1============== 2Memory Hotplug 3============== 4 5Created: Jul 28 2007 6Add description of notifier of memory hotplug Oct 11 2007 7 8This document is about memory hotplug including how-to-use and current status. 9Because Memory Hotplug is still under development, contents of this text will 10be changed often. 11 121. Introduction 13 1.1 purpose of memory hotplug 14 1.2. Phases of memory hotplug 15 1.3. Unit of Memory online/offline operation 162. Kernel Configuration 173. sysfs files for memory hotplug 184. Physical memory hot-add phase 19 4.1 Hardware(Firmware) Support 20 4.2 Notify memory hot-add event by hand 215. Logical Memory hot-add phase 22 5.1. State of memory 23 5.2. How to online memory 246. Logical memory remove 25 6.1 Memory offline and ZONE_MOVABLE 26 6.2. How to offline memory 277. Physical memory remove 288. Memory hotplug event notifier 299. Future Work List 30 31Note(1): x86_64's has special implementation for memory hotplug. 32 This text does not describe it. 33Note(2): This text assumes that sysfs is mounted at /sys. 34 35 36--------------- 371. Introduction 38--------------- 39 401.1 purpose of memory hotplug 41------------ 42Memory Hotplug allows users to increase/decrease the amount of memory. 43Generally, there are two purposes. 44 45(A) For changing the amount of memory. 46 This is to allow a feature like capacity on demand. 47(B) For installing/removing DIMMs or NUMA-nodes physically. 48 This is to exchange DIMMs/NUMA-nodes, reduce power consumption, etc. 49 50(A) is required by highly virtualized environments and (B) is required by 51hardware which supports memory power management. 52 53Linux memory hotplug is designed for both purpose. 54 55 561.2. Phases of memory hotplug 57--------------- 58There are 2 phases in Memory Hotplug. 59 1) Physical Memory Hotplug phase 60 2) Logical Memory Hotplug phase. 61 62The First phase is to communicate hardware/firmware and make/erase 63environment for hotplugged memory. Basically, this phase is necessary 64for the purpose (B), but this is good phase for communication between 65highly virtualized environments too. 66 67When memory is hotplugged, the kernel recognizes new memory, makes new memory 68management tables, and makes sysfs files for new memory's operation. 69 70If firmware supports notification of connection of new memory to OS, 71this phase is triggered automatically. ACPI can notify this event. If not, 72"probe" operation by system administration is used instead. 73(see Section 4.). 74 75Logical Memory Hotplug phase is to change memory state into 76available/unavailable for users. Amount of memory from user's view is 77changed by this phase. The kernel makes all memory in it as free pages 78when a memory range is available. 79 80In this document, this phase is described as online/offline. 81 82Logical Memory Hotplug phase is triggered by write of sysfs file by system 83administrator. For the hot-add case, it must be executed after Physical Hotplug 84phase by hand. 85(However, if you writes udev's hotplug scripts for memory hotplug, these 86 phases can be execute in seamless way.) 87 88 891.3. Unit of Memory online/offline operation 90------------ 91Memory hotplug uses SPARSEMEM memory model which allows memory to be divided 92into chunks of the same size. These chunks are called "sections". The size of 93a memory section is architecture dependent. For example, power uses 16MiB, ia64 94uses 1GiB. 95 96Memory sections are combined into chunks referred to as "memory blocks". The 97size of a memory block is architecture dependent and represents the logical 98unit upon which memory online/offline operations are to be performed. The 99default size of a memory block is the same as memory section size unless an 100architecture specifies otherwise. (see Section 3.) 101 102To determine the size (in bytes) of a memory block please read this file: 103 104/sys/devices/system/memory/block_size_bytes 105 106 107----------------------- 1082. Kernel Configuration 109----------------------- 110To use memory hotplug feature, kernel must be compiled with following 111config options. 112 113- For all memory hotplug 114 Memory model -> Sparse Memory (CONFIG_SPARSEMEM) 115 Allow for memory hot-add (CONFIG_MEMORY_HOTPLUG) 116 117- To enable memory removal, the followings are also necessary 118 Allow for memory hot remove (CONFIG_MEMORY_HOTREMOVE) 119 Page Migration (CONFIG_MIGRATION) 120 121- For ACPI memory hotplug, the followings are also necessary 122 Memory hotplug (under ACPI Support menu) (CONFIG_ACPI_HOTPLUG_MEMORY) 123 This option can be kernel module. 