Based on kernel version 3.16. Page generated on 2014-08-06 21:36 EST.
1 Memory Resource Controller 2 3 NOTE: The Memory Resource Controller has generically been referred to as the 4 memory controller in this document. Do not confuse memory controller 5 used here with the memory controller that is used in hardware. 6 7 (For editors) 8 In this document: 9 When we mention a cgroup (cgroupfs's directory) with memory controller, 10 we call it "memory cgroup". When you see git-log and source code, you'll 11 see patch's title and function names tend to use "memcg". 12 In this document, we avoid using it. 13 14 Benefits and Purpose of the memory controller 15 16 The memory controller isolates the memory behaviour of a group of tasks 17 from the rest of the system. The article on LWN  mentions some probable 18 uses of the memory controller. The memory controller can be used to 19 20 a. Isolate an application or a group of applications 21 Memory-hungry applications can be isolated and limited to a smaller 22 amount of memory. 23 b. Create a cgroup with a limited amount of memory; this can be used 24 as a good alternative to booting with mem=XXXX. 25 c. Virtualization solutions can control the amount of memory they want 26 to assign to a virtual machine instance. 27 d. A CD/DVD burner could control the amount of memory used by the 28 rest of the system to ensure that burning does not fail due to lack 29 of available memory. 30 e. There are several other use cases; find one or use the controller just 31 for fun (to learn and hack on the VM subsystem). 32 33 Current Status: linux-2.6.34-mmotm(development version of 2010/April) 34 35 Features: 36 - accounting anonymous pages, file caches, swap caches usage and limiting them. 37 - pages are linked to per-memcg LRU exclusively, and there is no global LRU. 38 - optionally, memory+swap usage can be accounted and limited. 39 - hierarchical accounting 40 - soft limit 41 - moving (recharging) account at moving a task is selectable. 42 - usage threshold notifier 43 - memory pressure notifier 44 - oom-killer disable knob and oom-notifier 45 - Root cgroup has no limit controls. 46 47 Kernel memory support is a work in progress, and the current version provides 48 basically functionality. (See Section 2.7) 49 50 Brief summary of control files. 51 52 tasks # attach a task(thread) and show list of threads 53 cgroup.procs # show list of processes 54 cgroup.event_control # an interface for event_fd() 55 memory.usage_in_bytes # show current res_counter usage for memory 56 (See 5.5 for details) 57 memory.memsw.usage_in_bytes # show current res_counter usage for memory+Swap 58 (See 5.5 for details) 59 memory.limit_in_bytes # set/show limit of memory usage 60 memory.memsw.limit_in_bytes # set/show limit of memory+Swap usage 61 memory.failcnt # show the number of memory usage hits limits 62 memory.memsw.failcnt # show the number of memory+Swap hits limits 63 memory.max_usage_in_bytes # show max memory usage recorded 64 memory.memsw.max_usage_in_bytes # show max memory+Swap usage recorded 65 memory.soft_limit_in_bytes # set/show soft limit of memory usage 66 memory.stat # show various statistics 67 memory.use_hierarchy # set/show hierarchical account enabled 68 memory.force_empty # trigger forced move charge to parent 69 memory.pressure_level # set memory pressure notifications 70 memory.swappiness # set/show swappiness parameter of vmscan 71 (See sysctl's vm.swappiness) 72 memory.move_charge_at_immigrate # set/show controls of moving charges 73 memory.oom_control # set/show oom controls. 74 memory.numa_stat # show the number of memory usage per numa node 75 76 memory.kmem.limit_in_bytes # set/show hard limit for kernel memory 77 memory.kmem.usage_in_bytes # show current kernel memory allocation 78 memory.kmem.failcnt # show the number of kernel memory usage hits limits 79 memory.kmem.max_usage_in_bytes # show max kernel memory usage recorded 80 81 memory.kmem.tcp.limit_in_bytes # set/show hard limit for tcp buf memory 82 memory.kmem.tcp.usage_in_bytes # show current tcp buf memory allocation 83 memory.kmem.tcp.failcnt # show the number of tcp buf memory usage hits limits 84 memory.kmem.tcp.max_usage_in_bytes # show max tcp buf memory usage recorded 85 86 1. History 87 88 The memory controller has a long history. A request for comments for the memory 89 controller was posted by Balbir Singh . At the time the RFC was posted 90 there were several implementations for memory control. The goal of the 91 RFC was to build consensus and agreement for the minimal features required 92 for memory control. The first RSS controller was posted by Balbir Singh 93 in Feb 2007. Pavel Emelianov  has since posted three versions of the 94 RSS controller. At OLS, at the resource management BoF, everyone suggested 95 that we handle both page cache and RSS together. Another request was raised 96 to allow user space handling of OOM. The current memory controller is 97 at version 6; it combines both mapped (RSS) and unmapped Page 98 Cache Control . 