1				CPUSETS
2				-------
3
4Copyright (C) 2004 BULL SA.
5Written by Simon.Derr@bull.net
6
7Portions Copyright (c) 2004-2006 Silicon Graphics, Inc.
8Modified by Paul Jackson <pj@sgi.com>
9Modified by Christoph Lameter <clameter@sgi.com>
10Modified by Paul Menage <menage@google.com>
11Modified by Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
12
13CONTENTS:
14=========
15
161. Cpusets
17  1.1 What are cpusets ?
18  1.2 Why are cpusets needed ?
19  1.3 How are cpusets implemented ?
20  1.4 What are exclusive cpusets ?
21  1.5 What is memory_pressure ?
22  1.6 What is memory spread ?
23  1.7 What is sched_load_balance ?
24  1.8 What is sched_relax_domain_level ?
25  1.9 How do I use cpusets ?
262. Usage Examples and Syntax
27  2.1 Basic Usage
28  2.2 Adding/removing cpus
29  2.3 Setting flags
30  2.4 Attaching processes
313. Questions
324. Contact
33
341. Cpusets
35==========
36
371.1 What are cpusets ?
38----------------------
39
40Cpusets provide a mechanism for assigning a set of CPUs and Memory
41Nodes to a set of tasks.   In this document "Memory Node" refers to
42an on-line node that contains memory.
43
44Cpusets constrain the CPU and Memory placement of tasks to only
45the resources within a task's current cpuset.  They form a nested
46hierarchy visible in a virtual file system.  These are the essential
47hooks, beyond what is already present, required to manage dynamic
48job placement on large systems.
49
50Cpusets use the generic cgroup subsystem described in
51Documentation/cgroups/cgroups.txt.
52
53Requests by a task, using the sched_setaffinity(2) system call to
54include CPUs in its CPU affinity mask, and using the mbind(2) and
55set_mempolicy(2) system calls to include Memory Nodes in its memory
56policy, are both filtered through that task's cpuset, filtering out any
57CPUs or Memory Nodes not in that cpuset.  The scheduler will not
58schedule a task on a CPU that is not allowed in its cpus_allowed
59vector, and the kernel page allocator will not allocate a page on a
60node that is not allowed in the requesting task's mems_allowed vector.
61
62User level code may create and destroy cpusets by name in the cgroup
63virtual file system, manage the attributes and permissions of these
64cpusets and which CPUs and Memory Nodes are assigned to each cpuset,
65specify and query to which cpuset a task is assigned, and list the
66task pids assigned to a cpuset.
67
68
691.2 Why are cpusets needed ?
70----------------------------
71
72The management of large computer systems, with many processors (CPUs),
73complex memory cache hierarchies and multiple Memory Nodes having
74non-uniform access times (NUMA) presents additional challenges for
75the efficient scheduling and memory placement of processes.
76
77Frequently more modest sized systems can be operated with adequate
78efficiency just by letting the operating system automatically share
79the available CPU and Memory resources amongst the requesting tasks.
80
81But larger systems, which benefit more from careful processor and
82memory placement to reduce memory access times and contention,
83and which typically represent a larger investment for the customer,
84can benefit from explicitly placing jobs on properly sized subsets of
85the system.
86
87This can be especially valuable on:
88
89    * Web Servers running multiple instances of the same web application,
90    * Servers running different applications (for instance, a web server
91      and a database), or
92    * NUMA systems running large HPC applications with demanding
93      performance characteristics.
94
95These subsets, or "soft partitions" must be able to be dynamically
96adjusted, as the job mix changes, without impacting other concurrently
97executing jobs. The location of the running jobs pages may also be moved
98when the memory locations are changed.
99
100The kernel cpuset patch provides the minimum essential kernel
101mechanisms required to efficiently implement such subsets.  It
102leverages existing CPU and Memory Placement facilities in the Linux
103kernel to avoid any additional impact on the critical scheduler or
104memory allocator code.
105
106
1071.3 How are cpusets implemented ?
108---------------------------------
109
110Cpusets provide a Linux kernel mechanism to constrain which CPUs and
111Memory Nodes are used by a process or set of processes.
112
113The Linux kernel already has a pair of mechanisms to specify on which
114CPUs a task may be scheduled (sched_setaffinity) and on which Memory
115Nodes it may obtain memory (mbind, set_mempolicy).
116
117Cpusets extends these two mechanisms as follows:
118
119 - Cpusets are sets of allowed CPUs and Memory Nodes, known to the
120   kernel.
121 - Each task in the system is attached to a cpuset, via a pointer
122   in the task structure to a reference counted cgroup structure.
123 - Calls to sched_setaffinity are filtered to just those CPUs
124   allowed in that task's cpuset.