124 125- As a related configuration, if your box has a feature of NUMA-node hotplug 126 via ACPI, then this option is necessary too. 127 ACPI0004,PNP0A05 and PNP0A06 Container Driver (under ACPI Support menu) 128 (CONFIG_ACPI_CONTAINER). 129 This option can be kernel module too. 130 131 132-------------------------------- 1333 sysfs files for memory hotplug 134-------------------------------- 135All memory blocks have their device information in sysfs. Each memory block 136is described under /sys/devices/system/memory as 137 138/sys/devices/system/memory/memoryXXX 139(XXX is the memory block id.) 140 141For the memory block covered by the sysfs directory. It is expected that all 142memory sections in this range are present and no memory holes exist in the 143range. Currently there is no way to determine if there is a memory hole, but 144the existence of one should not affect the hotplug capabilities of the memory 145block. 146 147For example, assume 1GiB memory block size. A device for a memory starting at 1480x100000000 is /sys/device/system/memory/memory4 149(0x100000000 / 1Gib = 4) 150This device covers address range [0x100000000 ... 0x140000000) 151 152Under each memory block, you can see 5 files: 153 154/sys/devices/system/memory/memoryXXX/phys_index 155/sys/devices/system/memory/memoryXXX/phys_device 156/sys/devices/system/memory/memoryXXX/state 157/sys/devices/system/memory/memoryXXX/removable 158/sys/devices/system/memory/memoryXXX/valid_zones 159 160'phys_index' : read-only and contains memory block id, same as XXX. 161'state' : read-write 162 at read: contains online/offline state of memory. 163 at write: user can specify "online_kernel", 164 "online_movable", "online", "offline" command 165 which will be performed on all sections in the block. 166'phys_device' : read-only: designed to show the name of physical memory 167 device. This is not well implemented now. 168'removable' : read-only: contains an integer value indicating 169 whether the memory block is removable or not 170 removable. A value of 1 indicates that the memory 171 block is removable and a value of 0 indicates that 172 it is not removable. A memory block is removable only if 173 every section in the block is removable. 174'valid_zones' : read-only: designed to show which zones this memory block 175 can be onlined to. 176 The first column shows it's default zone. 177 "memory6/valid_zones: Normal Movable" shows this memoryblock 178 can be onlined to ZONE_NORMAL by default and to ZONE_MOVABLE 179 by online_movable. 180 "memory7/valid_zones: Movable Normal" shows this memoryblock 181 can be onlined to ZONE_MOVABLE by default and to ZONE_NORMAL 182 by online_kernel. 183 184NOTE: 185 These directories/files appear after physical memory hotplug phase. 186 187If CONFIG_NUMA is enabled the memoryXXX/ directories can also be accessed 188via symbolic links located in the /sys/devices/system/node/node* directories. 189 190For example: 191/sys/devices/system/node/node0/memory9 -> ../../memory/memory9 192 193A backlink will also be created: 194/sys/devices/system/memory/memory9/node0 -> ../../node/node0 195 196 197-------------------------------- 1984. Physical memory hot-add phase 199-------------------------------- 200 2014.1 Hardware(Firmware) Support 202------------ 203On x86_64/ia64 platform, memory hotplug by ACPI is supported. 204 205In general, the firmware (ACPI) which supports memory hotplug defines 206memory class object of _HID "PNP0C80". When a notify is asserted to PNP0C80, 207Linux's ACPI handler does hot-add memory to the system and calls a hotplug udev 208script. This will be done automatically. 209 210But scripts for memory hotplug are not contained in generic udev package(now). 211You may have to write it by yourself or online/offline memory by hand. 212Please see "How to online memory", "How to offline memory" in this text. 213 214If firmware supports NUMA-node hotplug, and defines an object _HID "ACPI0004", 215"PNP0A05", or "PNP0A06", notification is asserted to it, and ACPI handler 216calls hotplug code for all of objects which are defined in it. 217If memory device is found, memory hotplug code will be called. 218 219 2204.2 Notify memory hot-add event by hand 221------------ 222On some architectures, the firmware may not notify the kernel of a memory 223hotplug event. Therefore, the memory "probe" interface is supported to 224explicitly notify the kernel. This interface depends on 225CONFIG_ARCH_MEMORY_PROBE and can be configured on powerpc, sh, and x86 226if hotplug is supported, although for x86 this should be handled by ACPI 227notification. 