99 100 2. Memory Control 101 102 Memory is a unique resource in the sense that it is present in a limited 103 amount. If a task requires a lot of CPU processing, the task can spread 104 its processing over a period of hours, days, months or years, but with 105 memory, the same physical memory needs to be reused to accomplish the task. 106 107 The memory controller implementation has been divided into phases. These 108 are: 109 110 1. Memory controller 111 2. mlock(2) controller 112 3. Kernel user memory accounting and slab control 113 4. user mappings length controller 114 115 The memory controller is the first controller developed. 116 117 2.1. Design 118 119 The core of the design is a counter called the res_counter. The res_counter 120 tracks the current memory usage and limit of the group of processes associated 121 with the controller. Each cgroup has a memory controller specific data 122 structure (mem_cgroup) associated with it. 123 124 2.2. Accounting 125 126 +--------------------+ 127 | mem_cgroup | 128 | (res_counter) | 129 +--------------------+ 130 / ^ \ 131 / | \ 132 +---------------+ | +---------------+ 133 | mm_struct | |.... | mm_struct | 134 | | | | | 135 +---------------+ | +---------------+ 136 | 137 + --------------+ 138 | 139 +---------------+ +------+--------+ 140 | page +----------> page_cgroup| 141 | | | | 142 +---------------+ +---------------+ 143 144 (Figure 1: Hierarchy of Accounting) 145 146 147 Figure 1 shows the important aspects of the controller 148 149 1. Accounting happens per cgroup 150 2. Each mm_struct knows about which cgroup it belongs to 151 3. Each page has a pointer to the page_cgroup, which in turn knows the 152 cgroup it belongs to 153 154 The accounting is done as follows: mem_cgroup_charge_common() is invoked to 155 set up the necessary data structures and check if the cgroup that is being 156 charged is over its limit. If it is, then reclaim is invoked on the cgroup. 157 More details can be found in the reclaim section of this document. 158 If everything goes well, a page meta-data-structure called page_cgroup is 159 updated. page_cgroup has its own LRU on cgroup. 160 (*) page_cgroup structure is allocated at boot/memory-hotplug time. 161 162 2.2.1 Accounting details 163 164 All mapped anon pages (RSS) and cache pages (Page Cache) are accounted. 165 Some pages which are never reclaimable and will not be on the LRU 166 are not accounted. We just account pages under usual VM management. 167 168 RSS pages are accounted at page_fault unless they've already been accounted 169 for earlier. A file page will be accounted for as Page Cache when it's 170 inserted into inode (radix-tree). While it's mapped into the page tables of 171 processes, duplicate accounting is carefully avoided. 172 173 An RSS page is unaccounted when it's fully unmapped. A PageCache page is 174 unaccounted when it's removed from radix-tree. Even if RSS pages are fully 175 unmapped (by kswapd), they may exist as SwapCache in the system until they 176 are really freed. Such SwapCaches are also accounted. 177 A swapped-in page is not accounted until it's mapped. 178 179 Note: The kernel does swapin-readahead and reads multiple swaps at once. 180 This means swapped-in pages may contain pages for other tasks than a task 181 causing page fault. So, we avoid accounting at swap-in I/O. 182 183 At page migration, accounting information is kept. 184 185 Note: we just account pages-on-LRU because our purpose is to control amount 186 of used pages; not-on-LRU pages tend to be out-of-control from VM view. 187 188 2.3 Shared Page Accounting 189 190 Shared pages are accounted on the basis of the first touch approach. The 191 cgroup that first touches a page is accounted for the page. The principle 192 behind this approach is that a cgroup that aggressively uses a shared 193 page will eventually get charged for it (once it is uncharged from 194 the cgroup that brought it in -- this will happen on memory pressure). 195 196 But see section 8.2: when moving a task to another cgroup, its pages may 197 be recharged to the new cgroup, if move_charge_at_immigrate has been chosen. 198 199 Exception: If CONFIG_MEMCG_SWAP is not used. 200 When you do swapoff and make swapped-out pages of shmem(tmpfs) to 201 be backed into memory in force, charges for pages are accounted against the 202 caller of swapoff rather than the users of shmem. 203 204 2.4 Swap Extension (CONFIG_MEMCG_SWAP) 205 206 Swap Extension allows you to record charge for swap. A swapped-in page is 207 charged back to original page allocator if possible. 208 209 When swap is accounted, following files are added. 210 - memory.memsw.usage_in_bytes. 