125 - Calls to mbind and set_mempolicy are filtered to just
126   those Memory Nodes allowed in that task's cpuset.
127 - The root cpuset contains all the systems CPUs and Memory
128   Nodes.
129 - For any cpuset, one can define child cpusets containing a subset
130   of the parents CPU and Memory Node resources.
131 - The hierarchy of cpusets can be mounted at /dev/cpuset, for
132   browsing and manipulation from user space.
133 - A cpuset may be marked exclusive, which ensures that no other
134   cpuset (except direct ancestors and descendants) may contain
135   any overlapping CPUs or Memory Nodes.
136 - You can list all the tasks (by pid) attached to any cpuset.
137
138The implementation of cpusets requires a few, simple hooks
139into the rest of the kernel, none in performance critical paths:
140
141 - in init/main.c, to initialize the root cpuset at system boot.
142 - in fork and exit, to attach and detach a task from its cpuset.
143 - in sched_setaffinity, to mask the requested CPUs by what's
144   allowed in that task's cpuset.
145 - in sched.c migrate_live_tasks(), to keep migrating tasks within
146   the CPUs allowed by their cpuset, if possible.
147 - in the mbind and set_mempolicy system calls, to mask the requested
148   Memory Nodes by what's allowed in that task's cpuset.
149 - in page_alloc.c, to restrict memory to allowed nodes.
150 - in vmscan.c, to restrict page recovery to the current cpuset.
151
152You should mount the "cgroup" filesystem type in order to enable
153browsing and modifying the cpusets presently known to the kernel.  No
154new system calls are added for cpusets - all support for querying and
155modifying cpusets is via this cpuset file system.
156
157The /proc/<pid>/status file for each task has four added lines,
158displaying the task's cpus_allowed (on which CPUs it may be scheduled)
159and mems_allowed (on which Memory Nodes it may obtain memory),
160in the two formats seen in the following example:
161
162  Cpus_allowed:   ffffffff,ffffffff,ffffffff,ffffffff
163  Cpus_allowed_list:      0-127
164  Mems_allowed:   ffffffff,ffffffff
165  Mems_allowed_list:      0-63
166
167Each cpuset is represented by a directory in the cgroup file system
168containing (on top of the standard cgroup files) the following
169files describing that cpuset:
170
171 - cpuset.cpus: list of CPUs in that cpuset
172 - cpuset.mems: list of Memory Nodes in that cpuset
173 - cpuset.memory_migrate flag: if set, move pages to cpusets nodes
174 - cpuset.cpu_exclusive flag: is cpu placement exclusive?
175 - cpuset.mem_exclusive flag: is memory placement exclusive?
176 - cpuset.mem_hardwall flag:  is memory allocation hardwalled
177 - cpuset.memory_pressure: measure of how much paging pressure in cpuset
178 - cpuset.memory_spread_page flag: if set, spread page cache evenly on allowed nodes
179 - cpuset.memory_spread_slab flag: if set, spread slab cache evenly on allowed nodes
180 - cpuset.sched_load_balance flag: if set, load balance within CPUs on that cpuset
181 - cpuset.sched_relax_domain_level: the searching range when migrating tasks
182
183In addition, only the root cpuset has the following file:
184 - cpuset.memory_pressure_enabled flag: compute memory_pressure?
185
186New cpusets are created using the mkdir system call or shell
187command.  The properties of a cpuset, such as its flags, allowed
188CPUs and Memory Nodes, and attached tasks, are modified by writing
189to the appropriate file in that cpusets directory, as listed above.
190
191The named hierarchical structure of nested cpusets allows partitioning
192a large system into nested, dynamically changeable, "soft-partitions".
193
194The attachment of each task, automatically inherited at fork by any
195children of that task, to a cpuset allows organizing the work load
196on a system into related sets of tasks such that each set is constrained
197to using the CPUs and Memory Nodes of a particular cpuset.  A task
198may be re-attached to any other cpuset, if allowed by the permissions
199on the necessary cpuset file system directories.
200
201Such management of a system "in the large" integrates smoothly with
202the detailed placement done on individual tasks and memory regions
203using the sched_setaffinity, mbind and set_mempolicy system calls.
204
205The following rules apply to each cpuset:
206
207 - Its CPUs and Memory Nodes must be a subset of its parents.
208 - It can't be marked exclusive unless its parent is.
209 - If its cpu or memory is exclusive, they may not overlap any sibling.
210
211These rules, and the natural hierarchy of cpusets, enable efficient
212enforcement of the exclusive guarantee, without having to scan all
213cpusets every time any of them change to ensure nothing overlaps a
214exclusive cpuset.  Also, the use of a Linux virtual file system (vfs)
215to represent the cpuset hierarchy provides for a familiar permission
216and name space for cpusets, with a minimum of additional kernel code.
