]>
Commit | Line | Data |
---|---|---|
c757249a SN |
1 | /* |
2 | * taskstats.c - Export per-task statistics to userland | |
3 | * | |
4 | * Copyright (C) Shailabh Nagar, IBM Corp. 2006 | |
5 | * (C) Balbir Singh, IBM Corp. 2006 | |
6 | * | |
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. | |
11 | * | |
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. | |
16 | * | |
17 | */ | |
18 | ||
19 | #include <linux/kernel.h> | |
20 | #include <linux/taskstats_kern.h> | |
6f44993f | 21 | #include <linux/delayacct.h> |
c757249a SN |
22 | #include <net/genetlink.h> |
23 | #include <asm/atomic.h> | |
24 | ||
25 | static DEFINE_PER_CPU(__u32, taskstats_seqnum) = { 0 }; | |
26 | static int family_registered; | |
27 | kmem_cache_t *taskstats_cache; | |
c757249a SN |
28 | |
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, | |
34 | }; | |
35 | ||
36 | static struct nla_policy taskstats_cmd_get_policy[TASKSTATS_CMD_ATTR_MAX+1] | |
37 | __read_mostly = { | |
38 | [TASKSTATS_CMD_ATTR_PID] = { .type = NLA_U32 }, | |
39 | [TASKSTATS_CMD_ATTR_TGID] = { .type = NLA_U32 }, | |
40 | }; | |
41 | ||
42 | ||
43 | static int prepare_reply(struct genl_info *info, u8 cmd, struct sk_buff **skbp, | |
44 | void **replyp, size_t size) | |
45 | { | |
46 | struct sk_buff *skb; | |
47 | void *reply; | |
48 | ||
49 | /* | |
50 | * If new attributes are added, please revisit this allocation | |
51 | */ | |
52 | skb = nlmsg_new(size); | |
53 | if (!skb) | |
54 | return -ENOMEM; | |
55 | ||
56 | if (!info) { | |
57 | int seq = get_cpu_var(taskstats_seqnum)++; | |
58 | put_cpu_var(taskstats_seqnum); | |
59 | ||
60 | reply = genlmsg_put(skb, 0, seq, | |
61 | family.id, 0, 0, | |
62 | cmd, family.version); | |
63 | } else | |
64 | reply = genlmsg_put(skb, info->snd_pid, info->snd_seq, | |
65 | family.id, 0, 0, | |
66 | cmd, family.version); | |
67 | if (reply == NULL) { | |
68 | nlmsg_free(skb); | |
69 | return -EINVAL; | |
70 | } | |
71 | ||
72 | *skbp = skb; | |
73 | *replyp = reply; | |
74 | return 0; | |
75 | } | |
76 | ||
77 | static int send_reply(struct sk_buff *skb, pid_t pid, int event) | |
78 | { | |
79 | struct genlmsghdr *genlhdr = nlmsg_data((struct nlmsghdr *)skb->data); | |
80 | void *reply; | |
81 | int rc; | |
82 | ||
83 | reply = genlmsg_data(genlhdr); | |
84 | ||
85 | rc = genlmsg_end(skb, reply); | |
86 | if (rc < 0) { | |
87 | nlmsg_free(skb); | |
88 | return rc; | |
89 | } | |
90 | ||
91 | if (event == TASKSTATS_MSG_MULTICAST) | |
92 | return genlmsg_multicast(skb, pid, TASKSTATS_LISTEN_GROUP); | |
93 | return genlmsg_unicast(skb, pid); | |
94 | } | |
95 | ||
96 | static int fill_pid(pid_t pid, struct task_struct *pidtsk, | |
97 | struct taskstats *stats) | |
98 | { | |
99 | int rc; | |
100 | struct task_struct *tsk = pidtsk; | |
101 | ||
102 | if (!pidtsk) { | |
103 | read_lock(&tasklist_lock); | |
104 | tsk = find_task_by_pid(pid); | |
105 | if (!tsk) { | |
106 | read_unlock(&tasklist_lock); | |
107 | return -ESRCH; | |
108 | } | |
109 | get_task_struct(tsk); | |
110 | read_unlock(&tasklist_lock); | |
111 | } else | |
112 | get_task_struct(tsk); | |
113 | ||
114 | /* | |
115 | * Each accounting subsystem adds calls to its functions to | |
116 | * fill in relevant parts of struct taskstsats as follows | |
117 | * | |
118 | * rc = per-task-foo(stats, tsk); | |
119 | * if (rc) | |
120 | * goto err; | |
121 | */ | |
122 | ||
6f44993f SN |
123 | rc = delayacct_add_tsk(stats, tsk); |
124 | stats->version = TASKSTATS_VERSION; | |
125 | ||
126 | /* Define err: label here if needed */ | |
c757249a SN |
127 | put_task_struct(tsk); |
128 | return rc; | |
129 | ||
130 | } | |
131 | ||
132 | static int fill_tgid(pid_t tgid, struct task_struct *tgidtsk, | |