211 - memory.memsw.limit_in_bytes. 212 213 memsw means memory+swap. Usage of memory+swap is limited by 214 memsw.limit_in_bytes. 215 216 Example: Assume a system with 4G of swap. A task which allocates 6G of memory 217 (by mistake) under 2G memory limitation will use all swap. 218 In this case, setting memsw.limit_in_bytes=3G will prevent bad use of swap. 219 By using the memsw limit, you can avoid system OOM which can be caused by swap 220 shortage. 221 222 * why 'memory+swap' rather than swap. 223 The global LRU(kswapd) can swap out arbitrary pages. Swap-out means 224 to move account from memory to swap...there is no change in usage of 225 memory+swap. In other words, when we want to limit the usage of swap without 226 affecting global LRU, memory+swap limit is better than just limiting swap from 227 an OS point of view. 228 229 * What happens when a cgroup hits memory.memsw.limit_in_bytes 230 When a cgroup hits memory.memsw.limit_in_bytes, it's useless to do swap-out 231 in this cgroup. Then, swap-out will not be done by cgroup routine and file 232 caches are dropped. But as mentioned above, global LRU can do swapout memory 233 from it for sanity of the system's memory management state. You can't forbid 234 it by cgroup. 235 236 2.5 Reclaim 237 238 Each cgroup maintains a per cgroup LRU which has the same structure as 239 global VM. When a cgroup goes over its limit, we first try 240 to reclaim memory from the cgroup so as to make space for the new 241 pages that the cgroup has touched. If the reclaim is unsuccessful, 242 an OOM routine is invoked to select and kill the bulkiest task in the 243 cgroup. (See 10. OOM Control below.) 244 245 The reclaim algorithm has not been modified for cgroups, except that 246 pages that are selected for reclaiming come from the per-cgroup LRU 247 list. 248 249 NOTE: Reclaim does not work for the root cgroup, since we cannot set any 250 limits on the root cgroup. 251 252 Note2: When panic_on_oom is set to "2", the whole system will panic. 253 254 When oom event notifier is registered, event will be delivered. 255 (See oom_control section) 256 257 2.6 Locking 258 259 lock_page_cgroup()/unlock_page_cgroup() should not be called under 260 mapping->tree_lock. 261 262 Other lock order is following: 263 PG_locked. 264 mm->page_table_lock 265 zone->lru_lock 266 lock_page_cgroup. 267 In many cases, just lock_page_cgroup() is called. 268 per-zone-per-cgroup LRU (cgroup's private LRU) is just guarded by 269 zone->lru_lock, it has no lock of its own. 270 271 2.7 Kernel Memory Extension (CONFIG_MEMCG_KMEM) 272 273 WARNING: Current implementation lacks reclaim support. That means allocation 274 attempts will fail when close to the limit even if there are plenty of 275 kmem available for reclaim. That makes this option unusable in real 276 life so DO NOT SELECT IT unless for development purposes. 277 278 With the Kernel memory extension, the Memory Controller is able to limit 279 the amount of kernel memory used by the system. Kernel memory is fundamentally 280 different than user memory, since it can't be swapped out, which makes it 281 possible to DoS the system by consuming too much of this precious resource. 282 283 Kernel memory won't be accounted at all until limit on a group is set. This 284 allows for existing setups to continue working without disruption. The limit 285 cannot be set if the cgroup have children, or if there are already tasks in the 286 cgroup. Attempting to set the limit under those conditions will return -EBUSY. 287 When use_hierarchy == 1 and a group is accounted, its children will 288 automatically be accounted regardless of their limit value. 289 290 After a group is first limited, it will be kept being accounted until it 291 is removed. The memory limitation itself, can of course be removed by writing 292 -1 to memory.kmem.limit_in_bytes. In this case, kmem will be accounted, but not 293 limited. 294 295 Kernel memory limits are not imposed for the root cgroup. Usage for the root 296 cgroup may or may not be accounted. The memory used is accumulated into 297 memory.kmem.usage_in_bytes, or in a separate counter when it makes sense. 298 (currently only for tcp). 299 The main "kmem" counter is fed into the main counter, so kmem charges will 300 also be visible from the user counter. 301 302 Currently no soft limit is implemented for kernel memory. It is future work 303 to trigger slab reclaim when those limits are reached. 304 305 2.7.1 Current Kernel Memory resources accounted 306 307 * stack pages: every process consumes some stack pages. By accounting into 308 kernel memory, we prevent new processes from being created when the kernel 309 memory usage is too high. 310 311 * slab pages: pages allocated by the SLAB or SLUB allocator are tracked. A copy 312 of each kmem_cache is created every time the cache is touched by the first time 313 from inside the memcg. The creation is done lazily, so some objects can still be 314 skipped while the cache is being created. All objects in a slab page should 315 belong to the same memcg. This only fails to hold when a task is migrated to a 316 different memcg during the page allocation by the cache. 