217
218The cpus and mems files in the root (top_cpuset) cpuset are
219read-only.  The cpus file automatically tracks the value of
220cpu_online_mask using a CPU hotplug notifier, and the mems file
221automatically tracks the value of node_states[N_MEMORY]--i.e.,
222nodes with memory--using the cpuset_track_online_nodes() hook.
223
224
2251.4 What are exclusive cpusets ?
226--------------------------------
227
228If a cpuset is cpu or mem exclusive, no other cpuset, other than
229a direct ancestor or descendant, may share any of the same CPUs or
230Memory Nodes.
231
232A cpuset that is cpuset.mem_exclusive *or* cpuset.mem_hardwall is "hardwalled",
233i.e. it restricts kernel allocations for page, buffer and other data
234commonly shared by the kernel across multiple users.  All cpusets,
235whether hardwalled or not, restrict allocations of memory for user
236space.  This enables configuring a system so that several independent
237jobs can share common kernel data, such as file system pages, while
238isolating each job's user allocation in its own cpuset.  To do this,
239construct a large mem_exclusive cpuset to hold all the jobs, and
240construct child, non-mem_exclusive cpusets for each individual job.
241Only a small amount of typical kernel memory, such as requests from
242interrupt handlers, is allowed to be taken outside even a
243mem_exclusive cpuset.
244
245
2461.5 What is memory_pressure ?
247-----------------------------
248The memory_pressure of a cpuset provides a simple per-cpuset metric
249of the rate that the tasks in a cpuset are attempting to free up in
250use memory on the nodes of the cpuset to satisfy additional memory
251requests.
252
253This enables batch managers monitoring jobs running in dedicated
254cpusets to efficiently detect what level of memory pressure that job
255is causing.
256
257This is useful both on tightly managed systems running a wide mix of
258submitted jobs, which may choose to terminate or re-prioritize jobs that
259are trying to use more memory than allowed on the nodes assigned to them,
260and with tightly coupled, long running, massively parallel scientific
261computing jobs that will dramatically fail to meet required performance
262goals if they start to use more memory than allowed to them.
263
264This mechanism provides a very economical way for the batch manager
265to monitor a cpuset for signs of memory pressure.  It's up to the
266batch manager or other user code to decide what to do about it and
267take action.
268
269==> Unless this feature is enabled by writing "1" to the special file
270    /dev/cpuset/memory_pressure_enabled, the hook in the rebalance
271    code of __alloc_pages() for this metric reduces to simply noticing
272    that the cpuset_memory_pressure_enabled flag is zero.  So only
273    systems that enable this feature will compute the metric.
274
275Why a per-cpuset, running average:
276
277    Because this meter is per-cpuset, rather than per-task or mm,
278    the system load imposed by a batch scheduler monitoring this
279    metric is sharply reduced on large systems, because a scan of
280    the tasklist can be avoided on each set of queries.
281
282    Because this meter is a running average, instead of an accumulating
283    counter, a batch scheduler can detect memory pressure with a
284    single read, instead of having to read and accumulate results
285    for a period of time.
286
287    Because this meter is per-cpuset rather than per-task or mm,
288    the batch scheduler can obtain the key information, memory
289    pressure in a cpuset, with a single read, rather than having to
290    query and accumulate results over all the (dynamically changing)
291    set of tasks in the cpuset.
292
293A per-cpuset simple digital filter (requires a spinlock and 3 words
294of data per-cpuset) is kept, and updated by any task attached to that
295cpuset, if it enters the synchronous (direct) page reclaim code.
296
297A per-cpuset file provides an integer number representing the recent
298(half-life of 10 seconds) rate of direct page reclaims caused by
299the tasks in the cpuset, in units of reclaims attempted per second,
300times 1000.
301
302
3031.6 What is memory spread ?
304---------------------------
305There are two boolean flag files per cpuset that control where the
306kernel allocates pages for the file system buffers and related in
307kernel data structures.  They are called 'cpuset.memory_spread_page' and
308'cpuset.memory_spread_slab'.
309
310If the per-cpuset boolean flag file 'cpuset.memory_spread_page' is set, then
311the kernel will spread the file system buffers (page cache) evenly
312over all the nodes that the faulting task is allowed to use, instead
313of preferring to put those pages on the node where the task is running.
314
315If the per-cpuset boolean flag file 'cpuset.memory_spread_slab' is set,
316then the kernel will spread some file system related slab caches,
317such as for inodes and dentries evenly over all the nodes that the
318faulting task is allowed to use, instead of preferring to put those
319pages on the node where the task is running.
