Commit cd831bd7874916c4eb3f0f47218e4a65c46bb905
1 parent
f8d39c01
Merge NBD client/server, by Laurent Vivier.
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@4834 c046a42c-6fe2-441c-8c8c-71466251a162
Showing
4 changed files
with
213 additions
and
11 deletions
nbd.c
@@ -25,6 +25,7 @@ | @@ -25,6 +25,7 @@ | ||
25 | #include <ctype.h> | 25 | #include <ctype.h> |
26 | #include <inttypes.h> | 26 | #include <inttypes.h> |
27 | #include <sys/socket.h> | 27 | #include <sys/socket.h> |
28 | +#include <sys/un.h> | ||
28 | #include <netinet/in.h> | 29 | #include <netinet/in.h> |
29 | #include <netinet/tcp.h> | 30 | #include <netinet/tcp.h> |
30 | #include <arpa/inet.h> | 31 | #include <arpa/inet.h> |
@@ -183,6 +184,65 @@ error: | @@ -183,6 +184,65 @@ error: | ||
183 | return -1; | 184 | return -1; |
184 | } | 185 | } |
185 | 186 | ||
187 | +int unix_socket_incoming(const char *path) | ||
188 | +{ | ||
189 | + int s; | ||
190 | + struct sockaddr_un addr; | ||
191 | + int serrno; | ||
192 | + | ||
193 | + s = socket(PF_UNIX, SOCK_STREAM, 0); | ||
194 | + if (s == -1) { | ||
195 | + return -1; | ||
196 | + } | ||
197 | + | ||
198 | + memset(&addr, 0, sizeof(addr)); | ||
199 | + addr.sun_family = AF_UNIX; | ||
200 | + pstrcpy(addr.sun_path, sizeof(addr.sun_path), path); | ||
201 | + | ||
202 | + if (bind(s, (struct sockaddr *)&addr, sizeof(addr)) == -1) { | ||
203 | + goto error; | ||
204 | + } | ||
205 | + | ||
206 | + if (listen(s, 128) == -1) { | ||
207 | + goto error; | ||
208 | + } | ||
209 | + | ||
210 | + return s; | ||
211 | +error: | ||
212 | + serrno = errno; | ||
213 | + close(s); | ||
214 | + errno = serrno; | ||
215 | + return -1; | ||
216 | +} | ||
217 | + | ||
218 | +int unix_socket_outgoing(const char *path) | ||
219 | +{ | ||
220 | + int s; | ||
221 | + struct sockaddr_un addr; | ||
222 | + int serrno; | ||
223 | + | ||
224 | + s = socket(PF_UNIX, SOCK_STREAM, 0); | ||
225 | + if (s == -1) { | ||
226 | + return -1; | ||
227 | + } | ||
228 | + | ||
229 | + memset(&addr, 0, sizeof(addr)); | ||
230 | + addr.sun_family = AF_UNIX; | ||
231 | + pstrcpy(addr.sun_path, sizeof(addr.sun_path), path); | ||
232 | + | ||
233 | + if (connect(s, (struct sockaddr *)&addr, sizeof(addr)) == -1) { | ||
234 | + goto error; | ||
235 | + } | ||
236 | + | ||
237 | + return s; | ||
238 | +error: | ||
239 | + serrno = errno; | ||
240 | + close(s); | ||
241 | + errno = serrno; | ||
242 | + return -1; | ||
243 | +} | ||
244 | + | ||
245 | + | ||
186 | /* Basic flow | 246 | /* Basic flow |
187 | 247 | ||
188 | Server Client | 248 | Server Client |
@@ -225,12 +285,10 @@ int nbd_negotiate(BlockDriverState *bs, int csock, off_t size) | @@ -225,12 +285,10 @@ int nbd_negotiate(BlockDriverState *bs, int csock, off_t size) | ||
225 | return 0; | 285 | return 0; |
226 | } | 286 | } |
227 | 287 | ||
228 | -int nbd_receive_negotiate(int fd, int csock) | 288 | +int nbd_receive_negotiate(int csock, off_t *size, size_t *blocksize) |
229 | { | 289 | { |
230 | char buf[8 + 8 + 8 + 128]; | 290 | char buf[8 + 8 + 8 + 128]; |
231 | uint64_t magic; | 291 | uint64_t magic; |
232 | - off_t size; | ||
233 | - size_t blocksize; | ||
234 | 292 | ||
235 | TRACE("Receiving negotation."); | 293 | TRACE("Receiving negotation."); |
236 | 294 | ||
@@ -241,8 +299,8 @@ int nbd_receive_negotiate(int fd, int csock) | @@ -241,8 +299,8 @@ int nbd_receive_negotiate(int fd, int csock) | ||
241 | } | 299 | } |
242 | 300 | ||
243 | magic = be64_to_cpup((uint64_t*)(buf + 8)); | 301 | magic = be64_to_cpup((uint64_t*)(buf + 8)); |
244 | - size = be64_to_cpup((uint64_t*)(buf + 16)); | ||
245 | - blocksize = 1024; | 302 | + *size = be64_to_cpup((uint64_t*)(buf + 16)); |
303 | + *blocksize = 1024; | ||
246 | 304 | ||
247 | TRACE("Magic is %c%c%c%c%c%c%c%c", | 305 | TRACE("Magic is %c%c%c%c%c%c%c%c", |
