Repository navigation
Expand file tree
/
Copy pathmnexec.c
More file actions
304 lines (263 loc) · 7.86 KB
/
Copy pathmnexec.c
File metadata and controls
304 lines (263 loc) · 7.86 KB
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
/* mnexec: execution utility for mininet
*
* Starts up programs and does things that are slow or
* difficult in Python, including:
*
* - closing all file descriptors except stdin/out/error
* - detaching from a controlling tty using setsid
* - running in network and other namespaces
* - printing out the pid of a process so we can identify it later
* - attaching to namespace(s) and cgroup
* - setting RT scheduling
*
* Partially based on public domain setsid(1)
*/
#define _GNU_SOURCE
#include <stdio.h>
#include <linux/sched.h>
#include <unistd.h>
#include <limits.h>
#include <syscall.h>
#include <fcntl.h>
#include <stdlib.h>
#include <sched.h>
#include <ctype.h>
#include <sys/mount.h>
#include <sys/wait.h>
#include <assert.h>
#include <string.h>
#include <sys/stat.h>
#if !defined(VERSION)
#define VERSION "(devel)"
#endif
void usage(char *name)
{
printf("Execution utility for Mininet\n\n"
"Usage: %s [-cdmnPpu] [-a pid] [-g group] [-r rtprio] cmd args...\n\n"
"Options:\n"
" -c: close all file descriptors except stdin/out/error\n"
" -d: detach from tty by calling setsid()\n"
" -m: run in a new mount namespace\n"
" -n: run in a new network namespace\n"
" -P: run in a new pid namespace (implies -m)\n"
" -u: run in a new UTS (ipc, hostname) namespace\n"
" -p: print ^A + pid\n"
" -a pid: attach to pid's namespaces\n"
" -g group: add to cgroup\n"
" -r rtprio: run with SCHED_RR (usually requires -g)\n"
" -v: print version\n",
name);
}
#if !defined(setns)
int setns(int fd, int nstype)
{
return syscall(__NR_setns, fd, nstype);
}
#endif
/* Validate alphanumeric path foo1/bar2/baz */
void validate(char *path)
{
char *s;
for (s=path; *s; s++) {
if (!isalnum(*s) && *s != '/') {
fprintf(stderr, "invalid path: %s\n", path);
exit(1);
}
}
}
/* Add our pid to cgroup */
void cgroup(char *gname)
{
static char path[PATH_MAX];
static char *groups[] = {
"cpu", "cpuacct", "cpuset", NULL
};
char **gptr;
pid_t pid = getpid();
int count = 0;
validate(gname);
for (gptr = groups; *gptr; gptr++) {
FILE *f;
snprintf(path, PATH_MAX, "/sys/fs/cgroup/%s/%s/tasks",
*gptr, gname);
f = fopen(path, "w");
if (f) {
count++;
fprintf(f, "%d\n", pid);
fclose(f);
}
}
if (!count) {
fprintf(stderr, "cgroup: could not add to cgroup %s\n",
gname);
exit(1);
}
}
/* Attach to ns 'name' if present; return 1 if pidns */
int attachns( pid_t pid, char *name ) {
char path[PATH_MAX];
int nsid, err;
int pidns = 0;
snprintf(path, PATH_MAX, "/proc/%d/ns/%s", pid, name) ;
if ((nsid = open(path, O_RDONLY)) < 0)
return nsid;
if (strcmp(name, "pid") == 0) {
struct stat buf1, buf2;
int stat1 = stat(path, &buf1);
int stat2 = stat("/proc/self/ns/pid", &buf2);
/* Don't reattach to the same pid ns */
if (stat1 == 0 && stat2 == 0 &&
buf1.st_dev == buf2.st_dev &&
buf1.st_ino == buf2.st_ino)
return 0;
pidns = 1;
}
if ((err = setns(nsid, 0)) < 0) {
perror("setns");
return err;
}
return pidns;
}
/* Attach to pid's namespaces - returns 1 if pidns */
int attach(int pid) {
char *cwd = get_current_dir_name();
char path[PATH_MAX];
int pidns = 0;