317 318 * sockets memory pressure: some sockets protocols have memory pressure 319 thresholds. The Memory Controller allows them to be controlled individually 320 per cgroup, instead of globally. 321 322 * tcp memory pressure: sockets memory pressure for the tcp protocol. 323 324 2.7.3 Common use cases 325 326 Because the "kmem" counter is fed to the main user counter, kernel memory can 327 never be limited completely independently of user memory. Say "U" is the user 328 limit, and "K" the kernel limit. There are three possible ways limits can be 329 set: 330 331 U != 0, K = unlimited: 332 This is the standard memcg limitation mechanism already present before kmem 333 accounting. Kernel memory is completely ignored. 334 335 U != 0, K < U: 336 Kernel memory is a subset of the user memory. This setup is useful in 337 deployments where the total amount of memory per-cgroup is overcommited. 338 Overcommiting kernel memory limits is definitely not recommended, since the 339 box can still run out of non-reclaimable memory. 340 In this case, the admin could set up K so that the sum of all groups is 341 never greater than the total memory, and freely set U at the cost of his 342 QoS. 343 344 U != 0, K >= U: 345 Since kmem charges will also be fed to the user counter and reclaim will be 346 triggered for the cgroup for both kinds of memory. This setup gives the 347 admin a unified view of memory, and it is also useful for people who just 348 want to track kernel memory usage. 349 350 3. User Interface 351 352 0. Configuration 353 354 a. Enable CONFIG_CGROUPS 355 b. Enable CONFIG_RESOURCE_COUNTERS 356 c. Enable CONFIG_MEMCG 357 d. Enable CONFIG_MEMCG_SWAP (to use swap extension) 358 d. Enable CONFIG_MEMCG_KMEM (to use kmem extension) 359 360 1. Prepare the cgroups (see cgroups.txt, Why are cgroups needed?) 361 # mount -t tmpfs none /sys/fs/cgroup 362 # mkdir /sys/fs/cgroup/memory 363 # mount -t cgroup none /sys/fs/cgroup/memory -o memory 364 365 2. Make the new group and move bash into it 366 # mkdir /sys/fs/cgroup/memory/0 367 # echo $$ > /sys/fs/cgroup/memory/0/tasks 368 369 Since now we're in the 0 cgroup, we can alter the memory limit: 370 # echo 4M > /sys/fs/cgroup/memory/0/memory.limit_in_bytes 371 372 NOTE: We can use a suffix (k, K, m, M, g or G) to indicate values in kilo, 373 mega or gigabytes. (Here, Kilo, Mega, Giga are Kibibytes, Mebibytes, Gibibytes.) 374 375 NOTE: We can write "-1" to reset the *.limit_in_bytes(unlimited). 376 NOTE: We cannot set limits on the root cgroup any more. 377 378 # cat /sys/fs/cgroup/memory/0/memory.limit_in_bytes 379 4194304 380 381 We can check the usage: 382 # cat /sys/fs/cgroup/memory/0/memory.usage_in_bytes 383 1216512 384 385 A successful write to this file does not guarantee a successful setting of 386 this limit to the value written into the file. This can be due to a 387 number of factors, such as rounding up to page boundaries or the total 388 availability of memory on the system. The user is required to re-read 389 this file after a write to guarantee the value committed by the kernel. 390 391 # echo 1 > memory.limit_in_bytes 392 # cat memory.limit_in_bytes 393 4096 394 395 The memory.failcnt field gives the number of times that the cgroup limit was 396 exceeded. 397 398 The memory.stat file gives accounting information. Now, the number of 399 caches, RSS and Active pages/Inactive pages are shown. 400 401 4. Testing 402 403 For testing features and implementation, see memcg_test.txt. 404 405 Performance test is also important. To see pure memory controller's overhead, 406 testing on tmpfs will give you good numbers of small overheads. 407 Example: do kernel make on tmpfs. 408 409 Page-fault scalability is also important. At measuring parallel 410 page fault test, multi-process test may be better than multi-thread 411 test because it has noise of shared objects/status. 412 413 But the above two are testing extreme situations. 414 Trying usual test under memory controller is always helpful. 415 416 4.1 Troubleshooting 417 418 Sometimes a user might find that the application under a cgroup is 419 terminated by the OOM killer. There are several causes for this: 420 421 1. The cgroup limit is too low (just too low to do anything useful) 422 2. The user is using anonymous memory and swap is turned off or too low 423 424 A sync followed by echo 1 > /proc/sys/vm/drop_caches will help get rid of 425 some of the pages cached in the cgroup (page cache pages). 426 427 To know what happens, disabling OOM_Kill as per "10. OOM Control" (below) and 428 seeing what happens will be helpful. 