133 | struct taskstats *stats) | |
134 | { | |
c757249a | 135 | struct task_struct *tsk, *first; |
ad4ecbcb | 136 | unsigned long flags; |
c757249a | 137 | |
ad4ecbcb SN |
138 | /* |
139 | * Add additional stats from live tasks except zombie thread group | |
140 | * leaders who are already counted with the dead tasks | |
141 | */ | |
c757249a | 142 | first = tgidtsk; |
c757249a | 143 | if (!first) { |
ad4ecbcb | 144 | read_lock(&tasklist_lock); |
c757249a SN |
145 | first = find_task_by_pid(tgid); |
146 | if (!first) { | |
147 | read_unlock(&tasklist_lock); | |
148 | return -ESRCH; | |
149 | } | |
ad4ecbcb SN |
150 | get_task_struct(first); |
151 | read_unlock(&tasklist_lock); | |
152 | } else | |
153 | get_task_struct(first); | |
154 | ||
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); | |
160 | ||
c757249a | 161 | tsk = first; |
ad4ecbcb | 162 | read_lock(&tasklist_lock); |
c757249a | 163 | do { |
ad4ecbcb SN |
164 | if (tsk->exit_state == EXIT_ZOMBIE && thread_group_leader(tsk)) |
165 | continue; | |
c757249a | 166 | /* |
ad4ecbcb | 167 | * Accounting subsystem can call its functions here to |
c757249a SN |
168 | * fill in relevant parts of struct taskstsats as follows |
169 | * | |
ad4ecbcb | 170 | * per-task-foo(stats, tsk); |
c757249a | 171 | */ |
ad4ecbcb | 172 | delayacct_add_tsk(stats, tsk); |
6f44993f | 173 | |
c757249a SN |
174 | } while_each_thread(first, tsk); |
175 | read_unlock(&tasklist_lock); | |
6f44993f SN |
176 | stats->version = TASKSTATS_VERSION; |
177 | ||
c757249a | 178 | /* |
ad4ecbcb SN |
179 | * Accounting subsytems can also add calls here to modify |
180 | * fields of taskstats. | |
c757249a SN |
181 | */ |
182 | ||
ad4ecbcb SN |
183 | return 0; |
184 | } | |
185 | ||
186 | ||
187 | static void fill_tgid_exit(struct task_struct *tsk) | |
188 | { | |
189 | unsigned long flags; | |
190 | ||
191 | spin_lock_irqsave(&tsk->signal->stats_lock, flags); | |
192 | if (!tsk->signal->stats) | |
193 | goto ret; | |
194 | ||
195 | /* | |
196 | * Each accounting subsystem calls its functions here to | |
197 | * accumalate its per-task stats for tsk, into the per-tgid structure | |
198 | * | |
199 | * per-task-foo(tsk->signal->stats, tsk); | |
200 | */ | |
201 | delayacct_add_tsk(tsk->signal->stats, tsk); | |
202 | ret: | |
203 | spin_unlock_irqrestore(&tsk->signal->stats_lock, flags); | |
204 | return; | |
c757249a SN |
205 | } |
206 | ||
ad4ecbcb | 207 | |
c757249a SN |
208 | static int taskstats_send_stats(struct sk_buff *skb, struct genl_info *info) |
209 | { | |
210 | int rc = 0; | |
211 | struct sk_buff *rep_skb; | |
212 | struct taskstats stats; | |
213 | void *reply; | |
214 | size_t size; | |
215 | struct nlattr *na; | |
216 | ||
217 | /* | |
218 | * Size includes space for nested attributes | |
219 | */ | |
220 | size = nla_total_size(sizeof(u32)) + | |
221 | nla_total_size(sizeof(struct taskstats)) + nla_total_size(0); | |
222 | ||
223 | memset(&stats, 0, sizeof(stats)); | |
224 | rc = prepare_reply(info, TASKSTATS_CMD_NEW, &rep_skb, &reply, size); | |
225 | if (rc < 0) | |
226 | return rc; | |
227 | ||
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); | |
231 | if (rc < 0) | |
232 | goto err; | |
233 | ||
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, | |
237 | 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); | |
241 | if (rc < 0) | |
242 | goto err; | |
243 | ||
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, | |
247 | stats); | |
248 | } else { | |
249 | rc = -EINVAL; | |
250 | goto err; | |
251 | } | |
252 | ||
253 | nla_nest_end(rep_skb, na); | |
254 | ||
255 | return send_reply(rep_skb, info->snd_pid, TASKSTATS_MSG_UNICAST); | |