248 | isprint(buf[0]) ? buf[0] : '.', | 306 | isprint(buf[0]) ? buf[0] : '.', |
@@ -254,7 +312,7 @@ int nbd_receive_negotiate(int fd, int csock) | @@ -254,7 +312,7 @@ int nbd_receive_negotiate(int fd, int csock) | ||
254 | isprint(buf[6]) ? buf[6] : '.', | 312 | isprint(buf[6]) ? buf[6] : '.', |
255 | isprint(buf[7]) ? buf[7] : '.'); | 313 | isprint(buf[7]) ? buf[7] : '.'); |
256 | TRACE("Magic is 0x%" PRIx64, magic); | 314 | TRACE("Magic is 0x%" PRIx64, magic); |
257 | - TRACE("Size is %" PRIu64, size); | 315 | + TRACE("Size is %" PRIu64, *size); |
258 | 316 | ||
259 | if (memcmp(buf, "NBDMAGIC", 8) != 0) { | 317 | if (memcmp(buf, "NBDMAGIC", 8) != 0) { |
260 | LOG("Invalid magic received"); | 318 | LOG("Invalid magic received"); |
@@ -269,7 +327,11 @@ int nbd_receive_negotiate(int fd, int csock) | @@ -269,7 +327,11 @@ int nbd_receive_negotiate(int fd, int csock) | ||
269 | errno = EINVAL; | 327 | errno = EINVAL; |
270 | return -1; | 328 | return -1; |
271 | } | 329 | } |
330 | + return 0; | ||
331 | +} | ||
272 | 332 | ||
333 | +int nbd_init(int fd, int csock, off_t size, size_t blocksize) | ||
334 | +{ | ||
273 | TRACE("Setting block size to %lu", (unsigned long)blocksize); | 335 | TRACE("Setting block size to %lu", (unsigned long)blocksize); |
274 | 336 | ||
275 | if (ioctl(fd, NBD_SET_BLKSIZE, blocksize) == -1) { | 337 | if (ioctl(fd, NBD_SET_BLKSIZE, blocksize) == -1) { |
nbd.h
@@ -27,9 +27,12 @@ | @@ -27,9 +27,12 @@ | ||
27 | #include "block_int.h" | 27 | #include "block_int.h" |
28 | 28 | ||
29 | int tcp_socket_incoming(const char *address, uint16_t port); | 29 | int tcp_socket_incoming(const char *address, uint16_t port); |
30 | +int unix_socket_outgoing(const char *path); | ||
31 | +int unix_socket_incoming(const char *path); | ||
30 | 32 | ||
31 | int nbd_negotiate(BlockDriverState *bs, int csock, off_t size); | 33 | int nbd_negotiate(BlockDriverState *bs, int csock, off_t size); |
32 | -int nbd_receive_negotiate(int fd, int csock); | 34 | +int nbd_receive_negotiate(int csock, off_t *size, size_t *blocksize); |
35 | +int nbd_init(int fd, int csock, off_t size, size_t blocksize); | ||
33 | int nbd_trip(BlockDriverState *bs, int csock, off_t size, uint64_t dev_offset, off_t *offset, bool readonly); | 36 | int nbd_trip(BlockDriverState *bs, int csock, off_t size, uint64_t dev_offset, off_t *offset, bool readonly); |
34 | int nbd_client(int fd, int csock); | 37 | int nbd_client(int fd, int csock); |
35 | int nbd_disconnect(int fd); | 38 | int nbd_disconnect(int fd); |
qemu-nbd.c
1 | -/*\ | 1 | +/* |
2 | * Copyright (C) 2005 Anthony Liguori <anthony@codemonkey.ws> | 2 | * Copyright (C) 2005 Anthony Liguori <anthony@codemonkey.ws> |
3 | * | 3 | * |
4 | * Network Block Device | 4 | * Network Block Device |
@@ -25,10 +25,14 @@ | @@ -25,10 +25,14 @@ | ||
25 | #include <stdio.h> | 25 | #include <stdio.h> |
26 | #include <getopt.h> | 26 | #include <getopt.h> |
27 | #include <err.h> | 27 | #include <err.h> |
28 | +#include <sys/types.h> | ||
28 | #include <sys/socket.h> | 29 | #include <sys/socket.h> |
29 | #include <netinet/in.h> | 30 | #include <netinet/in.h> |
30 | #include <netinet/tcp.h> | 31 | #include <netinet/tcp.h> |
31 | #include <arpa/inet.h> | 32 | #include <arpa/inet.h> |
33 | +#include <signal.h> | ||
34 | + | ||
35 | +#define SOCKET_PATH "/var/lock/qemu-nbd-%s" | ||
32 | 36 | ||
33 | int verbose; | 37 | int verbose; |
34 | 38 | ||
@@ -41,14 +45,18 @@ static void usage(const char *name) | @@ -41,14 +45,18 @@ static void usage(const char *name) | ||
41 | " -p, --port=PORT port to listen on (default `1024')\n" | 45 | " -p, --port=PORT port to listen on (default `1024')\n" |
42 | " -o, --offset=OFFSET offset into the image\n" | 46 | " -o, --offset=OFFSET offset into the image\n" |
