FreeBSD/Linux Kernel Cross Reference
sys/mm/oom_kill.c
1 /*
2 * linux/mm/oom_kill.c
3 *
4 * Copyright (C) 1998,2000 Rik van Riel
5 * Thanks go out to Claus Fischer for some serious inspiration and
6 * for goading me into coding this file...
7 *
8 * The routines in this file are used to kill a process when
9 * we're seriously out of memory. This gets called from kswapd()
10 * in linux/mm/vmscan.c when we really run out of memory.
11 *
12 * Since we won't call these routines often (on a well-configured
13 * machine) this file will double as a 'coding guide' and a signpost
14 * for newbie kernel hackers. It features several pointers to major
15 * kernel subsystems and hints as to where to find out what things do.
16 */
17
18 #include <linux/mm.h>
19 #include <linux/sched.h>
20 #include <linux/swap.h>
21 #include <linux/swapctl.h>
22 #include <linux/timex.h>
23
24 /* #define DEBUG */
25
26 /**
27 * int_sqrt - oom_kill.c internal function, rough approximation to sqrt
28 * @x: integer of which to calculate the sqrt
29 *
30 * A very rough approximation to the sqrt() function.
31 */
32 static unsigned int int_sqrt(unsigned int x)
33 {
34 unsigned int out = x;
35 while (x & ~(unsigned int)1) x >>=2, out >>=1;
36 if (x) out -= out >> 2;
37 return (out ? out : 1);
38 }
39
40 /**
41 * oom_badness - calculate a numeric value for how bad this task has been
42 * @p: task struct of which task we should calculate
43 *
44 * The formula used is relatively simple and documented inline in the
45 * function. The main rationale is that we want to select a good task
46 * to kill when we run out of memory.
47 *
48 * Good in this context means that:
49 * 1) we lose the minimum amount of work done
50 * 2) we recover a large amount of memory
51 * 3) we don't kill anything innocent of eating tons of memory
52 * 4) we want to kill the minimum amount of processes (one)
53 * 5) we try to kill the process the user expects us to kill, this
54 * algorithm has been meticulously tuned to meet the priniciple
55 * of least surprise ... (be careful when you change it)
56 */
57
58 static int badness(struct task_struct *p)
59 {
60 int points, cpu_time, run_time;
61
62 if (!p->mm)
63 return 0;
64
65 if (p->flags & PF_MEMDIE)
66 return 0;
67
68 /*
69 * The memory size of the process is the basis for the badness.
70 */
71 points = p->mm->total_vm;
72
73 /*
74 * CPU time is in seconds and run time is in minutes. There is no
75 * particular reason for this other than that it turned out to work
76 * very well in practice. This is not safe against jiffie wraps
77 * but we don't care _that_ much...
78 */
79 cpu_time = (p->times.tms_utime + p->times.tms_stime) >> (SHIFT_HZ + 3);
80 run_time = (jiffies - p->start_time) >> (SHIFT_HZ + 10);
81
82 points /= int_sqrt(cpu_time);
83 points /= int_sqrt(int_sqrt(run_time));
84
85 /*
86 * Niced processes are most likely less important, so double
87 * their badness points.
88 */
89 if (p->nice > 0)
90 points *= 2;
91
92 /*
93 * Superuser processes are usually more important, so we make it
94 * less likely that we kill those.
95 */
96 if (cap_t(p->cap_effective) & CAP_TO_MASK(CAP_SYS_ADMIN) ||
97 p->uid == 0 || p->euid == 0)
98 points /= 4;
99
100 /*
101 * We don't want to kill a process with direct hardware access.
102 * Not only could that mess up the hardware, but usually users
103 * tend to only have this flag set on applications they think
104 * of as important.
105 */
106 if (cap_t(p->cap_effective) & CAP_TO_MASK(CAP_SYS_RAWIO))
107 points /= 4;
108 #ifdef DEBUG
109 printk(KERN_DEBUG "OOMkill: task %d (%s) got %d points\n",
110 p->pid, p->comm, points);
111 #endif
112 return points;
113 }
114
115 /*
116 * Simple selection loop. We chose the process with the highest
117 * number of 'points'. We expect the caller will lock the tasklist.