429 430 4.2 Task migration 431 432 When a task migrates from one cgroup to another, its charge is not 433 carried forward by default. The pages allocated from the original cgroup still 434 remain charged to it, the charge is dropped when the page is freed or 435 reclaimed. 436 437 You can move charges of a task along with task migration. 438 See 8. "Move charges at task migration" 439 440 4.3 Removing a cgroup 441 442 A cgroup can be removed by rmdir, but as discussed in sections 4.1 and 4.2, a 443 cgroup might have some charge associated with it, even though all 444 tasks have migrated away from it. (because we charge against pages, not 445 against tasks.) 446 447 We move the stats to root (if use_hierarchy==0) or parent (if 448 use_hierarchy==1), and no change on the charge except uncharging 449 from the child. 450 451 Charges recorded in swap information is not updated at removal of cgroup. 452 Recorded information is discarded and a cgroup which uses swap (swapcache) 453 will be charged as a new owner of it. 454 455 About use_hierarchy, see Section 6. 456 457 5. Misc. interfaces. 458 459 5.1 force_empty 460 memory.force_empty interface is provided to make cgroup's memory usage empty. 461 When writing anything to this 462 463 # echo 0 > memory.force_empty 464 465 the cgroup will be reclaimed and as many pages reclaimed as possible. 466 467 The typical use case for this interface is before calling rmdir(). 468 Because rmdir() moves all pages to parent, some out-of-use page caches can be 469 moved to the parent. If you want to avoid that, force_empty will be useful. 470 471 Also, note that when memory.kmem.limit_in_bytes is set the charges due to 472 kernel pages will still be seen. This is not considered a failure and the 473 write will still return success. In this case, it is expected that 474 memory.kmem.usage_in_bytes == memory.usage_in_bytes. 475 476 About use_hierarchy, see Section 6. 477 478 5.2 stat file 479 480 memory.stat file includes following statistics 481 482 # per-memory cgroup local status 483 cache - # of bytes of page cache memory. 484 rss - # of bytes of anonymous and swap cache memory (includes 485 transparent hugepages). 486 rss_huge - # of bytes of anonymous transparent hugepages. 487 mapped_file - # of bytes of mapped file (includes tmpfs/shmem) 488 pgpgin - # of charging events to the memory cgroup. The charging 489 event happens each time a page is accounted as either mapped 490 anon page(RSS) or cache page(Page Cache) to the cgroup. 491 pgpgout - # of uncharging events to the memory cgroup. The uncharging 492 event happens each time a page is unaccounted from the cgroup. 493 swap - # of bytes of swap usage 494 writeback - # of bytes of file/anon cache that are queued for syncing to 495 disk. 496 inactive_anon - # of bytes of anonymous and swap cache memory on inactive 497 LRU list. 498 active_anon - # of bytes of anonymous and swap cache memory on active 499 LRU list. 500 inactive_file - # of bytes of file-backed memory on inactive LRU list. 501 active_file - # of bytes of file-backed memory on active LRU list. 502 unevictable - # of bytes of memory that cannot be reclaimed (mlocked etc). 503 504 # status considering hierarchy (see memory.use_hierarchy settings) 505 506 hierarchical_memory_limit - # of bytes of memory limit with regard to hierarchy 507 under which the memory cgroup is 508 hierarchical_memsw_limit - # of bytes of memory+swap limit with regard to 509 hierarchy under which memory cgroup is. 510 511 total_<counter> - # hierarchical version of <counter>, which in 512 addition to the cgroup's own value includes the 513 sum of all hierarchical children's values of 514 <counter>, i.e. total_cache 515 516 # The following additional stats are dependent on CONFIG_DEBUG_VM. 517 518 recent_rotated_anon - VM internal parameter. (see mm/vmscan.c) 519 recent_rotated_file - VM internal parameter. (see mm/vmscan.c) 520 recent_scanned_anon - VM internal parameter. (see mm/vmscan.c) 521 recent_scanned_file - VM internal parameter. (see mm/vmscan.c) 522 523 Memo: 524 recent_rotated means recent frequency of LRU rotation. 525 recent_scanned means recent # of scans to LRU. 526 showing for better debug please see the code for meanings. 527 528 Note: 529 Only anonymous and swap cache memory is listed as part of 'rss' stat. 530 This should not be confused with the true 'resident set size' or the 531 amount of physical memory used by the cgroup. 532 'rss + file_mapped" will give you resident set size of cgroup. 533 (Note: file and shmem may be shared among other cgroups. In that case, 534 file_mapped is accounted only when the memory cgroup is owner of page 535 cache.) 536 537 5.3 swappiness 538 539 Overrides /proc/sys/vm/swappiness for the particular group. The tunable 540 in the root cgroup corresponds to the global swappiness setting. 