320
321The setting of these flags does not affect anonymous data segment or
322stack segment pages of a task.
323
324By default, both kinds of memory spreading are off, and memory
325pages are allocated on the node local to where the task is running,
326except perhaps as modified by the task's NUMA mempolicy or cpuset
327configuration, so long as sufficient free memory pages are available.
328
329When new cpusets are created, they inherit the memory spread settings
330of their parent.
331
332Setting memory spreading causes allocations for the affected page
333or slab caches to ignore the task's NUMA mempolicy and be spread
334instead.    Tasks using mbind() or set_mempolicy() calls to set NUMA
335mempolicies will not notice any change in these calls as a result of
336their containing task's memory spread settings.  If memory spreading
337is turned off, then the currently specified NUMA mempolicy once again
338applies to memory page allocations.
339
340Both 'cpuset.memory_spread_page' and 'cpuset.memory_spread_slab' are boolean flag
341files.  By default they contain "0", meaning that the feature is off
342for that cpuset.  If a "1" is written to that file, then that turns
343the named feature on.
344
345The implementation is simple.
346
347Setting the flag 'cpuset.memory_spread_page' turns on a per-process flag
348PFA_SPREAD_PAGE for each task that is in that cpuset or subsequently
349joins that cpuset.  The page allocation calls for the page cache
350is modified to perform an inline check for this PFA_SPREAD_PAGE task
351flag, and if set, a call to a new routine cpuset_mem_spread_node()
352returns the node to prefer for the allocation.
353
354Similarly, setting 'cpuset.memory_spread_slab' turns on the flag
355PFA_SPREAD_SLAB, and appropriately marked slab caches will allocate
356pages from the node returned by cpuset_mem_spread_node().
357
358The cpuset_mem_spread_node() routine is also simple.  It uses the
359value of a per-task rotor cpuset_mem_spread_rotor to select the next
360node in the current task's mems_allowed to prefer for the allocation.
361
362This memory placement policy is also known (in other contexts) as
363round-robin or interleave.
364
365This policy can provide substantial improvements for jobs that need
366to place thread local data on the corresponding node, but that need
367to access large file system data sets that need to be spread across
368the several nodes in the jobs cpuset in order to fit.  Without this
369policy, especially for jobs that might have one thread reading in the
370data set, the memory allocation across the nodes in the jobs cpuset
371can become very uneven.
372
3731.7 What is sched_load_balance ?
374--------------------------------
375
376The kernel scheduler (kernel/sched/core.c) automatically load balances
377tasks.  If one CPU is underutilized, kernel code running on that
378CPU will look for tasks on other more overloaded CPUs and move those
379tasks to itself, within the constraints of such placement mechanisms
380as cpusets and sched_setaffinity.
381
382The algorithmic cost of load balancing and its impact on key shared
383kernel data structures such as the task list increases more than
384linearly with the number of CPUs being balanced.  So the scheduler
385has support to partition the systems CPUs into a number of sched
386domains such that it only load balances within each sched domain.
387Each sched domain covers some subset of the CPUs in the system;
388no two sched domains overlap; some CPUs might not be in any sched
389domain and hence won't be load balanced.
390
391Put simply, it costs less to balance between two smaller sched domains
392than one big one, but doing so means that overloads in one of the
393two domains won't be load balanced to the other one.
394
395By default, there is one sched domain covering all CPUs, including those
396marked isolated using the kernel boot time "isolcpus=" argument. However,
397the isolated CPUs will not participate in load balancing, and will not
398have tasks running on them unless explicitly assigned.
399
400This default load balancing across all CPUs is not well suited for
401the following two situations:
402 1) On large systems, load balancing across many CPUs is expensive.
403    If the system is managed using cpusets to place independent jobs
404    on separate sets of CPUs, full load balancing is unnecessary.
405 2) Systems supporting realtime on some CPUs need to minimize
406    system overhead on those CPUs, including avoiding task load
407    balancing if that is not needed.
408
409When the per-cpuset flag "cpuset.sched_load_balance" is enabled (the default
410setting), it requests that all the CPUs in that cpusets allowed 'cpuset.cpus'
411be contained in a single sched domain, ensuring that load balancing
412can move a task (not otherwised pinned, as by sched_setaffinity)
413from any CPU in that cpuset to any other.
414
415When the per-cpuset flag "cpuset.sched_load_balance" is disabled, then the
416scheduler will avoid load balancing across the CPUs in that cpuset,
417--except-- in so far as is necessary because some overlapping cpuset
418has "sched_load_balance" enabled.
