Pavel Emelianov | e552b66 | 2008-02-07 00:13:49 -0800 | [diff] [blame] | 1 | #ifndef __RES_COUNTER_H__ |
| 2 | #define __RES_COUNTER_H__ |
| 3 | |
| 4 | /* |
| 5 | * Resource Counters |
| 6 | * Contain common data types and routines for resource accounting |
| 7 | * |
| 8 | * Copyright 2007 OpenVZ SWsoft Inc |
| 9 | * |
| 10 | * Author: Pavel Emelianov <xemul@openvz.org> |
| 11 | * |
Li Zefan | 45ce80f | 2009-01-15 13:50:59 -0800 | [diff] [blame] | 12 | * See Documentation/cgroups/resource_counter.txt for more |
Pavel Emelyanov | faebe9f | 2008-04-29 01:00:18 -0700 | [diff] [blame] | 13 | * info about what this counter is. |
Pavel Emelianov | e552b66 | 2008-02-07 00:13:49 -0800 | [diff] [blame] | 14 | */ |
| 15 | |
| 16 | #include <linux/cgroup.h> |
| 17 | |
| 18 | /* |
| 19 | * The core object. the cgroup that wishes to account for some |
| 20 | * resource may include this counter into its structures and use |
| 21 | * the helpers described beyond |
| 22 | */ |
| 23 | |
| 24 | struct res_counter { |
| 25 | /* |
| 26 | * the current resource consumption level |
| 27 | */ |
Balbir Singh | 0eea103 | 2008-02-07 00:13:57 -0800 | [diff] [blame] | 28 | unsigned long long usage; |
Pavel Emelianov | e552b66 | 2008-02-07 00:13:49 -0800 | [diff] [blame] | 29 | /* |
Pavel Emelyanov | c84872e | 2008-04-29 01:00:17 -0700 | [diff] [blame] | 30 | * the maximal value of the usage from the counter creation |
| 31 | */ |
| 32 | unsigned long long max_usage; |
| 33 | /* |
Pavel Emelianov | e552b66 | 2008-02-07 00:13:49 -0800 | [diff] [blame] | 34 | * the limit that usage cannot exceed |
| 35 | */ |
Balbir Singh | 0eea103 | 2008-02-07 00:13:57 -0800 | [diff] [blame] | 36 | unsigned long long limit; |
Pavel Emelianov | e552b66 | 2008-02-07 00:13:49 -0800 | [diff] [blame] | 37 | /* |
| 38 | * the number of unsuccessful attempts to consume the resource |
| 39 | */ |
Balbir Singh | 0eea103 | 2008-02-07 00:13:57 -0800 | [diff] [blame] | 40 | unsigned long long failcnt; |
Pavel Emelianov | e552b66 | 2008-02-07 00:13:49 -0800 | [diff] [blame] | 41 | /* |
| 42 | * the lock to protect all of the above. |
| 43 | * the routines below consider this to be IRQ-safe |
| 44 | */ |
| 45 | spinlock_t lock; |
Balbir Singh | 28dbc4b | 2009-01-07 18:08:05 -0800 | [diff] [blame] | 46 | /* |
| 47 | * Parent counter, used for hierarchial resource accounting |
| 48 | */ |
| 49 | struct res_counter *parent; |
Pavel Emelianov | e552b66 | 2008-02-07 00:13:49 -0800 | [diff] [blame] | 50 | }; |
| 51 | |
Paul Menage | 2c7eabf | 2008-04-29 00:59:58 -0700 | [diff] [blame] | 52 | /** |
Pavel Emelianov | e552b66 | 2008-02-07 00:13:49 -0800 | [diff] [blame] | 53 | * Helpers to interact with userspace |
Paul Menage | 2c7eabf | 2008-04-29 00:59:58 -0700 | [diff] [blame] | 54 | * res_counter_read_u64() - returns the value of the specified member. |
Pavel Emelianov | e552b66 | 2008-02-07 00:13:49 -0800 | [diff] [blame] | 55 | * res_counter_read/_write - put/get the specified fields from the |
| 56 | * res_counter struct to/from the user |
| 57 | * |
| 58 | * @counter: the counter in question |
| 59 | * @member: the field to work with (see RES_xxx below) |
| 60 | * @buf: the buffer to opeate on,... |
| 61 | * @nbytes: its size... |
| 62 | * @pos: and the offset. |
| 63 | */ |
| 64 | |
Paul Menage | 2c7eabf | 2008-04-29 00:59:58 -0700 | [diff] [blame] | 65 | u64 res_counter_read_u64(struct res_counter *counter, int member); |
| 66 | |
Pavel Emelianov | e552b66 | 2008-02-07 00:13:49 -0800 | [diff] [blame] | 67 | ssize_t res_counter_read(struct res_counter *counter, int member, |
Balbir Singh | 0eea103 | 2008-02-07 00:13:57 -0800 | [diff] [blame] | 68 | const char __user *buf, size_t nbytes, loff_t *pos, |
| 69 | int (*read_strategy)(unsigned long long val, char *s)); |
Paul Menage | 856c13a | 2008-07-25 01:47:04 -0700 | [diff] [blame] | 70 | |
| 71 | typedef int (*write_strategy_fn)(const char *buf, unsigned long long *val); |
| 72 | |
| 73 | int res_counter_memparse_write_strategy(const char *buf, |
| 74 | unsigned long long *res); |
| 75 | |
| 76 | int res_counter_write(struct res_counter *counter, int member, |
| 77 | const char *buffer, write_strategy_fn write_strategy); |
Pavel Emelianov | e552b66 | 2008-02-07 00:13:49 -0800 | [diff] [blame] | 78 | |
| 79 | /* |
| 80 | * the field descriptors. one for each member of res_counter |
