/sys/devices/system/node/node0/memory9 -> ../../memory/memory9
216
217A backlink will also be created::
218
219	/sys/devices/system/memory/memory9/node0 -> ../../node/node0
220
221.. _memory_hotplug_physical_mem:
222
223Physical memory hot-add phase
224=============================
225
226Hardware(Firmware) Support
227--------------------------
228
229On x86_64/ia64 platform, memory hotplug by ACPI is supported.
230
231In general, the firmware (ACPI) which supports memory hotplug defines
232memory class object of _HID "PNP0C80". When a notify is asserted to PNP0C80,
233Linux's ACPI handler does hot-add memory to the system and calls a hotplug udev
234script. This will be done automatically.
235
236But scripts for memory hotplug are not contained in generic udev package(now).
237You may have to write it by yourself or online/offline memory by hand.
238Please see :ref:`memory_hotplug_how_to_online_memory` and
239:ref:`memory_hotplug_how_to_offline_memory`.
240
241If firmware supports NUMA-node hotplug, and defines an object _HID "ACPI0004",
242"PNP0A05", or "PNP0A06", notification is asserted to it, and ACPI handler
243calls hotplug code for all of objects which are defined in it.
244If memory device is found, memory hotplug code will be called.
245
246Notify memory hot-add event by hand
247-----------------------------------
248
249On some architectures, the firmware may not notify the kernel of a memory
250hotplug event.  Therefore, the memory "probe" interface is supported to
251explicitly notify the kernel.  This interface depends on
252CONFIG_ARCH_MEMORY_PROBE and can be configured on powerpc, sh, and x86
253if hotplug is supported, although for x86 this should be handled by ACPI
254notification.
255
256Probe interface is located at::
257
258	/sys/devices/system/memory/probe
259
260You can tell the physical address of new memory to the kernel by::
261
262	% echo start_address_of_new_memory > /sys/devices/system/memory/probe
263
264Then, [start_address_of_new_memory, start_address_of_new_memory +
265memory_block_size] memory range is hot-added. In this case, hotplug script is
266not called (in current implementation). You'll have to online memory by
267yourself.  Please see :ref:`memory_hotplug_how_to_online_memory`.
268
269Logical Memory hot-add phase
270============================
271
272State of memory
273---------------
274
275To see (online/offline) state of a memory block, read 'state' file::
276
277	% cat /sys/device/system/memory/memoryXXX/state
278
279
280- If the memory block is online, you'll read "online".
281- If the memory block is offline, you'll read "offline".
282
283
284.. _memory_hotplug_how_to_online_memory:
285
286How to online memory
287--------------------
288
289When the memory is hot-added, the kernel decides whether or not to "online"
290it according to the policy which can be read from "auto_online_blocks" file::
291
292	% cat /sys/devices/system/memory/auto_online_blocks
293
294The default depends on the CONFIG_MEMORY_HOTPLUG_DEFAULT_ONLINE kernel config
295option. If it is disabled the default is "offline" which means the newly added
296memory is not in a ready-to-use state and you have to "online" the newly added
297memory blocks manually. Automatic onlining can be requested by writing "online"
298to "auto_online_blocks" file::
299
300	% echo online > /sys/devices/system/memory/auto_online_blocks
301
302This sets a global policy and impacts all memory blocks that will subsequently
303be hotplugged. Currently offline blocks keep their state. It is possible, under
304certain circumstances, that some memory blocks will be added but will fail to
305online. User space tools can check their "state" files
306(``/sys/devices/system/memory/memoryXXX/state``) and try to online them manually.
307
308If the automatic onlining wasn't requested, failed, or some memory block was
309offlined it is possible to change the individual block's state by writing to the
310"state" file::
311
312	% echo online > /sys/devices/system/memory/memoryXXX/state
313
314This onlining will not change the ZONE type of the target memory block,
315If the memory block doesn't belong to any zone an appropriate kernel zone
316(usually ZONE_NORMAL) will be used unless movable_node kernel command line
317option is specified when ZONE_MOVABLE will be used.