419
420So, for example, if the top cpuset has the flag "cpuset.sched_load_balance"
421enabled, then the scheduler will have one sched domain covering all
422CPUs, and the setting of the "cpuset.sched_load_balance" flag in any other
423cpusets won't matter, as we're already fully load balancing.
424
425Therefore in the above two situations, the top cpuset flag
426"cpuset.sched_load_balance" should be disabled, and only some of the smaller,
427child cpusets have this flag enabled.
428
429When doing this, you don't usually want to leave any unpinned tasks in
430the top cpuset that might use non-trivial amounts of CPU, as such tasks
431may be artificially constrained to some subset of CPUs, depending on
432the particulars of this flag setting in descendant cpusets.  Even if
433such a task could use spare CPU cycles in some other CPUs, the kernel
434scheduler might not consider the possibility of load balancing that
435task to that underused CPU.
436
437Of course, tasks pinned to a particular CPU can be left in a cpuset
438that disables "cpuset.sched_load_balance" as those tasks aren't going anywhere
439else anyway.
440
441There is an impedance mismatch here, between cpusets and sched domains.
442Cpusets are hierarchical and nest.  Sched domains are flat; they don't
443overlap and each CPU is in at most one sched domain.
444
445It is necessary for sched domains to be flat because load balancing
446across partially overlapping sets of CPUs would risk unstable dynamics
447that would be beyond our understanding.  So if each of two partially
448overlapping cpusets enables the flag 'cpuset.sched_load_balance', then we
449form a single sched domain that is a superset of both.  We won't move
450a task to a CPU outside its cpuset, but the scheduler load balancing
451code might waste some compute cycles considering that possibility.
452
453This mismatch is why there is not a simple one-to-one relation
454between which cpusets have the flag "cpuset.sched_load_balance" enabled,
455and the sched domain configuration.  If a cpuset enables the flag, it
456will get balancing across all its CPUs, but if it disables the flag,
457it will only be assured of no load balancing if no other overlapping
458cpuset enables the flag.
459
460If two cpusets have partially overlapping 'cpuset.cpus' allowed, and only
461one of them has this flag enabled, then the other may find its
462tasks only partially load balanced, just on the overlapping CPUs.
463This is just the general case of the top_cpuset example given a few
464paragraphs above.  In the general case, as in the top cpuset case,
465don't leave tasks that might use non-trivial amounts of CPU in
466such partially load balanced cpusets, as they may be artificially
467constrained to some subset of the CPUs allowed to them, for lack of
468load balancing to the other CPUs.
469
470CPUs in "cpuset.isolcpus" were excluded from load balancing by the
471isolcpus= kernel boot option, and will never be load balanced regardless
472of the value of "cpuset.sched_load_balance" in any cpuset.
473
4741.7.1 sched_load_balance implementation details.
475------------------------------------------------
476
477The per-cpuset flag 'cpuset.sched_load_balance' defaults to enabled (contrary
478to most cpuset flags.)  When enabled for a cpuset, the kernel will
479ensure that it can load balance across all the CPUs in that cpuset
480(makes sure that all the CPUs in the cpus_allowed of that cpuset are
481in the same sched domain.)
482
483If two overlapping cpusets both have 'cpuset.sched_load_balance' enabled,
484then they will be (must be) both in the same sched domain.
485
486If, as is the default, the top cpuset has 'cpuset.sched_load_balance' enabled,
487then by the above that means there is a single sched domain covering
488the whole system, regardless of any other cpuset settings.
489
490The kernel commits to user space that it will avoid load balancing
491where it can.  It will pick as fine a granularity partition of sched
492domains as it can while still providing load balancing for any set
493of CPUs allowed to a cpuset having 'cpuset.sched_load_balance' enabled.
494
495The internal kernel cpuset to scheduler interface passes from the
496cpuset code to the scheduler code a partition of the load balanced
497CPUs in the system. This partition is a set of subsets (represented
498as an array of struct cpumask) of CPUs, pairwise disjoint, that cover
499all the CPUs that must be load balanced.
500
501The cpuset code builds a new such partition and passes it to the
502scheduler sched domain setup code, to have the sched domains rebuilt
503as necessary, whenever:
504 - the 'cpuset.sched_load_balance' flag of a cpuset with non-empty CPUs changes,
505 - or CPUs come or go from a cpuset with this flag enabled,
506 - or 'cpuset.sched_relax_domain_level' value of a cpuset with non-empty CPUs
507   and with this flag enabled changes,
508 - or a cpuset with non-empty CPUs and with this flag enabled is removed,
509 - or a cpu is offlined/onlined.
510
511This partition exactly defines what sched domains the scheduler should
512setup - one sched domain for each element (struct cpumask) in the
513partition.
