Lines Matching +full:memory +full:- +full:controller
2 Memory Resource Controller
12 The Memory Resource Controller has generically been referred to as the
13 memory controller in this document. Do not confuse memory controller
14 used here with the memory controller that is used in hardware.
17 When we mention a cgroup (cgroupfs's directory) with memory controller,
18 we call it "memory cgroup". When you see git-log and source code, you'll
22 Benefits and Purpose of the memory controller
25 The memory controller isolates the memory behaviour of a group of tasks
27 uses of the memory controller. The memory controller can be used to
30 Memory-hungry applications can be isolated and limited to a smaller
31 amount of memory.
32 b. Create a cgroup with a limited amount of memory; this can be used
34 c. Virtualization solutions can control the amount of memory they want
36 d. A CD/DVD burner could control the amount of memory used by the
38 of available memory.
39 e. There are several other use cases; find one or use the controller just
42 Current Status: linux-2.6.34-mmotm(development version of 2010/April)
46 - accounting anonymous pages, file caches, swap caches usage and limiting them.
47 - pages are linked to per-memcg LRU exclusively, and there is no global LRU.
48 - optionally, memory+swap usage can be accounted and limited.
49 - hierarchical accounting
50 - soft limit
51 - moving (recharging) account at moving a task is selectable.
52 - usage threshold notifier
53 - memory pressure notifier
54 - oom-killer disable knob and oom-notifier
55 - Root cgroup has no limit controls.
57 Kernel memory support is a work in progress, and the current version provides
68 memory.usage_in_bytes show current usage for memory
70 memory.memsw.usage_in_bytes show current usage for memory+Swap
72 memory.limit_in_bytes set/show limit of memory usage
73 memory.memsw.limit_in_bytes set/show limit of memory+Swap usage
74 memory.failcnt show the number of memory usage hits limits
75 memory.memsw.failcnt show the number of memory+Swap hits limits
76 memory.max_usage_in_bytes show max memory usage recorded
77 memory.memsw.max_usage_in_bytes show max memory+Swap usage recorded
78 memory.soft_limit_in_bytes set/show soft limit of memory usage
80 memory.stat show various statistics
81 memory.use_hierarchy set/show hierarchical account enabled
84 memory.force_empty trigger forced page reclaim
85 memory.pressure_level set memory pressure notifications
86 memory.swappiness set/show swappiness parameter of vmscan
88 memory.move_charge_at_immigrate set/show controls of moving charges
89 memory.oom_control set/show oom controls.
90 memory.numa_stat show the number of memory usage per numa
92 memory.kmem.limit_in_bytes This knob is deprecated and writing to
93 it will return -ENOTSUPP.
94 memory.kmem.usage_in_bytes show current kernel memory allocation
95 memory.kmem.failcnt show the number of kernel memory usage
97 memory.kmem.max_usage_in_bytes show max kernel memory usage recorded
99 memory.kmem.tcp.limit_in_bytes set/show hard limit for tcp buf memory
100 memory.kmem.tcp.usage_in_bytes show current tcp buf memory allocation
101 memory.kmem.tcp.failcnt show the number of tcp buf memory usage
103 memory.kmem.tcp.max_usage_in_bytes show max tcp buf memory usage recorded
109 The memory controller has a long history. A request for comments for the memory
110 controller was posted by Balbir Singh [1]. At the time the RFC was posted
111 there were several implementations for memory control. The goal of the
113 for memory control. The first RSS controller was posted by Balbir Singh[2]
115 RSS controller. At OLS, at the resource management BoF, everyone suggested
117 to allow user space handling of OOM. The current memory controller is
121 2. Memory Control
124 Memory is a unique resource in the sense that it is present in a limited
127 memory, the same physical memory needs to be reused to accomplish the task.
129 The memory controller implementation has been divided into phases. These
132 1. Memory controller
133 2. mlock(2) controller
134 3. Kernel user memory accounting and slab control
135 4. user mappings length controller
137 The memory controller is the first controller developed.
140 -----------
143 page_counter tracks the current memory usage and limit of the group of
144 processes associated with the controller. Each cgroup has a memory controller
148 ---------------
152 +--------------------+
155 +--------------------+
158 +---------------+ | +---------------+
161 +---------------+ | +---------------+
163 + --------------+
165 +---------------+ +------+--------+
166 | page +----------> page_cgroup|
168 +---------------+ +---------------+
173 Figure 1 shows the important aspects of the controller
184 If everything goes well, a page meta-data-structure called page_cgroup is
186 (*) page_cgroup structure is allocated at boot/memory-hotplug time.