318
319You can explicitly request to associate it with ZONE_MOVABLE by::
320
321	% echo online_movable > /sys/devices/system/memory/memoryXXX/state
322
323.. note:: current limit: this memory block must be adjacent to ZONE_MOVABLE
324
325Or you can explicitly request a kernel zone (usually ZONE_NORMAL) by::
326
327	% echo online_kernel > /sys/devices/system/memory/memoryXXX/state
328
329.. note:: current limit: this memory block must be adjacent to ZONE_NORMAL
330
331An explicit zone onlining can fail (e.g. when the range is already within
332and existing and incompatible zone already).
333
334After this, memory block XXX's state will be 'online' and the amount of
335available memory will be increased.
336
337This may be changed in future.
338
339Logical memory remove
340=====================
341
342Memory offline and ZONE_MOVABLE
343-------------------------------
344
345Memory offlining is more complicated than memory online. Because memory offline
346has to make the whole memory block be unused, memory offline can fail if
347the memory block includes memory which cannot be freed.
348
349In general, memory offline can use 2 techniques.
350
351(1) reclaim and free all memory in the memory block.
352(2) migrate all pages in the memory block.
353
354In the current implementation, Linux's memory offline uses method (2), freeing
355all  pages in the memory block by page migration. But not all pages are
356migratable. Under current Linux, migratable pages are anonymous pages and
357page caches. For offlining a memory block by migration, the kernel has to
358guarantee that the memory block contains only migratable pages.
359
360Now, a boot option for making a memory block which consists of migratable pages
361is supported. By specifying "kernelcore=" or "movablecore=" boot option, you can
362create ZONE_MOVABLE...a zone which is just used for movable pages.
363(See also Documentation/admin-guide/kernel-parameters.rst)
364
365Assume the system has "TOTAL" amount of memory at boot time, this boot option
366creates ZONE_MOVABLE as following.
367
3681) When kernelcore=YYYY boot option is used,
369   Size of memory not for movable pages (not for offline) is YYYY.
370   Size of memory for movable pages (for offline) is TOTAL-YYYY.
371
3722) When movablecore=ZZZZ boot option is used,
373   Size of memory not for movable pages (not for offline) is TOTAL - ZZZZ.
374   Size of memory for movable pages (for offline) is ZZZZ.
375
376.. note::
377
378   Unfortunately, there is no information to show which memory block belongs
379   to ZONE_MOVABLE. This is TBD.
380
381.. _memory_hotplug_how_to_offline_memory:
382
383How to offline memory
384---------------------
385
386You can offline a memory block by using the same sysfs interface that was used
387in memory onlining::
388
389	% echo offline > /sys/devices/system/memory/memoryXXX/state
390
391If offline succeeds, the state of the memory block is changed to be "offline".
392If it fails, some error core (like -EBUSY) will be returned by the kernel.
393Even if a memory block does not belong to ZONE_MOVABLE, you can try to offline
394it.  If it doesn't contain 'unmovable' memory, you'll get success.
395
396A memory block under ZONE_MOVABLE is considered to be able to be offlined
397easily.  But under some busy state, it may return -EBUSY. Even if a memory
398block cannot be offlined due to -EBUSY, you can retry offlining it and may be
399able to offline it (or not). (For example, a page is referred to by some kernel
400internal call and released soon.)
401
402Consideration:
403  Memory hotplug's design direction is to make the possibility of memory
404  offlining higher and to guarantee unplugging memory under any situation. But
405  it needs more work. Returning -EBUSY under some situation may be good because
406  the user can decide to retry more or not by himself. Currently, memory
407  offlining code does some amount of retry with 120 seconds timeout.
408
409Physical memory remove
410======================
411
412Need more implementation yet....
413 - Notification completion of remove works by OS to firmware.
414 - Guard from remove if not yet.
415
416Future Work
417===========
418
419  - allowing memory hot-add to ZONE_MOVABLE. maybe we need some switch like
420    sysctl or new control file.
421  - showing memory block and physical device relationship.
422  - test and make it better memory offlining.
423  - support HugeTLB page migration and offlining.
424  - memmap removing at memory offline.
425  - physical remove memory.
426