228 229Probe interface is located at 230/sys/devices/system/memory/probe 231 232You can tell the physical address of new memory to the kernel by 233 234% echo start_address_of_new_memory > /sys/devices/system/memory/probe 235 236Then, [start_address_of_new_memory, start_address_of_new_memory + 237memory_block_size] memory range is hot-added. In this case, hotplug script is 238not called (in current implementation). You'll have to online memory by 239yourself. Please see "How to online memory" in this text. 240 241 242------------------------------ 2435. Logical Memory hot-add phase 244------------------------------ 245 2465.1. State of memory 247------------ 248To see (online/offline) state of a memory block, read 'state' file. 249 250% cat /sys/device/system/memory/memoryXXX/state 251 252 253If the memory block is online, you'll read "online". 254If the memory block is offline, you'll read "offline". 255 256 2575.2. How to online memory 258------------ 259Even if the memory is hot-added, it is not at ready-to-use state. 260For using newly added memory, you have to "online" the memory block. 261 262For onlining, you have to write "online" to the memory block's state file as: 263 264% echo online > /sys/devices/system/memory/memoryXXX/state 265 266This onlining will not change the ZONE type of the target memory block, 267If the memory block is in ZONE_NORMAL, you can change it to ZONE_MOVABLE: 268 269% echo online_movable > /sys/devices/system/memory/memoryXXX/state 270(NOTE: current limit: this memory block must be adjacent to ZONE_MOVABLE) 271 272And if the memory block is in ZONE_MOVABLE, you can change it to ZONE_NORMAL: 273 274% echo online_kernel > /sys/devices/system/memory/memoryXXX/state 275(NOTE: current limit: this memory block must be adjacent to ZONE_NORMAL) 276 277After this, memory block XXX's state will be 'online' and the amount of 278available memory will be increased. 279 280Currently, newly added memory is added as ZONE_NORMAL (for powerpc, ZONE_DMA). 281This may be changed in future. 282 283 284 285------------------------ 2866. Logical memory remove 287------------------------ 288 2896.1 Memory offline and ZONE_MOVABLE 290------------ 291Memory offlining is more complicated than memory online. Because memory offline 292has to make the whole memory block be unused, memory offline can fail if 293the memory block includes memory which cannot be freed. 294 295In general, memory offline can use 2 techniques. 296 297(1) reclaim and free all memory in the memory block. 298(2) migrate all pages in the memory block. 299 300In the current implementation, Linux's memory offline uses method (2), freeing 301all pages in the memory block by page migration. But not all pages are 302migratable. Under current Linux, migratable pages are anonymous pages and 303page caches. For offlining a memory block by migration, the kernel has to 304guarantee that the memory block contains only migratable pages. 305 306Now, a boot option for making a memory block which consists of migratable pages 307is supported. By specifying "kernelcore=" or "movablecore=" boot option, you can 308create ZONE_MOVABLE...a zone which is just used for movable pages. 309(See also Documentation/kernel-parameters.txt) 310 311Assume the system has "TOTAL" amount of memory at boot time, this boot option 312creates ZONE_MOVABLE as following. 313 3141) When kernelcore=YYYY boot option is used, 315 Size of memory not for movable pages (not for offline) is YYYY. 316 Size of memory for movable pages (for offline) is TOTAL-YYYY. 317 3182) When movablecore=ZZZZ boot option is used, 319 Size of memory not for movable pages (not for offline) is TOTAL - ZZZZ. 320 Size of memory for movable pages (for offline) is ZZZZ. 321 322 323Note: Unfortunately, there is no information to show which memory block belongs 324to ZONE_MOVABLE. This is TBD. 325 326 3276.2. How to offline memory 328------------ 329You can offline a memory block by using the same sysfs interface that was used 330in memory onlining. 331 332% echo offline > /sys/devices/system/memory/memoryXXX/state 333 334If offline succeeds, the state of the memory block is changed to be "offline". 335If it fails, some error core (like -EBUSY) will be returned by the kernel. 336Even if a memory block does not belong to ZONE_MOVABLE, you can try to offline 337it. If it doesn't contain 'unmovable' memory, you'll get success. 