-
-
Notifications
You must be signed in to change notification settings - Fork 289
/
core-out-of-memory.c
416 lines (372 loc) · 10.1 KB
/
core-out-of-memory.c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
/*
* Copyright (C) 2013-2021 Canonical, Ltd.
* Copyright (C) 2022-2024 Colin Ian King.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version 2
* of the License, or (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
*
*/
#include "stress-ng.h"
#include "core-attribute.h"
#include "core-killpid.h"
#include "core-out-of-memory.h"
#if defined(HAVE_SYS_PROCCTL_H)
#include <sys/procctl.h>
#endif
#define WAIT_TIMEOUT (120) /* waitpid timeout, 2 minutes */
#if defined(__linux__)
#define OOM_SCORE_ADJ_MIN "-1000"
#define OOM_SCORE_ADJ_MAX "1000"
#define OOM_ADJ_NO_OOM "-17"
#define OOM_ADJ_MIN "-16"
#define OOM_ADJ_MAX "15"
/*
* stress_process_oomed()
* check if a process has been logged as OOM killed
*/
bool stress_process_oomed(const pid_t pid)
{
int fd;
bool oomed = false;
fd = open("/dev/kmsg", O_RDONLY | O_NONBLOCK);
if (fd < 0)
return oomed;
for (;;) {
char buf[4096], *ptr;
ssize_t ret;
ret = read(fd, buf, sizeof(buf) - 1);
if (ret < 0)
break;
buf[ret] = '\0';
/*
* Look for 'Out of memory: Kill process 22566'
*/
ptr = strstr(buf, "process");
if (ptr && (strstr(buf, "Out of memory") ||
strstr(buf, "oom_reaper"))) {
intmax_t oom_pid;
if (sscanf(ptr + 7, "%10" SCNdMAX, &oom_pid) == 1) {
if (oom_pid == (intmax_t)pid) {
oomed = true;
break;
}
}
}
}
(void)close(fd);
return oomed;
}
static inline const char *stress_args_name(stress_args_t *args)
{
return args ? args->name : "main";
}
/*
* stress_set_adjustment()
* try to set OOM adjustment, retry if EAGAIN or EINTR, give up
* after multiple retries. Returns 0 for success, -errno for
* on failure.
*/
static int stress_set_adjustment(
stress_args_t *args,
const char *procname,
const char *str)
{
const size_t len = strlen(str);
int i, saved_errno = 0;
for (i = 0; i < 32; i++) {
ssize_t n;
int fd;
fd = open(procname, O_WRONLY);
if (fd < 0)
return -errno;
n = write(fd, str, len);
saved_errno = errno;
(void)close(fd);
if (n > 0)
return 0;
if (n < 0) {
if ((saved_errno != EAGAIN) &&
(saved_errno != EINTR) &&
(saved_errno != EACCES)) {
if (args && (args->instance == 0))
pr_dbg("%s: can't set oom_score_adj, errno=%d (%s)\n",
stress_args_name(args), saved_errno, strerror(saved_errno));
return -saved_errno;
}
}
}
/* Unexpected failure, report why */
if ((args && (args->instance == 0)) && (saved_errno != EACCES))
pr_dbg("%s: can't set oom_score_adj, errno=%d (%s)\n", stress_args_name(args),
saved_errno, strerror(saved_errno));
return -1;
}
/*
* stress_set_oom_adjustment()
* attempt to stop oom killer
* if we have root privileges then try and make process
* unkillable by oom killer
* NOTE: null args -> main stress-ng process, otherwise a stressor
*/
void stress_set_oom_adjustment(stress_args_t *args, const bool killable)
{
bool high_priv;
bool make_killable = killable;
char *str;
int ret;
/*
* --no-oom-adjust option ignores any oom adjustments
