2 * taskstats.c - Export per-task statistics to userland
4 * Copyright (C) Shailabh Nagar, IBM Corp. 2006
5 * (C) Balbir Singh, IBM Corp. 2006
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
19 #include <linux/kernel.h>
20 #include <linux/taskstats_kern.h>
21 #include <linux/delayacct.h>
22 #include <net/genetlink.h>
23 #include <asm/atomic.h>
25 static DEFINE_PER_CPU(__u32
, taskstats_seqnum
) = { 0 };
26 static int family_registered
;
27 kmem_cache_t
*taskstats_cache
;
29 static struct genl_family family
= {
30 .id
= GENL_ID_GENERATE
,
31 .name
= TASKSTATS_GENL_NAME
,
32 .version
= TASKSTATS_GENL_VERSION
,
33 .maxattr
= TASKSTATS_CMD_ATTR_MAX
,
36 static struct nla_policy taskstats_cmd_get_policy
[TASKSTATS_CMD_ATTR_MAX
+1]
38 [TASKSTATS_CMD_ATTR_PID
] = { .type
= NLA_U32
},
39 [TASKSTATS_CMD_ATTR_TGID
] = { .type
= NLA_U32
},
43 static int prepare_reply(struct genl_info
*info
, u8 cmd
, struct sk_buff
**skbp
,
44 void **replyp
, size_t size
)
50 * If new attributes are added, please revisit this allocation
52 skb
= nlmsg_new(size
);
57 int seq
= get_cpu_var(taskstats_seqnum
)++;
58 put_cpu_var(taskstats_seqnum
);
60 reply
= genlmsg_put(skb
, 0, seq
,
64 reply
= genlmsg_put(skb
, info
->snd_pid
, info
->snd_seq
,
77 static int send_reply(struct sk_buff
*skb
, pid_t pid
, int event
)
79 struct genlmsghdr
*genlhdr
= nlmsg_data((struct nlmsghdr
*)skb
->data
);
83 reply
= genlmsg_data(genlhdr
);
85 rc
= genlmsg_end(skb
, reply
);
91 if (event
== TASKSTATS_MSG_MULTICAST
)
92 return genlmsg_multicast(skb
, pid
, TASKSTATS_LISTEN_GROUP
);
93 return genlmsg_unicast(skb
, pid
);
96 static int fill_pid(pid_t pid
, struct task_struct
*pidtsk
,
97 struct taskstats
*stats
)
100 struct task_struct
*tsk
= pidtsk
;
103 read_lock(&tasklist_lock
);
104 tsk
= find_task_by_pid(pid
);
106 read_unlock(&tasklist_lock
);
109 get_task_struct(tsk
);
110 read_unlock(&tasklist_lock
);
112 get_task_struct(tsk
);
115 * Each accounting subsystem adds calls to its functions to
116 * fill in relevant parts of struct taskstsats as follows
118 * rc = per-task-foo(stats, tsk);
123 rc
= delayacct_add_tsk(stats
, tsk
);
124 stats
->version
= TASKSTATS_VERSION
;
126 /* Define err: label here if needed */
127 put_task_struct(tsk
);
132 static int fill_tgid(pid_t tgid
, struct task_struct
*tgidtsk
,
133 struct taskstats
*stats
)
135 struct task_struct
*tsk
, *first
;
139 * Add additional stats from live tasks except zombie thread group
140 * leaders who are already counted with the dead tasks
144 read_lock(&tasklist_lock
);
145 first
= find_task_by_pid(tgid
);
147 read_unlock(&tasklist_lock
);
150 get_task_struct(first
);
151 read_unlock(&tasklist_lock
);
153 get_task_struct(first
);
155 /* Start with stats from dead tasks */
156 spin_lock_irqsave(&first
->signal
->stats_lock
, flags
);
157 if (first
->signal
->stats
)
158 memcpy(stats
, first
->signal
->stats
, sizeof(*stats
));
159 spin_unlock_irqrestore(&first
->signal
->stats_lock
, flags
);
162 read_lock(&tasklist_lock
);
164 if (tsk
->exit_state
== EXIT_ZOMBIE
&& thread_group_leader(tsk
))
167 * Accounting subsystem can call its functions here to
168 * fill in relevant parts of struct taskstsats as follows
170 * per-task-foo(stats, tsk);
172 delayacct_add_tsk(stats
, tsk
);
174 } while_each_thread(first
, tsk
);
175 read_unlock(&tasklist_lock
);
176 stats
->version
= TASKSTATS_VERSION
;
179 * Accounting subsytems can also add calls here to modify
180 * fields of taskstats.