338 339A memory block under ZONE_MOVABLE is considered to be able to be offlined 340easily. But under some busy state, it may return -EBUSY. Even if a memory 341block cannot be offlined due to -EBUSY, you can retry offlining it and may be 342able to offline it (or not). (For example, a page is referred to by some kernel 343internal call and released soon.) 344 345Consideration: 346Memory hotplug's design direction is to make the possibility of memory offlining 347higher and to guarantee unplugging memory under any situation. But it needs 348more work. Returning -EBUSY under some situation may be good because the user 349can decide to retry more or not by himself. Currently, memory offlining code 350does some amount of retry with 120 seconds timeout. 351 352------------------------- 3537. Physical memory remove 354------------------------- 355Need more implementation yet.... 356 - Notification completion of remove works by OS to firmware. 357 - Guard from remove if not yet. 358 359-------------------------------- 3608. Memory hotplug event notifier 361-------------------------------- 362Hotplugging events are sent to a notification queue. 363 364There are six types of notification defined in include/linux/memory.h: 365 366MEM_GOING_ONLINE 367 Generated before new memory becomes available in order to be able to 368 prepare subsystems to handle memory. The page allocator is still unable 369 to allocate from the new memory. 370 371MEM_CANCEL_ONLINE 372 Generated if MEMORY_GOING_ONLINE fails. 373 374MEM_ONLINE 375 Generated when memory has successfully brought online. The callback may 376 allocate pages from the new memory. 377 378MEM_GOING_OFFLINE 379 Generated to begin the process of offlining memory. Allocations are no 380 longer possible from the memory but some of the memory to be offlined 381 is still in use. The callback can be used to free memory known to a 382 subsystem from the indicated memory block. 383 384MEM_CANCEL_OFFLINE 385 Generated if MEMORY_GOING_OFFLINE fails. Memory is available again from 386 the memory block that we attempted to offline. 387 388MEM_OFFLINE 389 Generated after offlining memory is complete. 390 391A callback routine can be registered by calling 392 393 hotplug_memory_notifier(callback_func, priority) 394 395Callback functions with higher values of priority are called before callback 396functions with lower values. 397 398A callback function must have the following prototype: 399 400 int callback_func( 401 struct notifier_block *self, unsigned long action, void *arg); 402 403The first argument of the callback function (self) is a pointer to the block 404of the notifier chain that points to the callback function itself. 405The second argument (action) is one of the event types described above. 406The third argument (arg) passes a pointer of struct memory_notify. 407 408struct memory_notify { 409 unsigned long start_pfn; 410 unsigned long nr_pages; 411 int status_change_nid_normal; 412 int status_change_nid_high; 413 int status_change_nid; 414} 415 416start_pfn is start_pfn of online/offline memory. 417nr_pages is # of pages of online/offline memory. 418status_change_nid_normal is set node id when N_NORMAL_MEMORY of nodemask 419is (will be) set/clear, if this is -1, then nodemask status is not changed. 420status_change_nid_high is set node id when N_HIGH_MEMORY of nodemask 421is (will be) set/clear, if this is -1, then nodemask status is not changed. 422status_change_nid is set node id when N_MEMORY of nodemask is (will be) 423set/clear. It means a new(memoryless) node gets new memory by online and a 424node loses all memory. If this is -1, then nodemask status is not changed. 425If status_changed_nid* >= 0, callback should create/discard structures for the 426node if necessary. 427 428The callback routine shall return one of the values 429NOTIFY_DONE, NOTIFY_OK, NOTIFY_BAD, NOTIFY_STOP 430defined in include/linux/notifier.h 431 432NOTIFY_DONE and NOTIFY_OK have no effect on the further processing. 433 434NOTIFY_BAD is used as response to the MEM_GOING_ONLINE, MEM_GOING_OFFLINE, 435MEM_ONLINE, or MEM_OFFLINE action to cancel hotplugging. It stops 436further processing of the notification queue. 437 438NOTIFY_STOP stops further processing of the notification queue. 439 440-------------- 4419. Future Work 442-------------- 443 - allowing memory hot-add to ZONE_MOVABLE. maybe we need some switch like 444 sysctl or new control file. 445 - showing memory block and physical device relationship. 446 - test and make it better memory offlining. 447 - support HugeTLB page migration and offlining. 448 - memmap removing at memory offline. 449 - physical remove memory. 450 451