]> git.proxmox.com Git - qemu.git/blob - qemu-bridge-helper.c
Merge remote-tracking branch 'aneesh/for-upstream' into staging
[qemu.git] / qemu-bridge-helper.c
1 /*
2 * QEMU Bridge Helper
3 *
4 * Copyright IBM, Corp. 2011
5 *
6 * Authors:
7 * Anthony Liguori <aliguori@us.ibm.com>
8 * Richa Marwaha <rmarwah@linux.vnet.ibm.com>
9 * Corey Bryant <coreyb@linux.vnet.ibm.com>
10 *
11 * This work is licensed under the terms of the GNU GPL, version 2. See
12 * the COPYING file in the top-level directory.
13 *
14 */
15
16 #include "config-host.h"
17
18 #include <stdio.h>
19 #include <errno.h>
20 #include <fcntl.h>
21 #include <unistd.h>
22 #include <string.h>
23 #include <stdlib.h>
24 #include <stdbool.h>
25 #include <ctype.h>
26 #include <glib.h>
27
28 #include <sys/types.h>
29 #include <sys/ioctl.h>
30 #include <sys/socket.h>
31 #include <sys/un.h>
32 #include <sys/prctl.h>
33
34 #include <net/if.h>
35
36 #include <linux/sockios.h>
37
38 #include "qemu-queue.h"
39
40 #include "net/tap-linux.h"
41
42 #ifdef CONFIG_LIBCAP
43 #include <cap-ng.h>
44 #endif
45
46 #define DEFAULT_ACL_FILE CONFIG_QEMU_CONFDIR "/bridge.conf"
47
48 enum {
49 ACL_ALLOW = 0,
50 ACL_ALLOW_ALL,
51 ACL_DENY,
52 ACL_DENY_ALL,
53 };
54
55 typedef struct ACLRule {
56 int type;
57 char iface[IFNAMSIZ];
58 QSIMPLEQ_ENTRY(ACLRule) entry;
59 } ACLRule;
60
61 typedef QSIMPLEQ_HEAD(ACLList, ACLRule) ACLList;
62
63 static void usage(void)
64 {
65 fprintf(stderr,
66 "Usage: qemu-bridge-helper [--use-vnet] --br=bridge --fd=unixfd\n");
67 }
68
69 static int parse_acl_file(const char *filename, ACLList *acl_list)
70 {
71 FILE *f;
72 char line[4096];
73 ACLRule *acl_rule;
74
75 f = fopen(filename, "r");
76 if (f == NULL) {
77 return -1;
78 }
79
80 while (fgets(line, sizeof(line), f) != NULL) {
81 char *ptr = line;
82 char *cmd, *arg, *argend;
83
84 while (isspace(*ptr)) {
85 ptr++;
86 }
87
88 /* skip comments and empty lines */
89 if (*ptr == '#' || *ptr == 0) {
90 continue;
91 }
92
93 cmd = ptr;
94 arg = strchr(cmd, ' ');
95 if (arg == NULL) {
96 arg = strchr(cmd, '\t');
97 }
98
99 if (arg == NULL) {
100 fprintf(stderr, "Invalid config line:\n %s\n", line);
101 fclose(f);
102 errno = EINVAL;
103 return -1;
104 }
105
106 *arg = 0;
107 arg++;
108 while (isspace(*arg)) {
109 arg++;
110 }
111
112 argend = arg + strlen(arg);
113 while (arg != argend && isspace(*(argend - 1))) {
114 argend--;
115 }
116 *argend = 0;
117
118 if (strcmp(cmd, "deny") == 0) {
119 acl_rule = g_malloc(sizeof(*acl_rule));
120 if (strcmp(arg, "all") == 0) {
121 acl_rule->type = ACL_DENY_ALL;
122 } else {
123 acl_rule->type = ACL_DENY;
124 snprintf(acl_rule->iface, IFNAMSIZ, "%s", arg);
125 }
126 QSIMPLEQ_INSERT_TAIL(acl_list, acl_rule, entry);
127 } else if (strcmp(cmd, "allow") == 0) {
128 acl_rule = g_malloc(sizeof(*acl_rule));
129 if (strcmp(arg, "all") == 0) {
130 acl_rule->type = ACL_ALLOW_ALL;
131 } else {
132 acl_rule->type = ACL_ALLOW;
133 snprintf(acl_rule->iface, IFNAMSIZ, "%s", arg);
134 }
135 QSIMPLEQ_INSERT_TAIL(acl_list, acl_rule, entry);
136 } else if (strcmp(cmd, "include") == 0) {
137 /* ignore errors */
138 parse_acl_file(arg, acl_list);
139 } else {
140 fprintf(stderr, "Unknown command `%s'\n", cmd);
141 fclose(f);
142 errno = EINVAL;
143 return -1;
144 }
145 }
146
147 fclose(f);
148
149 return 0;
150 }
151