118 *
119 * (not docbooked, we don't want this one cluttering up the manual)
120 */
121 static struct task_struct * select_bad_process(void)
122 {
123 int maxpoints = 0;
124 struct task_struct *p = NULL;
125 struct task_struct *chosen = NULL;
126
127 for_each_task(p) {
128 if (p->pid) {
129 int points = badness(p);
130 if (points > maxpoints) {
131 chosen = p;
132 maxpoints = points;
133 }
134 }
135 }
136 return chosen;
137 }
138
139 /**
140 * We must be careful though to never send SIGKILL a process with
141 * CAP_SYS_RAW_IO set, send SIGTERM instead (but it's unlikely that
142 * we select a process with CAP_SYS_RAW_IO set).
143 */
144 void oom_kill_task(struct task_struct *p)
145 {
146 printk(KERN_ERR "Out of Memory: Killed process %d (%s).\n", p->pid, p->comm);
147
148 /*
149 * We give our sacrificial lamb high priority and access to
150 * all the memory it needs. That way it should be able to
151 * exit() and clear out its resources quickly...
152 */
153 p->counter = 5 * HZ;
154 p->flags |= PF_MEMALLOC | PF_MEMDIE;
155
156 /* This process has hardware access, be more careful. */
157 if (cap_t(p->cap_effective) & CAP_TO_MASK(CAP_SYS_RAWIO)) {
158 force_sig(SIGTERM, p);
159 } else {
160 force_sig(SIGKILL, p);
161 }
162 }
163
164 /**
165 * oom_kill - kill the "best" process when we run out of memory
166 *
167 * If we run out of memory, we have the choice between either
168 * killing a random task (bad), letting the system crash (worse)
169 * OR try to be smart about which process to kill. Note that we
170 * don't have to be perfect here, we just have to be good.
171 */
172 static void oom_kill(void)
173 {
174 struct task_struct *p, *q;
175
176 read_lock(&tasklist_lock);
177 p = select_bad_process();
178
179 /* Found nothing?!?! Either we hang forever, or we panic. */
180 if (p == NULL)
181 panic("Out of memory and no killable processes...\n");
182
183 /* kill all processes that share the ->mm (i.e. all threads) */
184 for_each_task(q) {
185 if (q->mm == p->mm)
186 oom_kill_task(q);
187 }
188 read_unlock(&tasklist_lock);
189
190 /*
191 * Make kswapd go out of the way, so "p" has a good chance of
192 * killing itself before someone else gets the chance to ask
193 * for more memory.
194 */
195 yield();
196 return;
197 }
198
199 /**
200 * out_of_memory - is the system out of memory?
201 */
202 void out_of_memory(void)
203 {
204 static unsigned long first, last, count, lastkill;
205 unsigned long now, since;
206
207 /*
208 * Enough swap space left? Not OOM.
209 */
210 if (nr_swap_pages > 0)
211 return;
212
213 now = jiffies;
214 since = now - last;
215 last = now;
216
217 /*
218 * If it's been a long time since last failure,
219 * we're not oom.
220 */
221 last = now;
222 if (since > 5*HZ)
223 goto reset;
224
225 /*
226 * If we haven't tried for at least one second,
227 * we're not really oom.
228 */
229 since = now - first;
230 if (since < HZ)
231 return;
232
233 /*
234 * If we have gotten only a few failures,
235 * we're not really oom.
236 */
237 if (++count < 10)
238 return;
239
240 /*
241 * If we just killed a process, wait a while
242 * to give that task a chance to exit. This
243 * avoids killing multiple processes needlessly.
244 */
245 since = now - lastkill;
246 if (since < HZ*5)
247 return;
248
249 /*
250 * Ok, really out of memory. Kill something.
251 */
252 lastkill = now;
253 oom_kill();
254
255 reset:
256 first = now;
257 count = 0;
258 }
Cache object: 88e9948792d72311fb547b41f1ba7525
|