| 81 | */ |
| 82 | |
| 83 | enum { |
| 84 | RES_USAGE, |
Pavel Emelyanov | c84872e | 2008-04-29 01:00:17 -0700 | [diff] [blame] | 85 | RES_MAX_USAGE, |
Pavel Emelianov | e552b66 | 2008-02-07 00:13:49 -0800 | [diff] [blame] | 86 | RES_LIMIT, |
| 87 | RES_FAILCNT, |
| 88 | }; |
| 89 | |
| 90 | /* |
| 91 | * helpers for accounting |
| 92 | */ |
| 93 | |
Balbir Singh | 28dbc4b | 2009-01-07 18:08:05 -0800 | [diff] [blame] | 94 | void res_counter_init(struct res_counter *counter, struct res_counter *parent); |
Pavel Emelianov | e552b66 | 2008-02-07 00:13:49 -0800 | [diff] [blame] | 95 | |
| 96 | /* |
| 97 | * charge - try to consume more resource. |
| 98 | * |
| 99 | * @counter: the counter |
| 100 | * @val: the amount of the resource. each controller defines its own |
| 101 | * units, e.g. numbers, bytes, Kbytes, etc |
| 102 | * |
| 103 | * returns 0 on success and <0 if the counter->usage will exceed the |
| 104 | * counter->limit _locked call expects the counter->lock to be taken |
| 105 | */ |
| 106 | |
Pavel Emelyanov | f2992db | 2008-07-25 01:46:55 -0700 | [diff] [blame] | 107 | int __must_check res_counter_charge_locked(struct res_counter *counter, |
| 108 | unsigned long val); |
| 109 | int __must_check res_counter_charge(struct res_counter *counter, |
Balbir Singh | 28dbc4b | 2009-01-07 18:08:05 -0800 | [diff] [blame] | 110 | unsigned long val, struct res_counter **limit_fail_at); |
Pavel Emelianov | e552b66 | 2008-02-07 00:13:49 -0800 | [diff] [blame] | 111 | |
| 112 | /* |
| 113 | * uncharge - tell that some portion of the resource is released |
| 114 | * |
| 115 | * @counter: the counter |
| 116 | * @val: the amount of the resource |
| 117 | * |
| 118 | * these calls check for usage underflow and show a warning on the console |
| 119 | * _locked call expects the counter->lock to be taken |
| 120 | */ |
| 121 | |
| 122 | void res_counter_uncharge_locked(struct res_counter *counter, unsigned long val); |
| 123 | void res_counter_uncharge(struct res_counter *counter, unsigned long val); |
| 124 | |
Balbir Singh | 66e1707 | 2008-02-07 00:13:56 -0800 | [diff] [blame] | 125 | static inline bool res_counter_limit_check_locked(struct res_counter *cnt) |
| 126 | { |
| 127 | if (cnt->usage < cnt->limit) |
| 128 | return true; |
| 129 | |
| 130 | return false; |
| 131 | } |
| 132 | |
| 133 | /* |
| 134 | * Helper function to detect if the cgroup is within it's limit or |
| 135 | * not. It's currently called from cgroup_rss_prepare() |
| 136 | */ |
| 137 | static inline bool res_counter_check_under_limit(struct res_counter *cnt) |
| 138 | { |
| 139 | bool ret; |
| 140 | unsigned long flags; |
| 141 | |
| 142 | spin_lock_irqsave(&cnt->lock, flags); |
| 143 | ret = res_counter_limit_check_locked(cnt); |
| 144 | spin_unlock_irqrestore(&cnt->lock, flags); |
| 145 | return ret; |
| 146 | } |
| 147 | |
Pavel Emelyanov | c84872e | 2008-04-29 01:00:17 -0700 | [diff] [blame] | 148 | static inline void res_counter_reset_max(struct res_counter *cnt) |
| 149 | { |
| 150 | unsigned long flags; |
| 151 | |
| 152 | spin_lock_irqsave(&cnt->lock, flags); |
| 153 | cnt->max_usage = cnt->usage; |
| 154 | spin_unlock_irqrestore(&cnt->lock, flags); |
| 155 | } |
| 156 | |
Pavel Emelyanov | 29f2a4d | 2008-04-29 01:00:21 -0700 | [diff] [blame] | 157 | static inline void res_counter_reset_failcnt(struct res_counter *cnt) |
| 158 | { |
| 159 | unsigned long flags; |
| 160 | |
| 161 | spin_lock_irqsave(&cnt->lock, flags); |
| 162 | cnt->failcnt = 0; |
| 163 | spin_unlock_irqrestore(&cnt->lock, flags); |
| 164 | } |
KAMEZAWA Hiroyuki | 12b9804 | 2008-07-25 01:47:19 -0700 | [diff] [blame] | 165 | |
| 166 | static inline int res_counter_set_limit(struct res_counter *cnt, |
| 167 | unsigned long long limit) |
| 168 | { |
| 169 | unsigned long flags; |
| 170 | int ret = -EBUSY; |
| 171 | |
| 172 | spin_lock_irqsave(&cnt->lock, flags); |
Li Zefan | 11d55d2 | 2008-09-05 14:00:18 -0700 | [diff] [blame] | 173 | if (cnt->usage <= limit) { |
KAMEZAWA Hiroyuki | 12b9804 | 2008-07-25 01:47:19 -0700 | [diff] [blame] | 174 | cnt->limit = limit; |
| 175 | ret = 0; |
| 176 | } |
| 177 | spin_unlock_irqrestore(&cnt->lock, flags); |
| 178 | return ret; |
| 179 | } |
| 180 | |
Pavel Emelianov | e552b66 | 2008-02-07 00:13:49 -0800 | [diff] [blame] | 181 | #endif |