189 ------------------------
197 inserted into inode (radix-tree). While it's mapped into the page tables of
201 unaccounted when it's removed from radix-tree. Even if RSS pages are fully
204 A swapped-in page is accounted after adding into swapcache.
206 Note: The kernel does swapin-readahead and reads multiple swaps at once.
212 Note: we just account pages-on-LRU because our purpose is to control amount
213 of used pages; not-on-LRU pages tend to be out-of-control from VM view.
216 --------------------------
222 the cgroup that brought it in -- this will happen on memory pressure).
228 --------------------------------------
235 - memory.memsw.usage_in_bytes.
236 - memory.memsw.limit_in_bytes.
238 memsw means memory+swap. Usage of memory+swap is limited by
241 Example: Assume a system with 4G of swap. A task which allocates 6G of memory
242 (by mistake) under 2G memory limitation will use all swap.
247 **why 'memory+swap' rather than swap**
249 The global LRU(kswapd) can swap out arbitrary pages. Swap-out means
250 to move account from memory to swap...there is no change in usage of
251 memory+swap. In other words, when we want to limit the usage of swap without
252 affecting global LRU, memory+swap limit is better than just limiting swap from
255 **What happens when a cgroup hits memory.memsw.limit_in_bytes**
257 When a cgroup hits memory.memsw.limit_in_bytes, it's useless to do swap-out
258 in this cgroup. Then, swap-out will not be done by cgroup routine and file
259 caches are dropped. But as mentioned above, global LRU can do swapout memory
260 from it for sanity of the system's memory management state. You can't forbid
264 -----------
268 to reclaim memory from the cgroup so as to make space for the new
274 pages that are selected for reclaiming come from the per-cgroup LRU
288 -----------
292 Page lock (PG_locked bit of page->flags)
293 mm->page_table_lock or split pte_lock
294 lock_page_memcg (memcg->move_lock)
295 mapping->i_pages lock
296 lruvec->lru_lock.
298 Per-node-per-memcgroup LRU (cgroup's private LRU) is guarded by
299 lruvec->lru_lock; PG_lru bit of page->flags is cleared before
300 isolating a page from its LRU under lruvec->lru_lock.
302 2.7 Kernel Memory Extension
303 -----------------------------------------------
305 With the Kernel memory extension, the Memory Controller is able to limit
306 the amount of kernel memory used by the system. Kernel memory is fundamentally
307 different than user memory, since it can't be swapped out, which makes it
310 Kernel memory accounting is enabled for all memory cgroups by default. But
311 it can be disabled system-wide by passing cgroup.memory=nokmem to the kernel
312 at boot time. In this case, kernel memory will not be accounted at all.
314 Kernel memory limits are not imposed for the root cgroup. Usage for the root
315 cgroup may or may not be accounted. The memory used is accumulated into
316 memory.kmem.usage_in_bytes, or in a separate counter when it makes sense.
322 Currently no soft limit is implemented for kernel memory. It is future work
325 2.7.1 Current Kernel Memory resources accounted
326 -----------------------------------------------
330 kernel memory, we prevent new processes from being created when the kernel
331 memory usage is too high.
341 sockets memory pressure:
342 some sockets protocols have memory pressure
343 thresholds. The Memory Controller allows them to be controlled individually
346 tcp memory pressure:
347 sockets memory pressure for the tcp protocol.
350 ----------------------
352 Because the "kmem" counter is fed to the main user counter, kernel memory can
353 never be limited completely independently of user memory. Say "U" is the user
359 accounting. Kernel memory is completely ignored.
362 Kernel memory is a subset of the user memory. This setup is useful in
363 deployments where the total amount of memory per-cgroup is overcommitted.
364 Overcommitting kernel memory limits is definitely not recommended, since the
365 box can still run out of non-reclaimable memory.
367 never greater than the total memory, and freely set U at the cost of his
371 In the current implementation, memory reclaim will NOT be
377 triggered for the cgroup for both kinds of memory. This setup gives the
378 admin a unified view of memory, and it is also useful for people who just
379 want to track kernel memory usage.