152 static bool has_vnet_hdr(int fd)
153 {
154 unsigned int features = 0;
155
156 if (ioctl(fd, TUNGETFEATURES, &features) == -1) {
157 return false;
158 }
159
160 if (!(features & IFF_VNET_HDR)) {
161 return false;
162 }
163
164 return true;
165 }
166
167 static void prep_ifreq(struct ifreq *ifr, const char *ifname)
168 {
169 memset(ifr, 0, sizeof(*ifr));
170 snprintf(ifr->ifr_name, IFNAMSIZ, "%s", ifname);
171 }
172
173 static int send_fd(int c, int fd)
174 {
175 char msgbuf[CMSG_SPACE(sizeof(fd))];
176 struct msghdr msg = {
177 .msg_control = msgbuf,
178 .msg_controllen = sizeof(msgbuf),
179 };
180 struct cmsghdr *cmsg;
181 struct iovec iov;
182 char req[1] = { 0x00 };
183
184 cmsg = CMSG_FIRSTHDR(&msg);
185 cmsg->cmsg_level = SOL_SOCKET;
186 cmsg->cmsg_type = SCM_RIGHTS;
187 cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
188 msg.msg_controllen = cmsg->cmsg_len;
189
190 iov.iov_base = req;
191 iov.iov_len = sizeof(req);
192
193 msg.msg_iov = &iov;
194 msg.msg_iovlen = 1;
195 memcpy(CMSG_DATA(cmsg), &fd, sizeof(fd));
196
197 return sendmsg(c, &msg, 0);
198 }
199
200 #ifdef CONFIG_LIBCAP
201 static int drop_privileges(void)
202 {
203 /* clear all capabilities */
204 capng_clear(CAPNG_SELECT_BOTH);
205
206 if (capng_update(CAPNG_ADD, CAPNG_EFFECTIVE | CAPNG_PERMITTED,
207 CAP_NET_ADMIN) < 0) {
208 return -1;
209 }
210
211 /* change to calling user's real uid and gid, retaining supplemental
212 * groups and CAP_NET_ADMIN */
213 if (capng_change_id(getuid(), getgid(), CAPNG_CLEAR_BOUNDING)) {
214 return -1;
215 }
216
217 return 0;
218 }
219 #endif
220
221 int main(int argc, char **argv)
222 {
223 struct ifreq ifr;
224 int fd, ctlfd, unixfd = -1;
225 int use_vnet = 0;
226 int mtu;
227 const char *bridge = NULL;
228 char iface[IFNAMSIZ];
229 int index;
230 ACLRule *acl_rule;
231 ACLList acl_list;
232 int access_allowed, access_denied;
233 int ret = EXIT_SUCCESS;
234
235 #ifdef CONFIG_LIBCAP
236 /* if we're run from an suid binary, immediately drop privileges preserving
237 * cap_net_admin */
238 if (geteuid() == 0 && getuid() != geteuid()) {
239 if (drop_privileges() == -1) {
240 fprintf(stderr, "failed to drop privileges\n");
241 return 1;
242 }
243 }
244 #endif
245
246 /* parse arguments */
247 for (index = 1; index < argc; index++) {
248 if (strcmp(argv[index], "--use-vnet") == 0) {
249 use_vnet = 1;
250 } else if (strncmp(argv[index], "--br=", 5) == 0) {
251 bridge = &argv[index][5];
252 } else if (strncmp(argv[index], "--fd=", 5) == 0) {
253 unixfd = atoi(&argv[index][5]);
254 } else {
255 usage();
256 return EXIT_FAILURE;
257 }
258 }
259
260 if (bridge == NULL || unixfd == -1) {
261 usage();
262 return EXIT_FAILURE;
263 }
264
265 /* parse default acl file */
266 QSIMPLEQ_INIT(&acl_list);
267 if (parse_acl_file(DEFAULT_ACL_FILE, &acl_list) == -1) {
268 fprintf(stderr, "failed to parse default acl file `%s'\n",
269 DEFAULT_ACL_FILE);
270 ret = EXIT_FAILURE;
271 goto cleanup;
272 }
273
274 /* validate bridge against acl -- default policy is to deny
275 * according acl policy if we have a deny and allow both
276 * then deny should always win over allow
277 */
278 access_allowed = 0;
279 access_denied = 0;
280 QSIMPLEQ_FOREACH(acl_rule, &acl_list, entry) {
281 switch (acl_rule->type) {
282 case ACL_ALLOW_ALL:
283 access_allowed = 1;
284 break;