*/
if (g_opt_flags & OPT_FLAGS_NO_OOM_ADJUST)
return;
high_priv = (getuid() == 0) && (geteuid() == 0);
/*
* main cannot be killable; if OPT_FLAGS_OOMABLE set make
* all child procs easily OOMable
*/
if (args && (g_opt_flags & OPT_FLAGS_OOMABLE))
make_killable = true;
/*
* Try modern oom interface
*/
if (make_killable)
str = OOM_SCORE_ADJ_MAX;
else
str = high_priv ? OOM_SCORE_ADJ_MIN : "0";
ret = stress_set_adjustment(args, "/proc/self/oom_score_adj", str);
/*
* Success or some random failure that's not -ENOENT
*/
if ((ret == 0) || (ret != -ENOENT))
return;
/*
* Fall back to old oom interface if we got -ENOENT
*/
if (make_killable)
str = OOM_ADJ_MAX;
else
str = high_priv ? OOM_ADJ_NO_OOM : OOM_ADJ_MIN;
(void)stress_set_adjustment(args, "/proc/self/oom_adj", str);
}
#elif defined(HAVE_SYS_PROCCTL_H) && \
defined(__FreeBSD__) && \
defined(PROC_SPROTECT) && \
(defined(PPROT_CLEAR) || \
defined(PPORT_SET))
void stress_set_oom_adjustment(stress_args_t *args, const bool killable)
{
bool make_killable = killable;
int flag;
if (g_opt_flags & OPT_FLAGS_NO_OOM_ADJUST)
return;
/*
* main cannot be killable; if OPT_FLAGS_OOMABLE set make
* all child procs easily OOMable
*/
if (args && (g_opt_flags & OPT_FLAGS_OOMABLE))
make_killable = true;
flag = make_killable ? PPROT_CLEAR : PPROT_SET;
(void)procctl(P_PID, 0, PROC_SPROTECT, &flag);
}
bool PURE stress_process_oomed(const pid_t pid)
{
(void)pid;
return false;
}
#else
void stress_set_oom_adjustment(stress_args_t *args, const bool killable)
{
(void)args;
(void)killable;
}
bool PURE stress_process_oomed(const pid_t pid)
{
(void)pid;
return false;
}
#endif
/*
* stress_oomable_child()
* generic way to run a process that is possibly going to be
* OOM'd and we retry if it gets killed.
*/
int stress_oomable_child(
stress_args_t *args,
void *context,
stress_oomable_child_func_t func,
const int flag)
{
pid_t pid;
int ooms = 0;
int segvs = 0;
int buserrs = 0;
int rc = EXIT_SUCCESS;
size_t signal_idx = 0;
const bool not_quiet = !(flag & STRESS_OOMABLE_QUIET);
const bool valid_timeout = (g_opt_timeout > 0);
/*
* Kill child multiple times, start with SIGALRM and work up
*/
static const int signals[] = {
SIGALRM,
SIGALRM,
SIGALRM,
SIGALRM,
SIGTERM,
SIGKILL
};
again:
if (!stress_continue(args))
return EXIT_SUCCESS;
if (valid_timeout && (stress_time_now() > args->time_end)) {
return EXIT_SUCCESS;
}
pid = fork();
if (pid < 0) {
/* Keep trying if we are out of resources */
if ((errno == EAGAIN) || (errno == ENOMEM)) {
/* don't retry for 1/10th sec */
(void)shim_usleep(100000);
goto again;
}
if (not_quiet)
pr_err("%s: fork failed: errno=%d: (%s)\n",
args->name, errno, strerror(errno));
return -1;
} else if (pid > 0) {
/* Parent, wait for child */
int status, ret;
double t_end = stress_time_now() + WAIT_TIMEOUT;
args->stats->s_pid.oomable_child = pid;
rewait:
stress_set_proc_state(args->name, STRESS_STATE_WAIT);
ret = waitpid(pid, &status, 0);
if (ret < 0) {
stress_set_proc_state(args->name, STRESS_STATE_RUN);
/* No longer alive? */
