| blob | dd49864979964f10ee097fd46779294e4b2cbe21 |
2 What is Linux Memory Policy?
4 In the Linux kernel, "memory policy" determines from which node the kernel will
5 allocate memory in a NUMA system or in an emulated NUMA system. Linux has
6 supported platforms with Non-Uniform Memory Access architectures since 2.4.?.
7 The current memory policy support was added to Linux 2.6 around May 2004. This
8 document attempts to describe the concepts and APIs of the 2.6 memory policy
9 support.
11 Memory policies should not be confused with cpusets (Documentation/cpusets.txt)
12 which is an administrative mechanism for restricting the nodes from which
13 memory may be allocated by a set of processes. Memory policies are a
14 programming interface that a NUMA-aware application can take advantage of. When
15 both cpusets and policies are applied to a task, the restrictions of the cpuset
16 takes priority. See "MEMORY POLICIES AND CPUSETS" below for more details.
18 MEMORY POLICY CONCEPTS
20 Scope of Memory Policies
22 The Linux kernel supports _scopes_ of memory policy, described here from
23 most general to most specific:
25 System Default Policy: this policy is "hard coded" into the kernel. It
26 is the policy that governs all page allocations that aren't controlled
27 by one of the more specific policy scopes discussed below. When the
28 system is "up and running", the system default policy will use "local
29 allocation" described below. However, during boot up, the system
30 default policy will be set to interleave allocations across all nodes
31 with "sufficient" memory, so as not to overload the initial boot node
32 with boot-time allocations.
34 Task/Process Policy: this is an optional, per-task policy. When defined
35 for a specific task, this policy controls all page allocations made by or
36 on behalf of the task that aren't controlled by a more specific scope.
37 If a task does not define a task policy, then all page allocations that
38 would have been controlled by the task policy "fall back" to the System
39 Default Policy.
41 The task policy applies to the entire address space of a task. Thus,
42 it is inheritable, and indeed is inherited, across both fork()
43 [clone() w/o the CLONE_VM flag] and exec*(). This allows a parent task
44 to establish the task policy for a child task exec()'d from an
45 executable image that has no awareness of memory policy. See the
46 MEMORY POLICY APIS section, below, for an overview of the system call
47 that a task may use to set/change it's task/process policy.
49 In a multi-threaded task, task policies apply only to the thread
50 [Linux kernel task] that installs the policy and any threads
51 subsequently created by that thread. Any sibling threads existing
52 at the time a new task policy is installed retain their current
53 policy.
55 A task policy applies only to pages allocated after the policy is
56 installed. Any pages already faulted in by the task when the task
57 changes its task policy remain where they were allocated based on
58 the policy at the time they were allocated.
60 VMA Policy: A "VMA" or "Virtual Memory Area" refers to a range of a task's
61 virtual adddress space. A task may define a specific policy for a range
62 of its virtual address space. See the MEMORY POLICIES APIS section,
63 below, for an overview of the mbind() system call used to set a VMA
64 policy.
66 A VMA policy will govern the allocation of pages that back this region of
67 the address space. Any regions of the task's address space that don't
68 have an explicit VMA policy will fall back to the task policy, which may
69 itself fall back to the System Default Policy.
71 VMA policies have a few complicating details:
73 VMA policy applies ONLY to anonymous pages. These include pages
74 allocated for anonymous segments, such as the task stack and heap, and
75 any regions of the address space mmap()ed with the MAP_ANONYMOUS flag.
76 If a VMA policy is applied to a file mapping, it will be ignored if
77 the mapping used the MAP_SHARED flag. If the file mapping used the
78 MAP_PRIVATE flag, the VMA policy will only be applied when an
79 anonymous page is allocated on an attempt to write to the mapping--
80 i.e., at Copy-On-Write.
82 VMA policies are shared between all tasks that share a virtual address
83 space--a.k.a. threads--independent of when the policy is installed; and
84 they are inherited across fork(). However, because VMA policies refer
85 to a specific region of a task's address space, and because the address
86 space is discarded and recreated on exec*(), VMA policies are NOT
87 inheritable across exec(). Thus, only NUMA-aware applications may
88 use VMA policies.
90 A task may install a new VMA policy on a sub-range of a previously
91 mmap()ed region. When this happens, Linux splits the existing virtual
92 memory area into 2 or 3 VMAs, each with it's own policy.
94 By default, VMA policy applies only to pages allocated after the policy
95 is installed. Any pages already faulted into the VMA range remain
96 where they were allocated based on the policy at the time they were
97 allocated. However, since 2.6.16, Linux supports page migration via
98 the mbind() system call, so that page contents can be moved to match
99 a newly installed policy.