285 case ACL_ALLOW:
286 if (strcmp(bridge, acl_rule->iface) == 0) {
287 access_allowed = 1;
288 }
289 break;
290 case ACL_DENY_ALL:
291 access_denied = 1;
292 break;
293 case ACL_DENY:
294 if (strcmp(bridge, acl_rule->iface) == 0) {
295 access_denied = 1;
296 }
297 break;
298 }
299 }
300
301 if ((access_allowed == 0) || (access_denied == 1)) {
302 fprintf(stderr, "access denied by acl file\n");
303 ret = EXIT_FAILURE;
304 goto cleanup;
305 }
306
307 /* open a socket to use to control the network interfaces */
308 ctlfd = socket(AF_INET, SOCK_STREAM, 0);
309 if (ctlfd == -1) {
310 fprintf(stderr, "failed to open control socket: %s\n", strerror(errno));
311 ret = EXIT_FAILURE;
312 goto cleanup;
313 }
314
315 /* open the tap device */
316 fd = open("/dev/net/tun", O_RDWR);
317 if (fd == -1) {
318 fprintf(stderr, "failed to open /dev/net/tun: %s\n", strerror(errno));
319 ret = EXIT_FAILURE;
320 goto cleanup;
321 }
322
323 /* request a tap device, disable PI, and add vnet header support if
324 * requested and it's available. */
325 prep_ifreq(&ifr, "tap%d");
326 ifr.ifr_flags = IFF_TAP|IFF_NO_PI;
327 if (use_vnet && has_vnet_hdr(fd)) {
328 ifr.ifr_flags |= IFF_VNET_HDR;
329 }
330
331 if (ioctl(fd, TUNSETIFF, &ifr) == -1) {
332 fprintf(stderr, "failed to create tun device: %s\n", strerror(errno));
333 ret = EXIT_FAILURE;
334 goto cleanup;
335 }
336
337 /* save tap device name */
338 snprintf(iface, sizeof(iface), "%s", ifr.ifr_name);
339
340 /* get the mtu of the bridge */
341 prep_ifreq(&ifr, bridge);
342 if (ioctl(ctlfd, SIOCGIFMTU, &ifr) == -1) {
343 fprintf(stderr, "failed to get mtu of bridge `%s': %s\n",
344 bridge, strerror(errno));
345 ret = EXIT_FAILURE;
346 goto cleanup;
347 }
348
349 /* save mtu */
350 mtu = ifr.ifr_mtu;
351
352 /* set the mtu of the interface based on the bridge */
353 prep_ifreq(&ifr, iface);
354 ifr.ifr_mtu = mtu;
355 if (ioctl(ctlfd, SIOCSIFMTU, &ifr) == -1) {
356 fprintf(stderr, "failed to set mtu of device `%s' to %d: %s\n",
357 iface, mtu, strerror(errno));
358 ret = EXIT_FAILURE;
359 goto cleanup;
360 }
361
362 /* add the interface to the bridge */
363 prep_ifreq(&ifr, bridge);
364 ifr.ifr_ifindex = if_nametoindex(iface);
365
366 if (ioctl(ctlfd, SIOCBRADDIF, &ifr) == -1) {
367 fprintf(stderr, "failed to add interface `%s' to bridge `%s': %s\n",
368 iface, bridge, strerror(errno));
369 ret = EXIT_FAILURE;
370 goto cleanup;
371 }
372
373 /* bring the interface up */
374 prep_ifreq(&ifr, iface);
375 if (ioctl(ctlfd, SIOCGIFFLAGS, &ifr) == -1) {
376 fprintf(stderr, "failed to get interface flags for `%s': %s\n",
377 iface, strerror(errno));
378 ret = EXIT_FAILURE;
379 goto cleanup;
380 }
381
382 ifr.ifr_flags |= IFF_UP;
383 if (ioctl(ctlfd, SIOCSIFFLAGS, &ifr) == -1) {
384 fprintf(stderr, "failed to bring up interface `%s': %s\n",
385 iface, strerror(errno));
386 ret = EXIT_FAILURE;
387 goto cleanup;
388 }
389
390 /* write fd to the domain socket */
391 if (send_fd(unixfd, fd) == -1) {
392 fprintf(stderr, "failed to write fd to unix socket: %s\n",
393 strerror(errno));
394 ret = EXIT_FAILURE;
395 goto cleanup;
396 }
397
398 /* ... */
399
400 /* profit! */
401
402 cleanup:
403
404 while ((acl_rule = QSIMPLEQ_FIRST(&acl_list)) != NULL) {
405 QSIMPLEQ_REMOVE_HEAD(&acl_list, entry);
406 g_free(acl_rule);
407 }
408
409 return ret;
410 }