541 542 Please note that unlike during the global reclaim, limit reclaim 543 enforces that 0 swappiness really prevents from any swapping even if 544 there is a swap storage available. This might lead to memcg OOM killer 545 if there are no file pages to reclaim. 546 547 5.4 failcnt 548 549 A memory cgroup provides memory.failcnt and memory.memsw.failcnt files. 550 This failcnt(== failure count) shows the number of times that a usage counter 551 hit its limit. When a memory cgroup hits a limit, failcnt increases and 552 memory under it will be reclaimed. 553 554 You can reset failcnt by writing 0 to failcnt file. 555 # echo 0 > .../memory.failcnt 556 557 5.5 usage_in_bytes 558 559 For efficiency, as other kernel components, memory cgroup uses some optimization 560 to avoid unnecessary cacheline false sharing. usage_in_bytes is affected by the 561 method and doesn't show 'exact' value of memory (and swap) usage, it's a fuzz 562 value for efficient access. (Of course, when necessary, it's synchronized.) 563 If you want to know more exact memory usage, you should use RSS+CACHE(+SWAP) 564 value in memory.stat(see 5.2). 565 566 5.6 numa_stat 567 568 This is similar to numa_maps but operates on a per-memcg basis. This is 569 useful for providing visibility into the numa locality information within 570 an memcg since the pages are allowed to be allocated from any physical 571 node. One of the use cases is evaluating application performance by 572 combining this information with the application's CPU allocation. 573 574 Each memcg's numa_stat file includes "total", "file", "anon" and "unevictable" 575 per-node page counts including "hierarchical_<counter>" which sums up all 576 hierarchical children's values in addition to the memcg's own value. 577 578 The output format of memory.numa_stat is: 579 580 total=<total pages> N0=<node 0 pages> N1=<node 1 pages> ... 581 file=<total file pages> N0=<node 0 pages> N1=<node 1 pages> ... 582 anon=<total anon pages> N0=<node 0 pages> N1=<node 1 pages> ... 583 unevictable=<total anon pages> N0=<node 0 pages> N1=<node 1 pages> ... 584 hierarchical_<counter>=<counter pages> N0=<node 0 pages> N1=<node 1 pages> ... 585 586 The "total" count is sum of file + anon + unevictable. 587 588 6. Hierarchy support 589 590 The memory controller supports a deep hierarchy and hierarchical accounting. 591 The hierarchy is created by creating the appropriate cgroups in the 592 cgroup filesystem. Consider for example, the following cgroup filesystem 593 hierarchy 594 595 root 596 / | \ 597 / | \ 598 a b c 599 | \ 600 | \ 601 d e 602 603 In the diagram above, with hierarchical accounting enabled, all memory 604 usage of e, is accounted to its ancestors up until the root (i.e, c and root), 605 that has memory.use_hierarchy enabled. If one of the ancestors goes over its 606 limit, the reclaim algorithm reclaims from the tasks in the ancestor and the 607 children of the ancestor. 608 609 6.1 Enabling hierarchical accounting and reclaim 610 611 A memory cgroup by default disables the hierarchy feature. Support 612 can be enabled by writing 1 to memory.use_hierarchy file of the root cgroup 613 614 # echo 1 > memory.use_hierarchy 615 616 The feature can be disabled by 617 618 # echo 0 > memory.use_hierarchy 619 620 NOTE1: Enabling/disabling will fail if either the cgroup already has other 621 cgroups created below it, or if the parent cgroup has use_hierarchy 622 enabled. 623 624 NOTE2: When panic_on_oom is set to "2", the whole system will panic in 625 case of an OOM event in any cgroup. 626 627 7. Soft limits 628 629 Soft limits allow for greater sharing of memory. The idea behind soft limits 630 is to allow control groups to use as much of the memory as needed, provided 631 632 a. There is no memory contention 633 b. They do not exceed their hard limit 634 635 When the system detects memory contention or low memory, control groups 636 are pushed back to their soft limits. If the soft limit of each control 637 group is very high, they are pushed back as much as possible to make 638 sure that one control group does not starve the others of memory. 639 640 Please note that soft limits is a best-effort feature; it comes with 641 no guarantees, but it does its best to make sure that when memory is 642 heavily contended for, memory is allocated based on the soft limit 643 hints/setup. Currently soft limit based reclaim is set up such that 644 it gets invoked from balance_pgdat (kswapd). 645 646 7.1 Interface 647 648 Soft limits can be setup by using the following commands (in this example we 649 assume a soft limit of 256 MiB) 650 651 # echo 256M > memory.soft_limit_in_bytes 652 653 If we want to change this to 1G, we can at any time use 654 655 # echo 1G > memory.soft_limit_in_bytes 656 657 NOTE1: Soft limits take effect over a long period of time, since they involve 658 reclaiming memory for balancing between memory cgroups 659 NOTE2: It is recommended to set the soft limit always below the hard limit, 660 otherwise the hard limit will take precedence. 