attachns(pid, "net");
attachns(pid, "uts");
if (attachns(pid, "pid") == 1)
pidns = 1;
if (attachns(pid, "mnt") != 0) {
/* Plan B: chroot into pid's root file system */
sprintf(path, "/proc/%d/root", pid);
if (chroot(path) < 0) {
perror(path);
}
}
/* chdir to correct working directory */
if (chdir(cwd) != 0)
perror(cwd);
return pidns;
}
int main(int argc, char *argv[])
{
int c;
/* Argument flags */
int flags = 0;
int closefds = 0;
int attachpid = 0;
char *cgrouparg = NULL;
int detachtty = 0;
int printpid = 0;
int rtprio = 0;
int pidns = 0;
int dofork = 0;
while ((c = getopt(argc, argv, "+cdmnPpa:g:r:uvh")) != -1)
switch(c) {
case 'c': closefds = 1; break;
case 'd': detachtty = 1; break;
case 'm': flags |= CLONE_NEWNS; break;
case 'n': flags |= CLONE_NEWNET; break;
case 'p': printpid = 1; break;
case 'P': flags |= CLONE_NEWPID | CLONE_NEWNS; break;
case 'a': attachpid = atoi(optarg);break;
case 'g': cgrouparg = optarg ; break;
case 'r': rtprio = atoi(optarg); break;
case 'u': flags |= CLONE_NEWUTS; break;
case 'v': printf("%s\n", VERSION); exit(0);
case 'h': usage(argv[0]); exit(0);
default: usage(argv[0]); exit(1);
}
if (closefds) {
/* close file descriptors except stdin/out/error */
int fd;
for (fd = getdtablesize(); fd > 2; fd--) close(fd);
}
if (attachpid)
/* Attach to existing namespace(s) */
pidns = attach(attachpid);
else {
/* Create new namespace(s) */
if (unshare(flags) == -1) {
perror("unshare");
return 1;
}
}
if (flags & CLONE_NEWPID || pidns)
/* pidns requires fork/wait; child will be pid 1 */
dofork = 1;
if (detachtty && getpgrp() == getpid())
/* Fork so that we will no longer be pgroup leader */
dofork = 1;
else
/* We don't need a new session, only a new pgroup */
detachtty = 0;
if (detachtty)
/* Create a new session - and by implication a new process group */
setsid();
else
/* Use a new process group (in the current session)
* so Mininet can use killpg without unintended effects */
setpgid(0, 0);
if (dofork) {
/* Fork and then wait if necessary */
pid_t pid = fork();
switch(pid) {
int status;
case -1:
perror("fork");
return 1;
case 0:
/* Child continues below */
break;
default:
/* We print the *child pid* in *parent's pidns* if needed */
if (printpid) {
printf("\001%d\n", pid);
fflush(stdout);
}
/* For pid namespace, we need to fork and wait for child ;-( */
if (flags & CLONE_NEWPID || pidns) {
wait(&status);
}
return 0;
}
}
if (printpid && !dofork) {
/* Print child pid if we didn't fork/aren't in a pidns */
assert(!pidns);
printf("\001%d\n", getpid());
fflush(stdout);
}
if (flags & CLONE_NEWPID) {
/* Child remounts /proc for ps */
if (mount("proc", "/proc", "proc", MS_MGC_VAL, NULL) == -1) {
perror("mountproc");
}
}
/* Attach to cgroup if necessary */
if (cgrouparg) cgroup(cgrouparg);
if (flags & CLONE_NEWNET & CLONE_NEWNS) {
/* Mount sysfs to pick up the new network namespace */
if (mount("sysfs", "/sys", "sysfs", MS_MGC_VAL, NULL) == -1) {
perror("mount");
return 1;
}
}
if (rtprio != 0) {
/* Set RT scheduling priority */
static struct sched_param sp;
sp.sched_priority = atoi(optarg);
if (sched_setscheduler(getpid(), SCHED_RR, &sp) < 0) {
perror("sched_setscheduler");
return 1;
}
}
if (optind < argc) {
execvp(argv[optind], &argv[optind]);
perror(argv[optind]);
return 1;
}
usage(argv[0]);
return 0;
}