101 Shared Policy: Conceptually, shared policies apply to "memory objects"
102 mapped shared into one or more tasks' distinct address spaces. An
103 application installs a shared policies the same way as VMA policies--using
104 the mbind() system call specifying a range of virtual addresses that map
105 the shared object. However, unlike VMA policies, which can be considered
106 to be an attribute of a range of a task's address space, shared policies
107 apply directly to the shared object. Thus, all tasks that attach to the
108 object share the policy, and all pages allocated for the shared object,
109 by any task, will obey the shared policy.
111 As of 2.6.22, only shared memory segments, created by shmget() or
112 mmap(MAP_ANONYMOUS|MAP_SHARED), support shared policy. When shared
113 policy support was added to Linux, the associated data structures were
114 added to hugetlbfs shmem segments. At the time, hugetlbfs did not
115 support allocation at fault time--a.k.a lazy allocation--so hugetlbfs
116 shmem segments were never "hooked up" to the shared policy support.
117 Although hugetlbfs segments now support lazy allocation, their support
118 for shared policy has not been completed.
120 As mentioned above [re: VMA policies], allocations of page cache
121 pages for regular files mmap()ed with MAP_SHARED ignore any VMA
122 policy installed on the virtual address range backed by the shared
123 file mapping. Rather, shared page cache pages, including pages backing
124 private mappings that have not yet been written by the task, follow
125 task policy, if any, else System Default Policy.
127 The shared policy infrastructure supports different policies on subset
128 ranges of the shared object. However, Linux still splits the VMA of
129 the task that installs the policy for each range of distinct policy.
130 Thus, different tasks that attach to a shared memory segment can have
131 different VMA configurations mapping that one shared object. This
132 can be seen by examining the /proc/<pid>/numa_maps of tasks sharing
133 a shared memory region, when one task has installed shared policy on
134 one or more ranges of the region.
136 Components of Memory Policies
138 A Linux memory policy is a tuple consisting of a "mode" and an optional set
139 of nodes. The mode determine the behavior of the policy, while the
140 optional set of nodes can be viewed as the arguments to the behavior.
142 Internally, memory policies are implemented by a reference counted
143 structure, struct mempolicy. Details of this structure will be discussed
144 in context, below, as required to explain the behavior.
146 Note: in some functions AND in the struct mempolicy itself, the mode
147 is called "policy". However, to avoid confusion with the policy tuple,
148 this document will continue to use the term "mode".
150 Linux memory policy supports the following 4 behavioral modes:
152 Default Mode--MPOL_DEFAULT: The behavior specified by this mode is
153 context or scope dependent.
155 As mentioned in the Policy Scope section above, during normal
156 system operation, the System Default Policy is hard coded to
157 contain the Default mode.
159 In this context, default mode means "local" allocation--that is
160 attempt to allocate the page from the node associated with the cpu
161 where the fault occurs. If the "local" node has no memory, or the
162 node's memory can be exhausted [no free pages available], local
163 allocation will "fallback to"--attempt to allocate pages from--
164 "nearby" nodes, in order of increasing "distance".
166 Implementation detail -- subject to change: "Fallback" uses
167 a per node list of sibling nodes--called zonelists--built at
168 boot time, or when nodes or memory are added or removed from
169 the system [memory hotplug]. These per node zonelist are
170 constructed with nodes in order of increasing distance based
171 on information provided by the platform firmware.
173 When a task/process policy or a shared policy contains the Default
174 mode, this also means "local allocation", as described above.
176 In the context of a VMA, Default mode means "fall back to task
177 policy"--which may or may not specify Default mode. Thus, Default
178 mode can not be counted on to mean local allocation when used
179 on a non-shared region of the address space. However, see
180 MPOL_PREFERRED below.
182 The Default mode does not use the optional set of nodes.
184 MPOL_BIND: This mode specifies that memory must come from the
185 set of nodes specified by the policy.
187 The memory policy APIs do not specify an order in which the nodes
188 will be searched. However, unlike "local allocation", the Bind
189 policy does not consider the distance between the nodes. Rather,
190 allocations will fallback to the nodes specified by the policy in
191 order of numeric node id. Like everything in Linux, this is subject
192 to change.
194 MPOL_PREFERRED: This mode specifies that the allocation should be
195 attempted from the single node specified in the policy. If that
196 allocation fails, the kernel will search other nodes, exactly as
197 it would for a local allocation that started at the preferred node
198 in increasing distance from the preferred node. "Local" allocation
199 policy can be viewed as a Preferred policy that starts at the node
200 containing the cpu where the allocation takes place.
202 Internally, the Preferred policy uses a single node--the
203 preferred_node member of struct mempolicy. A "distinguished
204 value of this preferred_node, currently '-1', is interpreted
205 as "the node containing the cpu where the allocation takes
206 place"--local allocation. This is the way to specify
207 local allocation for a specific range of addresses--i.e. for
208 VMA policies.
210 MPOL_INTERLEAVED: This mode specifies that page allocations be
211 interleaved, on a page granularity, across the nodes specified in
212 the policy. This mode also behaves slightly differently, based on
213 the context where it is used:
215 For allocation of anonymous pages and shared memory pages,
216 Interleave mode indexes the set of nodes specified by the policy
217 using the page offset of the faulting address into the segment
218 [VMA] containing the address modulo the number of nodes specified
219 by the policy. It then attempts to allocate a page, starting at
220 the selected node, as if the node had been specified by a Preferred
221 policy or had been selected by a local allocation. That is,
222 allocation will follow the per node zonelist.