385 ------------------
391 -------------------------------------------------------------------
395 # mount -t tmpfs none /sys/fs/cgroup
396 # mkdir /sys/fs/cgroup/memory
397 # mount -t cgroup none /sys/fs/cgroup/memory -o memory
401 # mkdir /sys/fs/cgroup/memory/0
402 # echo $$ > /sys/fs/cgroup/memory/0/tasks
404 Since now we're in the 0 cgroup, we can alter the memory limit::
406 # echo 4M > /sys/fs/cgroup/memory/0/memory.limit_in_bytes
414 We can write "-1" to reset the ``*.limit_in_bytes(unlimited)``.
421 # cat /sys/fs/cgroup/memory/0/memory.limit_in_bytes
426 # cat /sys/fs/cgroup/memory/0/memory.usage_in_bytes
432 availability of memory on the system. The user is required to re-read
435 # echo 1 > memory.limit_in_bytes
436 # cat memory.limit_in_bytes
439 The memory.failcnt field gives the number of times that the cgroup limit was
442 The memory.stat file gives accounting information. Now, the number of
450 Performance test is also important. To see pure memory controller's overhead,
454 Page-fault scalability is also important. At measuring parallel
455 page fault test, multi-process test may be better than multi-thread
459 Trying usual test under memory controller is always helpful.
462 -------------------
468 2. The user is using anonymous memory and swap is turned off or too low
477 ------------------
488 ---------------------
506 ---------------
507 memory.force_empty interface is provided to make cgroup's memory usage empty.
510 # echo 0 > memory.force_empty
516 charged file caches. Some out-of-use page caches may keep charged until
517 memory pressure happens. If you want to avoid that, force_empty will be useful.
520 -------------
522 memory.stat file includes following statistics
524 per-memory cgroup local status
528 cache # of bytes of page cache memory.
529 rss # of bytes of anonymous and swap cache memory (includes
533 pgpgin # of charging events to the memory cgroup. The charging
536 pgpgout # of uncharging events to the memory cgroup. The uncharging
542 inactive_anon # of bytes of anonymous and swap cache memory on inactive
544 active_anon # of bytes of anonymous and swap cache memory on active
546 inactive_file # of bytes of file-backed memory on inactive LRU list.
547 active_file # of bytes of file-backed memory on active LRU list.
548 unevictable # of bytes of memory that cannot be reclaimed (mlocked etc).
551 status considering hierarchy (see memory.use_hierarchy settings)
555 hierarchical_memory_limit # of bytes of memory limit with regard to hierarchy
556 under which the memory cgroup is
557 hierarchical_memsw_limit # of bytes of memory+swap limit with regard to
558 hierarchy under which memory cgroup is.
582 Only anonymous and swap cache memory is listed as part of 'rss' stat.
584 amount of physical memory used by the cgroup.
589 mapped_file is accounted only when the memory cgroup is owner of page
593 --------------
604 -----------
606 A memory cgroup provides memory.failcnt and memory.memsw.failcnt files.
608 hit its limit. When a memory cgroup hits a limit, failcnt increases and
609 memory under it will be reclaimed.
613 # echo 0 > .../memory.failcnt
616 ------------------
618 For efficiency, as other kernel components, memory cgroup uses some optimization
620 method and doesn't show 'exact' value of memory (and swap) usage, it's a fuzz
622 If you want to know more exact memory usage, you should use RSS+CACHE(+SWAP)
623 value in memory.stat(see 5.2).
626 -------------
628 This is similar to numa_maps but operates on a per-memcg basis. This is
635 per-node page counts including "hierarchical_<counter>" which sums up all
638 The output format of memory.numa_stat is::
651 The memory controller supports a deep hierarchy and hierarchical accounting.
664 In the diagram above, with hierarchical accounting enabled, all memory
670 ---------------------------------------
676 For compatibility reasons writing 1 to memory.use_hierarchy will always pass::
678 # echo 1 > memory.use_hierarchy
683 Soft limits allow for greater sharing of memory. The idea behind soft limits
684 is to allow control groups to use as much of the memory as needed, provided
686 a. There is no memory contention
689 When the system detects memory contention or low memory, control groups
692 sure that one control group does not starve the others of memory.
694 Please note that soft limits is a best-effort feature; it comes with
695 no guarantees, but it does its best to make sure that when memory is
696 heavily contended for, memory is allocated based on the soft limit
701 -------------
706 # echo 256M > memory.soft_limit_in_bytes
710 # echo 1G > memory.soft_limit_in_bytes
714 reclaiming memory for balancing between memory cgroups
728 -------------
731 writing to memory.move_charge_at_immigrate of the destination cgroup.
735 # echo (some positive value) > memory.move_charge_at_immigrate
741 Charges are moved only when you move mm->owner, in other words,
745 try to make space by reclaiming memory. Task migration may fail if we
752 # echo 0 > memory.move_charge_at_immigrate
755 --------------------------------------
760 (old) memory cgroup.