if (errno == ECHILD)
goto report;
if ((errno != EINTR) && not_quiet)
pr_dbg("%s: waitpid(): errno=%d (%s)\n",
args->name, errno, strerror(errno));
(void)stress_kill_sig(pid, signals[signal_idx]);
if (signal_idx < SIZEOF_ARRAY(signals))
signal_idx++;
else if (stress_time_now() > t_end) {
pr_warn("cannot terminate process %ju, gave up after %d seconds\n", (intmax_t)pid, WAIT_TIMEOUT);
goto report;
}
/*
* First time round do fast re-wait
* in case child can be reaped quickly,
* there after do slow backoff on each
* iteration until we give up and do
* the final SIGKILL
*/
if (signal_idx > 1)
(void)shim_usleep(500000);
goto rewait;
} else if (WIFSIGNALED(status)) {
stress_set_proc_state(args->name, STRESS_STATE_RUN);
if (not_quiet)
pr_dbg("%s: child died: %s (instance %d)\n",
args->name, stress_strsignal(WTERMSIG(status)),
args->instance);
/* Bus error death? retry */
if (WTERMSIG(status) == SIGBUS) {
buserrs++;
goto again;
}
/* If we got killed by OOM killer, re-start */
if ((signals[signal_idx] != SIGKILL) && (WTERMSIG(status) == SIGKILL)) {
/*
* The --oomable flag was enabled, so
* the behaviour here is to no longer
* retry. The exit return is EXIT_SUCCESS
* because the child is allowed to terminate
* by being OOM'd.
*/
if (g_opt_flags & OPT_FLAGS_OOMABLE) {
stress_log_system_mem_info();
if (not_quiet)
pr_dbg("%s: assuming killed by OOM "
"killer, bailing out "
"(instance %d)\n",
args->name, args->instance);
stress_clean_dir(args->name, args->pid, args->instance);
return EXIT_SUCCESS;
} else {
stress_log_system_mem_info();
if (not_quiet)
pr_dbg("%s: assuming killed by OOM "
"killer, restarting again "
"(instance %d)\n",
args->name, args->instance);
ooms++;
goto again;
}
}
/* If we got killed by sigsegv, re-start */
if (WTERMSIG(status) == SIGSEGV) {
if (not_quiet)
pr_dbg("%s: killed by SIGSEGV, "
"restarting again "
"(instance %d)\n",
args->name, args->instance);
segvs++;
goto again;
}
}
rc = WEXITSTATUS(status);
} else {
/* Child */
int ret;
if (!stress_continue(args)) {
stress_set_proc_state(args->name, STRESS_STATE_EXIT);
_exit(EXIT_SUCCESS);
}
stress_parent_died_alarm();
/* Make sure this is killable by OOM killer */
stress_set_oom_adjustment(args, true);
/* Explicitly drop capabilities, makes it more OOM-able */
if (flag & STRESS_OOMABLE_DROP_CAP) {
VOID_RET(int, stress_drop_capabilities(args->name));
}
/*
* Process may have exceeded run time by the time it was
* fully runnable, so check for this before doing expensive
* stressor invocation
*/
if (!stress_continue(args) ||
(valid_timeout && (stress_time_now() > args->time_end))) {
stress_set_proc_state(args->name, STRESS_STATE_EXIT);
_exit(EXIT_SUCCESS);
}
/* ..and finally re-start the stressor */
ret = func(args, context);
pr_fail_check(&rc);
if (rc != EXIT_SUCCESS)
ret = rc;
stress_set_proc_state(args->name, STRESS_STATE_EXIT);
_exit(ret);
}
report:
if ((ooms + segvs + buserrs > 0) && not_quiet) {
pr_dbg("%s: OOM restarts: %d"
", SIGSEGV restarts: %d"
", SIGBUS restarts: %d\n",
args->name, ooms, segvs, buserrs);
}
return rc;
}