1 ================= 1 =================
2 Scheduler Domains 2 Scheduler Domains
3 ================= 3 =================
4 4
5 Each CPU has a "base" scheduling domain (struc 5 Each CPU has a "base" scheduling domain (struct sched_domain). The domain
6 hierarchy is built from these base domains via 6 hierarchy is built from these base domains via the ->parent pointer. ->parent
7 MUST be NULL terminated, and domain structures 7 MUST be NULL terminated, and domain structures should be per-CPU as they are
8 locklessly updated. 8 locklessly updated.
9 9
10 Each scheduling domain spans a number of CPUs 10 Each scheduling domain spans a number of CPUs (stored in the ->span field).
11 A domain's span MUST be a superset of it child 11 A domain's span MUST be a superset of it child's span (this restriction could
12 be relaxed if the need arises), and a base dom 12 be relaxed if the need arises), and a base domain for CPU i MUST span at least
13 i. The top domain for each CPU will generally 13 i. The top domain for each CPU will generally span all CPUs in the system
14 although strictly it doesn't have to, but this 14 although strictly it doesn't have to, but this could lead to a case where some
15 CPUs will never be given tasks to run unless t 15 CPUs will never be given tasks to run unless the CPUs allowed mask is
16 explicitly set. A sched domain's span means "b 16 explicitly set. A sched domain's span means "balance process load among these
17 CPUs". 17 CPUs".
18 18
19 Each scheduling domain must have one or more C 19 Each scheduling domain must have one or more CPU groups (struct sched_group)
20 which are organised as a circular one way link 20 which are organised as a circular one way linked list from the ->groups
21 pointer. The union of cpumasks of these groups 21 pointer. The union of cpumasks of these groups MUST be the same as the
22 domain's span. The group pointed to by the ->g 22 domain's span. The group pointed to by the ->groups pointer MUST contain the CPU
23 to which the domain belongs. Groups may be sha 23 to which the domain belongs. Groups may be shared among CPUs as they contain
24 read only data after they have been set up. Th 24 read only data after they have been set up. The intersection of cpumasks from
25 any two of these groups may be non empty. If t 25 any two of these groups may be non empty. If this is the case the SD_OVERLAP
26 flag is set on the corresponding scheduling do 26 flag is set on the corresponding scheduling domain and its groups may not be
27 shared between CPUs. 27 shared between CPUs.
28 28
29 Balancing within a sched domain occurs between 29 Balancing within a sched domain occurs between groups. That is, each group
30 is treated as one entity. The load of a group 30 is treated as one entity. The load of a group is defined as the sum of the
31 load of each of its member CPUs, and only when 31 load of each of its member CPUs, and only when the load of a group becomes
32 out of balance are tasks moved between groups. 32 out of balance are tasks moved between groups.
33 33
34 In kernel/sched/core.c, sched_balance_trigger( 34 In kernel/sched/core.c, sched_balance_trigger() is run periodically on each CPU
35 through sched_tick(). It raises a softirq afte 35 through sched_tick(). It raises a softirq after the next regularly scheduled
36 rebalancing event for the current runqueue has 36 rebalancing event for the current runqueue has arrived. The actual load
37 balancing workhorse, sched_balance_softirq()-> 37 balancing workhorse, sched_balance_softirq()->sched_balance_domains(), is then run
38 in softirq context (SCHED_SOFTIRQ). 38 in softirq context (SCHED_SOFTIRQ).
39 39
40 The latter function takes two arguments: the r 40 The latter function takes two arguments: the runqueue of current CPU and whether
41 the CPU was idle at the time the sched_tick() 41 the CPU was idle at the time the sched_tick() happened and iterates over all
42 sched domains our CPU is on, starting from its 42 sched domains our CPU is on, starting from its base domain and going up the ->parent
43 chain. While doing that, it checks to see if t 43 chain. While doing that, it checks to see if the current domain has exhausted its
44 rebalance interval. If so, it runs sched_balan 44 rebalance interval. If so, it runs sched_balance_rq() on that domain. It then checks
45 the parent sched_domain (if it exists), and th 45 the parent sched_domain (if it exists), and the parent of the parent and so
46 forth. 46 forth.
47 47
48 Initially, sched_balance_rq() finds the busies 48 Initially, sched_balance_rq() finds the busiest group in the current sched domain.
49 If it succeeds, it looks for the busiest runqu 49 If it succeeds, it looks for the busiest runqueue of all the CPUs' runqueues in
50 that group. If it manages to find such a runqu 50 that group. If it manages to find such a runqueue, it locks both our initial
51 CPU's runqueue and the newly found busiest one 51 CPU's runqueue and the newly found busiest one and starts moving tasks from it
52 to our runqueue. The exact number of tasks amo 52 to our runqueue. The exact number of tasks amounts to an imbalance previously
53 computed while iterating over this sched domai 53 computed while iterating over this sched domain's groups.
54 54
55 Implementing sched domains 55 Implementing sched domains
56 ========================== 56 ==========================
57 57
58 The "base" domain will "span" the first level 58 The "base" domain will "span" the first level of the hierarchy. In the case
59 of SMT, you'll span all siblings of the physic 59 of SMT, you'll span all siblings of the physical CPU, with each group being
60 a single virtual CPU. 60 a single virtual CPU.
61 61
62 In SMP, the parent of the base domain will spa 62 In SMP, the parent of the base domain will span all physical CPUs in the
63 node. Each group being a single physical CPU. 63 node. Each group being a single physical CPU. Then with NUMA, the parent
64 of the SMP domain will span the entire machine 64 of the SMP domain will span the entire machine, with each group having the
65 cpumask of a node. Or, you could do multi-leve 65 cpumask of a node. Or, you could do multi-level NUMA or Opteron, for example,
66 might have just one domain covering its one NU 66 might have just one domain covering its one NUMA level.
67 67
68 The implementor should read comments in includ 68 The implementor should read comments in include/linux/sched/sd_flags.h:
69 SD_* to get an idea of the specifics and what 69 SD_* to get an idea of the specifics and what to tune for the SD flags
70 of a sched_domain. 70 of a sched_domain.
71 71
72 Architectures may override the generic domain 72 Architectures may override the generic domain builder and the default SD flags
73 for a given topology level by creating a sched 73 for a given topology level by creating a sched_domain_topology_level array and
74 calling set_sched_topology() with this array a 74 calling set_sched_topology() with this array as the parameter.
75 75
76 The sched-domains debugging infrastructure can 76 The sched-domains debugging infrastructure can be enabled by enabling
77 CONFIG_SCHED_DEBUG and adding 'sched_verbose' 77 CONFIG_SCHED_DEBUG and adding 'sched_verbose' to your cmdline. If you
78 forgot to tweak your cmdline, you can also fli 78 forgot to tweak your cmdline, you can also flip the
79 /sys/kernel/debug/sched/verbose knob. This ena 79 /sys/kernel/debug/sched/verbose knob. This enables an error checking parse of
80 the sched domains which should catch most poss 80 the sched domains which should catch most possible errors (described above). It
81 also prints out the domain structure in a visu 81 also prints out the domain structure in a visual format.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.