661 662 8. Move charges at task migration 663 664 Users can move charges associated with a task along with task migration, that 665 is, uncharge task's pages from the old cgroup and charge them to the new cgroup. 666 This feature is not supported in !CONFIG_MMU environments because of lack of 667 page tables. 668 669 8.1 Interface 670 671 This feature is disabled by default. It can be enabled (and disabled again) by 672 writing to memory.move_charge_at_immigrate of the destination cgroup. 673 674 If you want to enable it: 675 676 # echo (some positive value) > memory.move_charge_at_immigrate 677 678 Note: Each bits of move_charge_at_immigrate has its own meaning about what type 679 of charges should be moved. See 8.2 for details. 680 Note: Charges are moved only when you move mm->owner, in other words, 681 a leader of a thread group. 682 Note: If we cannot find enough space for the task in the destination cgroup, we 683 try to make space by reclaiming memory. Task migration may fail if we 684 cannot make enough space. 685 Note: It can take several seconds if you move charges much. 686 687 And if you want disable it again: 688 689 # echo 0 > memory.move_charge_at_immigrate 690 691 8.2 Type of charges which can be moved 692 693 Each bit in move_charge_at_immigrate has its own meaning about what type of 694 charges should be moved. But in any case, it must be noted that an account of 695 a page or a swap can be moved only when it is charged to the task's current 696 (old) memory cgroup. 697 698 bit | what type of charges would be moved ? 699 -----+------------------------------------------------------------------------ 700 0 | A charge of an anonymous page (or swap of it) used by the target task. 701 | You must enable Swap Extension (see 2.4) to enable move of swap charges. 702 -----+------------------------------------------------------------------------ 703 1 | A charge of file pages (normal file, tmpfs file (e.g. ipc shared memory) 704 | and swaps of tmpfs file) mmapped by the target task. Unlike the case of 705 | anonymous pages, file pages (and swaps) in the range mmapped by the task 706 | will be moved even if the task hasn't done page fault, i.e. they might 707 | not be the task's "RSS", but other task's "RSS" that maps the same file. 708 | And mapcount of the page is ignored (the page can be moved even if 709 | page_mapcount(page) > 1). You must enable Swap Extension (see 2.4) to 710 | enable move of swap charges. 711 712 8.3 TODO 713 714 - All of moving charge operations are done under cgroup_mutex. It's not good 715 behavior to hold the mutex too long, so we may need some trick. 716 717 9. Memory thresholds 718 719 Memory cgroup implements memory thresholds using the cgroups notification 720 API (see cgroups.txt). It allows to register multiple memory and memsw 721 thresholds and gets notifications when it crosses. 722 723 To register a threshold, an application must: 724 - create an eventfd using eventfd(2); 725 - open memory.usage_in_bytes or memory.memsw.usage_in_bytes; 726 - write string like "<event_fd> <fd of memory.usage_in_bytes> <threshold>" to 727 cgroup.event_control. 728 729 Application will be notified through eventfd when memory usage crosses 730 threshold in any direction. 731 732 It's applicable for root and non-root cgroup. 733 734 10. OOM Control 735 736 memory.oom_control file is for OOM notification and other controls. 737 738 Memory cgroup implements OOM notifier using the cgroup notification 739 API (See cgroups.txt). It allows to register multiple OOM notification 740 delivery and gets notification when OOM happens. 741 742 To register a notifier, an application must: 743 - create an eventfd using eventfd(2) 744 - open memory.oom_control file 745 - write string like "<event_fd> <fd of memory.oom_control>" to 746 cgroup.event_control 747 748 The application will be notified through eventfd when OOM happens. 749 OOM notification doesn't work for the root cgroup. 750 751 You can disable the OOM-killer by writing "1" to memory.oom_control file, as: 752 753 #echo 1 > memory.oom_control 754 755 If OOM-killer is disabled, tasks under cgroup will hang/sleep 756 in memory cgroup's OOM-waitqueue when they request accountable memory. 757 758 For running them, you have to relax the memory cgroup's OOM status by 759 * enlarge limit or reduce usage. 760 To reduce usage, 761 * kill some tasks. 762 * move some tasks to other group with account migration. 763 * remove some files (on tmpfs?) 764 765 Then, stopped tasks will work again. 766 767 At reading, current status of OOM is shown. 