187 static void fill_tgid_exit(struct task_struct
*tsk
)
191 spin_lock_irqsave(&tsk
->signal
->stats_lock
, flags
);
192 if (!tsk
->signal
->stats
)
196 * Each accounting subsystem calls its functions here to
197 * accumalate its per-task stats for tsk, into the per-tgid structure
199 * per-task-foo(tsk->signal->stats, tsk);
201 delayacct_add_tsk(tsk
->signal
->stats
, tsk
);
203 spin_unlock_irqrestore(&tsk
->signal
->stats_lock
, flags
);
208 static int taskstats_send_stats(struct sk_buff
*skb
, struct genl_info
*info
)
211 struct sk_buff
*rep_skb
;
212 struct taskstats stats
;
218 * Size includes space for nested attributes
220 size
= nla_total_size(sizeof(u32
)) +
221 nla_total_size(sizeof(struct taskstats
)) + nla_total_size(0);
223 memset(&stats
, 0, sizeof(stats
));
224 rc
= prepare_reply(info
, TASKSTATS_CMD_NEW
, &rep_skb
, &reply
, size
);
228 if (info
->attrs
[TASKSTATS_CMD_ATTR_PID
]) {
229 u32 pid
= nla_get_u32(info
->attrs
[TASKSTATS_CMD_ATTR_PID
]);
230 rc
= fill_pid(pid
, NULL
, &stats
);
234 na
= nla_nest_start(rep_skb
, TASKSTATS_TYPE_AGGR_PID
);
235 NLA_PUT_U32(rep_skb
, TASKSTATS_TYPE_PID
, pid
);
236 NLA_PUT_TYPE(rep_skb
, struct taskstats
, TASKSTATS_TYPE_STATS
,
238 } else if (info
->attrs
[TASKSTATS_CMD_ATTR_TGID
]) {
239 u32 tgid
= nla_get_u32(info
->attrs
[TASKSTATS_CMD_ATTR_TGID
]);
240 rc
= fill_tgid(tgid
, NULL
, &stats
);
244 na
= nla_nest_start(rep_skb
, TASKSTATS_TYPE_AGGR_TGID
);
245 NLA_PUT_U32(rep_skb
, TASKSTATS_TYPE_TGID
, tgid
);
246 NLA_PUT_TYPE(rep_skb
, struct taskstats
, TASKSTATS_TYPE_STATS
,
253 nla_nest_end(rep_skb
, na
);
255 return send_reply(rep_skb
, info
->snd_pid
, TASKSTATS_MSG_UNICAST
);
258 return genlmsg_cancel(rep_skb
, reply
);
264 /* Send pid data out on exit */
265 void taskstats_exit_send(struct task_struct
*tsk
, struct taskstats
*tidstats
,
269 struct sk_buff
*rep_skb
;
276 if (!family_registered
|| !tidstats
)
279 spin_lock_irqsave(&tsk
->signal
->stats_lock
, flags
);
280 is_thread_group
= tsk
->signal
->stats
? 1 : 0;
281 spin_unlock_irqrestore(&tsk
->signal
->stats_lock
, flags
);
285 * Size includes space for nested attributes
287 size
= nla_total_size(sizeof(u32
)) +
288 nla_total_size(sizeof(struct taskstats
)) + nla_total_size(0);
291 size
= 2 * size
; /* PID + STATS + TGID + STATS */
293 rc
= prepare_reply(NULL
, TASKSTATS_CMD_NEW
, &rep_skb
, &reply
, size
);
297 rc
= fill_pid(tsk
->pid
, tsk
, tidstats
);
301 na
= nla_nest_start(rep_skb
, TASKSTATS_TYPE_AGGR_PID
);
302 NLA_PUT_U32(rep_skb
, TASKSTATS_TYPE_PID
, (u32
)tsk
->pid
);
303 NLA_PUT_TYPE(rep_skb
, struct taskstats
, TASKSTATS_TYPE_STATS
,
305 nla_nest_end(rep_skb
, na
);
307 if (!is_thread_group
)
311 * tsk has/had a thread group so fill the tsk->signal->stats structure
312 * Doesn't matter if tsk is the leader or the last group member leaving
319 na
= nla_nest_start(rep_skb
, TASKSTATS_TYPE_AGGR_TGID
);
320 NLA_PUT_U32(rep_skb
, TASKSTATS_TYPE_TGID
, (u32
)tsk
->tgid
);
321 /* No locking needed for tsk->signal->stats since group is dead */
322 NLA_PUT_TYPE(rep_skb
, struct taskstats
, TASKSTATS_TYPE_STATS
,
323 *tsk
->signal
->stats
);
324 nla_nest_end(rep_skb
, na
);
327 send_reply(rep_skb
, 0, TASKSTATS_MSG_MULTICAST
);
331 genlmsg_cancel(rep_skb
, reply
);
339 static struct genl_ops taskstats_ops
= {
340 .cmd
= TASKSTATS_CMD_GET
,
341 .doit
= taskstats_send_stats
,
342 .policy
= taskstats_cmd_get_policy
,
345 /* Needed early in initialization */
346 void __init
taskstats_init_early(void)
348 taskstats_cache
= kmem_cache_create("taskstats_cache",
349 sizeof(struct taskstats
),
350 0, SLAB_PANIC
, NULL
, NULL
);
353 static int __init
taskstats_init(void)
357 rc
= genl_register_family(&family
);
361 rc
= genl_register_ops(&family
, &taskstats_ops
);
365 family_registered
= 1;
368 genl_unregister_family(&family
);
373 * late initcall ensures initialization of statistics collection
374 * mechanisms precedes initialization of the taskstats interface
376 late_initcall(taskstats_init
);