224 For allocation of page cache pages, Interleave mode indexes the set
225 of nodes specified by the policy using a node counter maintained
226 per task. This counter wraps around to the lowest specified node
227 after it reaches the highest specified node. This will tend to
228 spread the pages out over the nodes specified by the policy based
229 on the order in which they are allocated, rather than based on any
230 page offset into an address range or file. During system boot up,
231 the temporary interleaved system default policy works in this
232 mode.
234 MEMORY POLICY APIs
236 Linux supports 3 system calls for controlling memory policy. These APIS
237 always affect only the calling task, the calling task's address space, or
238 some shared object mapped into the calling task's address space.
240 Note: the headers that define these APIs and the parameter data types
241 for user space applications reside in a package that is not part of
242 the Linux kernel. The kernel system call interfaces, with the 'sys_'
243 prefix, are defined in <linux/syscalls.h>; the mode and flag
244 definitions are defined in <linux/mempolicy.h>.
246 Set [Task] Memory Policy:
248 long set_mempolicy(int mode, const unsigned long *nmask,
249 unsigned long maxnode);
251 Set's the calling task's "task/process memory policy" to mode
252 specified by the 'mode' argument and the set of nodes defined
253 by 'nmask'. 'nmask' points to a bit mask of node ids containing
254 at least 'maxnode' ids.
256 See the set_mempolicy(2) man page for more details
259 Get [Task] Memory Policy or Related Information
261 long get_mempolicy(int *mode,
262 const unsigned long *nmask, unsigned long maxnode,
263 void *addr, int flags);
265 Queries the "task/process memory policy" of the calling task, or
266 the policy or location of a specified virtual address, depending
267 on the 'flags' argument.
269 See the get_mempolicy(2) man page for more details
272 Install VMA/Shared Policy for a Range of Task's Address Space
274 long mbind(void *start, unsigned long len, int mode,
275 const unsigned long *nmask, unsigned long maxnode,
276 unsigned flags);
278 mbind() installs the policy specified by (mode, nmask, maxnodes) as
279 a VMA policy for the range of the calling task's address space
280 specified by the 'start' and 'len' arguments. Additional actions
281 may be requested via the 'flags' argument.
283 See the mbind(2) man page for more details.
285 MEMORY POLICY COMMAND LINE INTERFACE
287 Although not strictly part of the Linux implementation of memory policy,
288 a command line tool, numactl(8), exists that allows one to:
290 + set the task policy for a specified program via set_mempolicy(2), fork(2) and
291 exec(2)
293 + set the shared policy for a shared memory segment via mbind(2)
295 The numactl(8) tool is packages with the run-time version of the library
296 containing the memory policy system call wrappers. Some distributions
297 package the headers and compile-time libraries in a separate development
298 package.
301 MEMORY POLICIES AND CPUSETS
303 Memory policies work within cpusets as described above. For memory policies
304 that require a node or set of nodes, the nodes are restricted to the set of
305 nodes whose memories are allowed by the cpuset constraints. If the nodemask
306 specified for the policy contains nodes that are not allowed by the cpuset, or
307 the intersection of the set of nodes specified for the policy and the set of
308 nodes with memory is the empty set, the policy is considered invalid
309 and cannot be installed.
311 The interaction of memory policies and cpusets can be problematic for a
312 couple of reasons:
314 1) the memory policy APIs take physical node id's as arguments. As mentioned
315 above, it is illegal to specify nodes that are not allowed in the cpuset.
316 The application must query the allowed nodes using the get_mempolicy()
317 API with the MPOL_F_MEMS_ALLOWED flag to determine the allowed nodes and
318 restrict itself to those nodes. However, the resources available to a
319 cpuset can be changed by the system administrator, or a workload manager
320 application, at any time. So, a task may still get errors attempting to
321 specify policy nodes, and must query the allowed memories again.
323 2) when tasks in two cpusets share access to a memory region, such as shared
324 memory segments created by shmget() of mmap() with the MAP_ANONYMOUS and
325 MAP_SHARED flags, and any of the tasks install shared policy on the region,
326 only nodes whose memories are allowed in both cpusets may be used in the
327 policies. Obtaining this information requires "stepping outside" the
328 memory policy APIs to use the cpuset information and requires that one
329 know in what cpusets other task might be attaching to the shared region.
330 Furthermore, if the cpusets' allowed memory sets are disjoint, "local"
331 allocation is the only valid policy.