514
515The scheduler remembers the currently active sched domain partitions.
516When the scheduler routine partition_sched_domains() is invoked from
517the cpuset code to update these sched domains, it compares the new
518partition requested with the current, and updates its sched domains,
519removing the old and adding the new, for each change.
520
521
5221.8 What is sched_relax_domain_level ?
523--------------------------------------
524
525In sched domain, the scheduler migrates tasks in 2 ways; periodic load
526balance on tick, and at time of some schedule events.
527
528When a task is woken up, scheduler try to move the task on idle CPU.
529For example, if a task A running on CPU X activates another task B
530on the same CPU X, and if CPU Y is X's sibling and performing idle,
531then scheduler migrate task B to CPU Y so that task B can start on
532CPU Y without waiting task A on CPU X.
533
534And if a CPU run out of tasks in its runqueue, the CPU try to pull
535extra tasks from other busy CPUs to help them before it is going to
536be idle.
537
538Of course it takes some searching cost to find movable tasks and/or
539idle CPUs, the scheduler might not search all CPUs in the domain
540every time.  In fact, in some architectures, the searching ranges on
541events are limited in the same socket or node where the CPU locates,
542while the load balance on tick searches all.
543
544For example, assume CPU Z is relatively far from CPU X.  Even if CPU Z
545is idle while CPU X and the siblings are busy, scheduler can't migrate
546woken task B from X to Z since it is out of its searching range.
547As the result, task B on CPU X need to wait task A or wait load balance
548on the next tick.  For some applications in special situation, waiting
5491 tick may be too long.
550
551The 'cpuset.sched_relax_domain_level' file allows you to request changing
552this searching range as you like.  This file takes int value which
553indicates size of searching range in levels ideally as follows,
554otherwise initial value -1 that indicates the cpuset has no request.
555
556  -1  : no request. use system default or follow request of others.
557   0  : no search.
558   1  : search siblings (hyperthreads in a core).
559   2  : search cores in a package.
560   3  : search cpus in a node [= system wide on non-NUMA system]
561   4  : search nodes in a chunk of node [on NUMA system]
562   5  : search system wide [on NUMA system]
563
564The system default is architecture dependent.  The system default
565can be changed using the relax_domain_level= boot parameter.
566
567This file is per-cpuset and affect the sched domain where the cpuset
568belongs to.  Therefore if the flag 'cpuset.sched_load_balance' of a cpuset
569is disabled, then 'cpuset.sched_relax_domain_level' have no effect since
570there is no sched domain belonging the cpuset.
571
572If multiple cpusets are overlapping and hence they form a single sched
573domain, the largest value among those is used.  Be careful, if one
574requests 0 and others are -1 then 0 is used.
575
576Note that modifying this file will have both good and bad effects,
577and whether it is acceptable or not depends on your situation.
578Don't modify this file if you are not sure.
579
580If your situation is:
581 - The migration costs between each cpu can be assumed considerably
582   small(for you) due to your special application's behavior or
583   special hardware support for CPU cache etc.
584 - The searching cost doesn't have impact(for you) or you can make
585   the searching cost enough small by managing cpuset to compact etc.
586 - The latency is required even it sacrifices cache hit rate etc.
587then increasing 'sched_relax_domain_level' would benefit you.
588
589
5901.9 How do I use cpusets ?
591--------------------------
592
593In order to minimize the impact of cpusets on critical kernel
594code, such as the scheduler, and due to the fact that the kernel
595does not support one task updating the memory placement of another
596task directly, the impact on a task of changing its cpuset CPU
597or Memory Node placement, or of changing to which cpuset a task
598is attached, is subtle.
599
600If a cpuset has its Memory Nodes modified, then for each task attached
601to that cpuset, the next time that the kernel attempts to allocate
602a page of memory for that task, the kernel will notice the change
603in the task's cpuset, and update its per-task memory placement to
604remain within the new cpusets memory placement.  If the task was using
605mempolicy MPOL_BIND, and the nodes to which it was bound overlap with
606its new cpuset, then the task will continue to use whatever subset
607of MPOL_BIND nodes are still allowed in the new cpuset.  If the task
608was using MPOL_BIND and now none of its MPOL_BIND nodes are allowed
609in the new cpuset, then the task will be essentially treated as if it
610was MPOL_BIND bound to the new cpuset (even though its NUMA placement,
611as queried by get_mempolicy(), doesn't change).  If a task is moved
612from one cpuset to another, then the kernel will adjust the task's
613memory placement, as above, the next time that the kernel attempts
614to allocate a page of memory for that task.