256 | ||
257 | nla_put_failure: | |
258 | return genlmsg_cancel(rep_skb, reply); | |
259 | err: | |
260 | nlmsg_free(rep_skb); | |
261 | return rc; | |
262 | } | |
263 | ||
264 | /* Send pid data out on exit */ | |
265 | void taskstats_exit_send(struct task_struct *tsk, struct taskstats *tidstats, | |
ad4ecbcb | 266 | int group_dead) |
c757249a SN |
267 | { |
268 | int rc; | |
269 | struct sk_buff *rep_skb; | |
270 | void *reply; | |
271 | size_t size; | |
272 | int is_thread_group; | |
273 | struct nlattr *na; | |
ad4ecbcb | 274 | unsigned long flags; |
c757249a SN |
275 | |
276 | if (!family_registered || !tidstats) | |
277 | return; | |
278 | ||
ad4ecbcb SN |
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); | |
c757249a | 282 | |
ad4ecbcb | 283 | rc = 0; |
c757249a SN |
284 | /* |
285 | * Size includes space for nested attributes | |
286 | */ | |
287 | size = nla_total_size(sizeof(u32)) + | |
288 | nla_total_size(sizeof(struct taskstats)) + nla_total_size(0); | |
289 | ||
290 | if (is_thread_group) | |
291 | size = 2 * size; /* PID + STATS + TGID + STATS */ | |
292 | ||
293 | rc = prepare_reply(NULL, TASKSTATS_CMD_NEW, &rep_skb, &reply, size); | |
294 | if (rc < 0) | |
295 | goto ret; | |
296 | ||
297 | rc = fill_pid(tsk->pid, tsk, tidstats); | |
298 | if (rc < 0) | |
299 | goto err_skb; | |
300 | ||
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, | |
304 | *tidstats); | |
305 | nla_nest_end(rep_skb, na); | |
306 | ||
ad4ecbcb SN |
307 | if (!is_thread_group) |
308 | goto send; | |
c757249a | 309 | |
c757249a | 310 | /* |
ad4ecbcb SN |
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 | |
c757249a | 313 | */ |
ad4ecbcb SN |
314 | |
315 | fill_tgid_exit(tsk); | |
316 | if (!group_dead) | |
317 | goto send; | |
c757249a SN |
318 | |
319 | na = nla_nest_start(rep_skb, TASKSTATS_TYPE_AGGR_TGID); | |
320 | NLA_PUT_U32(rep_skb, TASKSTATS_TYPE_TGID, (u32)tsk->tgid); | |
ad4ecbcb | 321 | /* No locking needed for tsk->signal->stats since group is dead */ |
c757249a | 322 | NLA_PUT_TYPE(rep_skb, struct taskstats, TASKSTATS_TYPE_STATS, |
ad4ecbcb | 323 | *tsk->signal->stats); |
c757249a SN |
324 | nla_nest_end(rep_skb, na); |
325 | ||
ad4ecbcb | 326 | send: |
c757249a | 327 | send_reply(rep_skb, 0, TASKSTATS_MSG_MULTICAST); |
ad4ecbcb | 328 | return; |
c757249a SN |
329 | |
330 | nla_put_failure: | |
331 | genlmsg_cancel(rep_skb, reply); | |
332 | goto ret; | |
333 | err_skb: | |
334 | nlmsg_free(rep_skb); | |
335 | ret: | |
c757249a SN |
336 | return; |
337 | } | |
338 | ||
339 | static struct genl_ops taskstats_ops = { | |
340 | .cmd = TASKSTATS_CMD_GET, | |
341 | .doit = taskstats_send_stats, | |
342 | .policy = taskstats_cmd_get_policy, | |
343 | }; | |
344 | ||
345 | /* Needed early in initialization */ | |
346 | void __init taskstats_init_early(void) | |
347 | { | |
348 | taskstats_cache = kmem_cache_create("taskstats_cache", | |
349 | sizeof(struct taskstats), | |
350 | 0, SLAB_PANIC, NULL, NULL); | |
351 | } | |
352 | ||
353 | static int __init taskstats_init(void) | |
354 | { | |
355 | int rc; | |
356 | ||
357 | rc = genl_register_family(&family); | |
358 | if (rc) | |
359 | return rc; | |
360 | ||
361 | rc = genl_register_ops(&family, &taskstats_ops); | |
362 | if (rc < 0) | |
363 | goto err; | |
364 | ||
365 | family_registered = 1; | |
366 | return 0; | |
367 | err: | |
368 | genl_unregister_family(&family); | |
369 | return rc; | |
370 | } | |
371 | ||
372 | /* | |
373 | * late initcall ensures initialization of statistics collection | |
374 | * mechanisms precedes initialization of the taskstats interface | |
375 | */ | |
376 | late_initcall(taskstats_init); |