762 +---+--------------------------------------------------------------------------+
767 +---+--------------------------------------------------------------------------+
768 | 1 | A charge of file pages (normal file, tmpfs file (e.g. ipc shared memory) |
776 +---+--------------------------------------------------------------------------+
779 --------
781 - All of moving charge operations are done under cgroup_mutex. It's not good
784 9. Memory thresholds
787 Memory cgroup implements memory thresholds using the cgroups notification
788 API (see cgroups.txt). It allows to register multiple memory and memsw
793 - create an eventfd using eventfd(2);
794 - open memory.usage_in_bytes or memory.memsw.usage_in_bytes;
795 - write string like "<event_fd> <fd of memory.usage_in_bytes> <threshold>" to
798 Application will be notified through eventfd when memory usage crosses
801 It's applicable for root and non-root cgroup.
806 memory.oom_control file is for OOM notification and other controls.
808 Memory cgroup implements OOM notifier using the cgroup notification
814 - create an eventfd using eventfd(2)
815 - open memory.oom_control file
816 - write string like "<event_fd> <fd of memory.oom_control>" to
822 You can disable the OOM-killer by writing "1" to memory.oom_control file, as:
824 #echo 1 > memory.oom_control
826 If OOM-killer is disabled, tasks under cgroup will hang/sleep
827 in memory cgroup's OOM-waitqueue when they request accountable memory.
829 For running them, you have to relax the memory cgroup's OOM status by
843 - oom_kill_disable 0 or 1
844 (if 1, oom-killer is disabled)
845 - under_oom 0 or 1
846 (if 1, the memory cgroup is under OOM, tasks may be stopped.)
847 - oom_kill integer counter
851 11. Memory Pressure
854 The pressure level notifications can be used to monitor the memory
856 different strategies of managing their memory resources. The pressure
859 The "low" level means that the system is reclaiming memory for new
865 The "medium" level means that the system is experiencing medium memory
868 vmstat/zoneinfo/memcg or internal memory usage statistics and free any
869 resources that can be easily reconstructed or re-read from a disk.
872 about to out of memory (OOM) or even the in-kernel OOM killer is on its
878 events are not pass-through. For example, you have three cgroups: A->B->C. Now
883 especially bad if we are low on memory or thrashing. Group B, will receive
888 - "default": this is the default behavior specified above. This mode is the
892 - "hierarchy": events always propagate up to the root, similar to the default
895 example, groups A, B, and C will receive notification of memory pressure.
897 - "local": events are pass-through, i.e. they only receive notifications when
898 memory pressure is experienced in the memcg for which the notification is
900 registered for "local" notification and the group experiences memory
906 specified by a comma-delimited string, i.e. "low,hierarchy" specifies
907 hierarchical, pass-through, notification for all ancestor memcgs. Notification
908 that is the default, non pass-through behavior, does not specify a mode.
909 "medium,local" specifies pass-through notification for the medium level.
911 The file memory.pressure_level is only used to setup an eventfd. To
914 - create an eventfd using eventfd(2);
915 - open memory.pressure_level;
916 - write string as "<event_fd> <fd of memory.pressure_level> <level[,mode]>"
919 Application will be notified through eventfd when memory pressure is at
921 memory.pressure_level are no implemented.
926 memory limit, sets up a notification in the cgroup and then makes child
929 # cd /sys/fs/cgroup/memory/
932 # cgroup_event_listener memory.pressure_level low,hierarchy &
933 # echo 8000000 > memory.limit_in_bytes
934 # echo 8000000 > memory.memsw.limit_in_bytes
938 (Expect a bunch of notifications, and eventually, the oom-killer will
944 1. Make per-cgroup scanner reclaim not-shared pages first
945 2. Teach controller to account for shared-pages
952 Overall, the memory controller has been a stable controller and has been
958 1. Singh, Balbir. RFC: Memory Controller, http://lwn.net/Articles/206697/
959 2. Singh, Balbir. Memory Controller (RSS Control),
963 4. Emelianov, Pavel. RSS controller based on process cgroups (v2)
965 5. Emelianov, Pavel. RSS controller based on process cgroups (v3)
970 8. Singh, Balbir. RSS controller v2 test results (lmbench),
972 9. Singh, Balbir. RSS controller v2 AIM9 results
974 10. Singh, Balbir. Memory controller v6 test results,
975 https://lore.kernel.org/r/20070819094658.654.84837.sendpatchset@balbir-laptop
976 11. Singh, Balbir. Memory controller introduction (v6),
977 https://lore.kernel.org/r/20070817084228.26003.12568.sendpatchset@balbir-laptop
978 12. Corbet, Jonathan, Controlling memory use in cgroups,