615
616If a cpuset has its 'cpuset.cpus' modified, then each task in that cpuset
617will have its allowed CPU placement changed immediately.  Similarly,
618if a task's pid is written to another cpusets 'cpuset.tasks' file, then its
619allowed CPU placement is changed immediately.  If such a task had been
620bound to some subset of its cpuset using the sched_setaffinity() call,
621the task will be allowed to run on any CPU allowed in its new cpuset,
622negating the effect of the prior sched_setaffinity() call.
623
624In summary, the memory placement of a task whose cpuset is changed is
625updated by the kernel, on the next allocation of a page for that task,
626and the processor placement is updated immediately.
627
628Normally, once a page is allocated (given a physical page
629of main memory) then that page stays on whatever node it
630was allocated, so long as it remains allocated, even if the
631cpusets memory placement policy 'cpuset.mems' subsequently changes.
632If the cpuset flag file 'cpuset.memory_migrate' is set true, then when
633tasks are attached to that cpuset, any pages that task had
634allocated to it on nodes in its previous cpuset are migrated
635to the task's new cpuset. The relative placement of the page within
636the cpuset is preserved during these migration operations if possible.
637For example if the page was on the second valid node of the prior cpuset
638then the page will be placed on the second valid node of the new cpuset.
639
640Also if 'cpuset.memory_migrate' is set true, then if that cpuset's
641'cpuset.mems' file is modified, pages allocated to tasks in that
642cpuset, that were on nodes in the previous setting of 'cpuset.mems',
643will be moved to nodes in the new setting of 'mems.'
644Pages that were not in the task's prior cpuset, or in the cpuset's
645prior 'cpuset.mems' setting, will not be moved.
646
647There is an exception to the above.  If hotplug functionality is used
648to remove all the CPUs that are currently assigned to a cpuset,
649then all the tasks in that cpuset will be moved to the nearest ancestor
650with non-empty cpus.  But the moving of some (or all) tasks might fail if
651cpuset is bound with another cgroup subsystem which has some restrictions
652on task attaching.  In this failing case, those tasks will stay
653in the original cpuset, and the kernel will automatically update
654their cpus_allowed to allow all online CPUs.  When memory hotplug
655functionality for removing Memory Nodes is available, a similar exception
656is expected to apply there as well.  In general, the kernel prefers to
657violate cpuset placement, over starving a task that has had all
658its allowed CPUs or Memory Nodes taken offline.
659
660There is a second exception to the above.  GFP_ATOMIC requests are
661kernel internal allocations that must be satisfied, immediately.
662The kernel may drop some request, in rare cases even panic, if a
663GFP_ATOMIC alloc fails.  If the request cannot be satisfied within
664the current task's cpuset, then we relax the cpuset, and look for
665memory anywhere we can find it.  It's better to violate the cpuset
666than stress the kernel.
667
668To start a new job that is to be contained within a cpuset, the steps are:
669
670 1) mkdir /sys/fs/cgroup/cpuset
671 2) mount -t cgroup -ocpuset cpuset /sys/fs/cgroup/cpuset
672 3) Create the new cpuset by doing mkdir's and write's (or echo's) in
673    the /sys/fs/cgroup/cpuset virtual file system.
674 4) Start a task that will be the "founding father" of the new job.
675 5) Attach that task to the new cpuset by writing its pid to the
676    /sys/fs/cgroup/cpuset tasks file for that cpuset.
677 6) fork, exec or clone the job tasks from this founding father task.
678
679For example, the following sequence of commands will setup a cpuset
680named "Charlie", containing just CPUs 2 and 3, and Memory Node 1,
681and then start a subshell 'sh' in that cpuset:
682
683  mount -t cgroup -ocpuset cpuset /sys/fs/cgroup/cpuset
684  cd /sys/fs/cgroup/cpuset
685  mkdir Charlie
686  cd Charlie
687  /bin/echo 2-3 > cpuset.cpus
688  /bin/echo 1 > cpuset.mems
689  /bin/echo $$ > tasks
690  sh
691  # The subshell 'sh' is now running in cpuset Charlie
692  # The next line should display '/Charlie'
693  cat /proc/self/cpuset
694
695There are ways to query or modify cpusets:
696 - via the cpuset file system directly, using the various cd, mkdir, echo,
697   cat, rmdir commands from the shell, or their equivalent from C.
698 - via the C library libcpuset.
699 - via the C library libcgroup.
700   (http://sourceforge.net/projects/libcg/)
701 - via the python application cset.
702   (http://code.google.com/p/cpuset/)
703
704The sched_setaffinity calls can also be done at the shell prompt using
705SGI's runon or Robert Love's taskset.  The mbind and set_mempolicy
706calls can be done at the shell prompt using the numactl command
707(part of Andi Kleen's numa package).