768 oom_kill_disable 0 or 1 (if 1, oom-killer is disabled) 769 under_oom 0 or 1 (if 1, the memory cgroup is under OOM, tasks may 770 be stopped.) 771 772 11. Memory Pressure 773 774 The pressure level notifications can be used to monitor the memory 775 allocation cost; based on the pressure, applications can implement 776 different strategies of managing their memory resources. The pressure 777 levels are defined as following: 778 779 The "low" level means that the system is reclaiming memory for new 780 allocations. Monitoring this reclaiming activity might be useful for 781 maintaining cache level. Upon notification, the program (typically 782 "Activity Manager") might analyze vmstat and act in advance (i.e. 783 prematurely shutdown unimportant services). 784 785 The "medium" level means that the system is experiencing medium memory 786 pressure, the system might be making swap, paging out active file caches, 787 etc. Upon this event applications may decide to further analyze 788 vmstat/zoneinfo/memcg or internal memory usage statistics and free any 789 resources that can be easily reconstructed or re-read from a disk. 790 791 The "critical" level means that the system is actively thrashing, it is 792 about to out of memory (OOM) or even the in-kernel OOM killer is on its 793 way to trigger. Applications should do whatever they can to help the 794 system. It might be too late to consult with vmstat or any other 795 statistics, so it's advisable to take an immediate action. 796 797 The events are propagated upward until the event is handled, i.e. the 798 events are not pass-through. Here is what this means: for example you have 799 three cgroups: A->B->C. Now you set up an event listener on cgroups A, B 800 and C, and suppose group C experiences some pressure. In this situation, 801 only group C will receive the notification, i.e. groups A and B will not 802 receive it. This is done to avoid excessive "broadcasting" of messages, 803 which disturbs the system and which is especially bad if we are low on 804 memory or thrashing. So, organize the cgroups wisely, or propagate the 805 events manually (or, ask us to implement the pass-through events, 806 explaining why would you need them.) 807 808 The file memory.pressure_level is only used to setup an eventfd. To 809 register a notification, an application must: 810 811 - create an eventfd using eventfd(2); 812 - open memory.pressure_level; 813 - write string like "<event_fd> <fd of memory.pressure_level> <level>" 814 to cgroup.event_control. 815 816 Application will be notified through eventfd when memory pressure is at 817 the specific level (or higher). Read/write operations to 818 memory.pressure_level are no implemented. 819 820 Test: 821 822 Here is a small script example that makes a new cgroup, sets up a 823 memory limit, sets up a notification in the cgroup and then makes child 824 cgroup experience a critical pressure: 825 826 # cd /sys/fs/cgroup/memory/ 827 # mkdir foo 828 # cd foo 829 # cgroup_event_listener memory.pressure_level low & 830 # echo 8000000 > memory.limit_in_bytes 831 # echo 8000000 > memory.memsw.limit_in_bytes 832 # echo $$ > tasks 833 # dd if=/dev/zero | read x 834 835 (Expect a bunch of notifications, and eventually, the oom-killer will 836 trigger.) 837 838 12. TODO 839 840 1. Make per-cgroup scanner reclaim not-shared pages first 841 2. Teach controller to account for shared-pages 842 3. Start reclamation in the background when the limit is 843 not yet hit but the usage is getting closer 844 845 Summary 846 847 Overall, the memory controller has been a stable controller and has been 848 commented and discussed quite extensively in the community. 849 850 References 851 852 1. Singh, Balbir. RFC: Memory Controller, http://lwn.net/Articles/206697/ 853 2. Singh, Balbir. Memory Controller (RSS Control), 854 http://lwn.net/Articles/222762/ 855 3. Emelianov, Pavel. Resource controllers based on process cgroups 856 http://lkml.org/lkml/2007/3/6/198 857 4. Emelianov, Pavel. RSS controller based on process cgroups (v2) 858 http://lkml.org/lkml/2007/4/9/78 859 5. Emelianov, Pavel. RSS controller based on process cgroups (v3) 860 http://lkml.org/lkml/2007/5/30/244 861 6. Menage, Paul. Control Groups v10, http://lwn.net/Articles/236032/ 862 7. Vaidyanathan, Srinivasan, Control Groups: Pagecache accounting and control 863 subsystem (v3), http://lwn.net/Articles/235534/ 864 8. Singh, Balbir. RSS controller v2 test results (lmbench), 865 http://lkml.org/lkml/2007/5/17/232 866 9. Singh, Balbir. RSS controller v2 AIM9 results 867 http://lkml.org/lkml/2007/5/18/1 868 10. Singh, Balbir. Memory controller v6 test results, 869 http://lkml.org/lkml/2007/8/19/36 870 11. Singh, Balbir. Memory controller introduction (v6), 871 http://lkml.org/lkml/2007/8/17/69 872 12. Corbet, Jonathan, Controlling memory use in cgroups, 873 http://lwn.net/Articles/243795/