4 The resource counter, declared at include/linux/res_counter.h,
5 is supposed to facilitate the resource management by controllers
6 by providing common stuff for accounting.
8 This "stuff" includes the res_counter structure and routines
13 1. Crucial parts of the res_counter structure
15 a. unsigned long long usage
17 The usage value shows the amount of a resource that is consumed
18 by a group at a given time. The units of measurement should be
19 determined by the controller that uses this counter. E.g. it can
20 be bytes, items or any other unit the controller operates on.
22 b. unsigned long long max_usage
24 The maximal value of the usage over time.
26 This value is useful when gathering statistical information about
27 the particular group, as it shows the actual resource requirements
28 for a particular group, not just some usage snapshot.
30 c. unsigned long long limit
32 The maximal allowed amount of resource to consume by the group. In
33 case the group requests for more resources, so that the usage value
34 would exceed the limit, the resource allocation is rejected (see
37 d. unsigned long long failcnt
39 The failcnt stands for "failures counter". This is the number of
40 resource allocation attempts that failed.
44 Protects changes of the above values.
48 2. Basic accounting routines
50 a. void res_counter_init(struct res_counter *rc,
51 struct res_counter *rc_parent)
53 Initializes the resource counter. As usual, should be the first
54 routine called for a new counter.
56 The struct res_counter *parent can be used to define a hierarchical
57 child -> parent relationship directly in the res_counter structure,
58 NULL can be used to define no relationship.
60 c. int res_counter_charge(struct res_counter *rc, unsigned long val,
61 struct res_counter **limit_fail_at)
63 When a resource is about to be allocated it has to be accounted
64 with the appropriate resource counter (controller should determine
65 which one to use on its own). This operation is called "charging".
67 This is not very important which operation - resource allocation
68 or charging - is performed first, but
69 * if the allocation is performed first, this may create a
70 temporary resource over-usage by the time resource counter is
72 * if the charging is performed first, then it should be uncharged
73 on error path (if the one is called).
75 If the charging fails and a hierarchical dependency exists, the
76 limit_fail_at parameter is set to the particular res_counter element
77 where the charging failed.
79 d. int res_counter_charge_locked
80 (struct res_counter *rc, unsigned long val, bool force)
82 The same as res_counter_charge(), but it must not acquire/release the
83 res_counter->lock internally (it must be called with res_counter->lock
84 held). The force parameter indicates whether we can bypass the limit.
86 e. void res_counter_uncharge[_locked]
87 (struct res_counter *rc, unsigned long val)
89 When a resource is released (freed) it should be de-accounted
90 from the resource counter it was accounted to. This is called
93 The _locked routines imply that the res_counter->lock is taken.
95 f. void res_counter_uncharge_until
96 (struct res_counter *rc, struct res_counter *top,
99 Almost same as res_cunter_uncharge() but propagation of uncharge
100 stops when rc == top. This is useful when kill a res_coutner in
103 2.1 Other accounting routines
105 There are more routines that may help you with common needs, like
106 checking whether the limit is reached or resetting the max_usage
107 value. They are all declared in include/linux/res_counter.h.
111 3. Analyzing the resource counter registrations
113 a. If the failcnt value constantly grows, this means that the counter's
114 limit is too tight. Either the group is misbehaving and consumes too
115 many resources, or the configuration is not suitable for the group
116 and the limit should be increased.
118 b. The max_usage value can be used to quickly tune the group. One may
119 set the limits to maximal values and either load the container with
120 a common pattern or leave one for a while. After this the max_usage
121 value shows the amount of memory the container would require during
124 Setting the limit a bit above this value gives a pretty good
125 configuration that works in most of the cases.
127 c. If the max_usage is much less than the limit, but the failcnt value
128 is growing, then the group tries to allocate a big chunk of resource
131 d. If the max_usage is much less than the limit, but the failcnt value
132 is 0, then this group is given too high limit, that it does not
133 require. It is better to lower the limit a bit leaving more resource
138 4. Communication with the control groups subsystem (cgroups)
140 All the resource controllers that are using cgroups and resource counters
141 should provide files (in the cgroup filesystem) to work with the resource
142 counter fields. They are recommended to adhere to the following rules:
147 ---------------------------------------------------
148 usage usage_in_<unit_of_measurement>
149 max_usage max_usage_in_<unit_of_measurement>
150 limit limit_in_<unit_of_measurement>
154 b. Reading from file should show the corresponding field value in the
159 Field Expected behavior
160 ----------------------------------
162 max_usage reset to usage
164 failcnt reset to zero
170 a. Declare a task group (take a look at cgroups subsystem for this) and
171 fold a res_counter into it
174 struct res_counter res;
179 b. Put hooks in resource allocation/release paths
181 int alloc_something(...)
183 if (res_counter_charge(res_counter_ptr, amount) < 0)
186 <allocate the resource and return to the caller>
189 void release_something(...)
191 res_counter_uncharge(res_counter_ptr, amount);
193 <release the resource>
196 In order to keep the usage value self-consistent, both the
197 "res_counter_ptr" and the "amount" in release_something() should be
198 the same as they were in the alloc_something() when the releasing
199 resource was allocated.
201 c. Provide the way to read res_counter values and set them (the cgroups
202 still can help with it).
204 c. Compile and run :)