708
7092. Usage Examples and Syntax
710============================
711
7122.1 Basic Usage
713---------------
714
715Creating, modifying, using the cpusets can be done through the cpuset
716virtual filesystem.
717
718To mount it, type:
719# mount -t cgroup -o cpuset cpuset /sys/fs/cgroup/cpuset
720
721Then under /sys/fs/cgroup/cpuset you can find a tree that corresponds to the
722tree of the cpusets in the system. For instance, /sys/fs/cgroup/cpuset
723is the cpuset that holds the whole system.
724
725If you want to create a new cpuset under /sys/fs/cgroup/cpuset:
726# cd /sys/fs/cgroup/cpuset
727# mkdir my_cpuset
728
729Now you want to do something with this cpuset.
730# cd my_cpuset
731
732In this directory you can find several files:
733# ls
734cgroup.clone_children  cpuset.memory_pressure
735cgroup.event_control   cpuset.memory_spread_page
736cgroup.procs           cpuset.memory_spread_slab
737cpuset.cpu_exclusive   cpuset.mems
738cpuset.cpus            cpuset.sched_load_balance
739cpuset.mem_exclusive   cpuset.sched_relax_domain_level
740cpuset.mem_hardwall    notify_on_release
741cpuset.memory_migrate  tasks
742
743Reading them will give you information about the state of this cpuset:
744the CPUs and Memory Nodes it can use, the processes that are using
745it, its properties.  By writing to these files you can manipulate
746the cpuset.
747
748Set some flags:
749# /bin/echo 1 > cpuset.cpu_exclusive
750
751Add some cpus:
752# /bin/echo 0-7 > cpuset.cpus
753
754Add some mems:
755# /bin/echo 0-7 > cpuset.mems
756
757Now attach your shell to this cpuset:
758# /bin/echo $$ > tasks
759
760You can also create cpusets inside your cpuset by using mkdir in this
761directory.
762# mkdir my_sub_cs
763
764To remove a cpuset, just use rmdir:
765# rmdir my_sub_cs
766This will fail if the cpuset is in use (has cpusets inside, or has
767processes attached).
768
769Note that for legacy reasons, the "cpuset" filesystem exists as a
770wrapper around the cgroup filesystem.
771
772The command
773
774mount -t cpuset X /sys/fs/cgroup/cpuset
775
776is equivalent to
777
778mount -t cgroup -ocpuset,noprefix X /sys/fs/cgroup/cpuset
779echo "/sbin/cpuset_release_agent" > /sys/fs/cgroup/cpuset/release_agent
780
7812.2 Adding/removing cpus
782------------------------
783
784This is the syntax to use when writing in the cpus or mems files
785in cpuset directories:
786
787# /bin/echo 1-4 > cpuset.cpus		-> set cpus list to cpus 1,2,3,4
788# /bin/echo 1,2,3,4 > cpuset.cpus	-> set cpus list to cpus 1,2,3,4
789
790To add a CPU to a cpuset, write the new list of CPUs including the
791CPU to be added. To add 6 to the above cpuset:
792
793# /bin/echo 1-4,6 > cpuset.cpus	-> set cpus list to cpus 1,2,3,4,6
794
795Similarly to remove a CPU from a cpuset, write the new list of CPUs
796without the CPU to be removed.
797
798To remove all the CPUs:
799
800# /bin/echo "" > cpuset.cpus		-> clear cpus list
801
8022.3 Setting flags
803-----------------
804
805The syntax is very simple:
806
807# /bin/echo 1 > cpuset.cpu_exclusive 	-> set flag 'cpuset.cpu_exclusive'
808# /bin/echo 0 > cpuset.cpu_exclusive 	-> unset flag 'cpuset.cpu_exclusive'
809
8102.4 Attaching processes
811-----------------------
812
813# /bin/echo PID > tasks
814
815Note that it is PID, not PIDs. You can only attach ONE task at a time.
816If you have several tasks to attach, you have to do it one after another:
817
818# /bin/echo PID1 > tasks
819# /bin/echo PID2 > tasks
820	...
821# /bin/echo PIDn > tasks
822
823
8243. Questions
825============
826
827Q: what's up with this '/bin/echo' ?
828A: bash's builtin 'echo' command does not check calls to write() against
829   errors. If you use it in the cpuset file system, you won't be
830   able to tell whether a command succeeded or failed.
831
832Q: When I attach processes, only the first of the line gets really attached !
833A: We can only return one error code per call to write(). So you should also
834   put only ONE pid.
835
8364. Contact
837==========
838
839Web: http://www.bullopensource.org/cpuset
840