43 | " -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n" | 47 | " -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n" |
48 | +" -k, --socket=PATH path to the unix socket\n" | ||
49 | +" (default '"SOCKET_PATH"')\n" | ||
44 | " -r, --read-only export read-only\n" | 50 | " -r, --read-only export read-only\n" |
45 | " -P, --partition=NUM only expose partition NUM\n" | 51 | " -P, --partition=NUM only expose partition NUM\n" |
52 | +" -c, --connect=DEV connect FILE to the local NBD device DEV\n" | ||
53 | +" -d, --disconnect disconnect the specified device\n" | ||
46 | " -v, --verbose display extra debugging information\n" | 54 | " -v, --verbose display extra debugging information\n" |
47 | " -h, --help display this help and exit\n" | 55 | " -h, --help display this help and exit\n" |
48 | " -V, --version output version information and exit\n" | 56 | " -V, --version output version information and exit\n" |
49 | "\n" | 57 | "\n" |
50 | "Report bugs to <anthony@codemonkey.ws>\n" | 58 | "Report bugs to <anthony@codemonkey.ws>\n" |
51 | - , name); | 59 | + , name, "DEVICE"); |
52 | } | 60 | } |
53 | 61 | ||
54 | static void version(const char *name) | 62 | static void version(const char *name) |
@@ -144,27 +152,51 @@ static int find_partition(BlockDriverState *bs, int partition, | @@ -144,27 +152,51 @@ static int find_partition(BlockDriverState *bs, int partition, | ||
144 | return -1; | 152 | return -1; |
145 | } | 153 | } |
146 | 154 | ||
155 | +static void show_parts(const char *device) | ||
156 | +{ | ||
157 | + if (fork() == 0) { | ||
158 | + int nbd; | ||
159 | + | ||
160 | + /* linux just needs an open() to trigger | ||
161 | + * the partition table update | ||
162 | + * but remember to load the module with max_part != 0 : | ||
163 | + * modprobe nbd max_part=63 | ||
164 | + */ | ||
165 | + nbd = open(device, O_RDWR); | ||
166 | + if (nbd != -1) | ||
167 | + close(nbd); | ||
168 | + exit(0); | ||
169 | + } | ||
170 | +} | ||
171 | + | ||
147 | int main(int argc, char **argv) | 172 | int main(int argc, char **argv) |
148 | { | 173 | { |
149 | BlockDriverState *bs; | 174 | BlockDriverState *bs; |
150 | off_t dev_offset = 0; | 175 | off_t dev_offset = 0; |
151 | off_t offset = 0; | 176 | off_t offset = 0; |
152 | bool readonly = false; | 177 | bool readonly = false; |
178 | + bool disconnect = false; | ||
153 | const char *bindto = "0.0.0.0"; | 179 | const char *bindto = "0.0.0.0"; |
154 | int port = 1024; | 180 | int port = 1024; |
155 | int sock, csock; | 181 | int sock, csock; |
156 | struct sockaddr_in addr; | 182 | struct sockaddr_in addr; |
157 | socklen_t addr_len = sizeof(addr); | 183 | socklen_t addr_len = sizeof(addr); |
158 | off_t fd_size; | 184 | off_t fd_size; |
159 | - const char *sopt = "hVbo:p:rsP:v"; | 185 | + char *device = NULL; |
186 | + char *socket = NULL; | ||
187 | + char sockpath[128]; | ||
188 | + const char *sopt = "hVbo:p:rsP:c:dvk:"; | ||
160 | struct option lopt[] = { | 189 | struct option lopt[] = { |
161 | { "help", 0, 0, 'h' }, | 190 | { "help", 0, 0, 'h' }, |
162 | { "version", 0, 0, 'V' }, | 191 | { "version", 0, 0, 'V' }, |
163 | { "bind", 1, 0, 'b' }, | 192 | { "bind", 1, 0, 'b' }, |
164 | { "port", 1, 0, 'p' }, | 193 | { "port", 1, 0, 'p' }, |
194 | + { "socket", 1, 0, 'k' }, | ||
165 | { "offset", 1, 0, 'o' }, | 195 | { "offset", 1, 0, 'o' }, |
166 | { "read-only", 0, 0, 'r' }, | 196 | { "read-only", 0, 0, 'r' }, |
167 | { "partition", 1, 0, 'P' }, | 197 | { "partition", 1, 0, 'P' }, |
198 | + { "connect", 1, 0, 'c' }, | ||
199 | + { "disconnect", 0, 0, 'd' }, | ||
168 | { "snapshot", 0, 0, 's' }, | 200 | { "snapshot", 0, 0, 's' }, |
169 | { "verbose", 0, 0, 'v' }, | 201 | { "verbose", 0, 0, 'v' }, |
170 | { NULL, 0, 0, 0 } | 202 | { NULL, 0, 0, 0 } |
@@ -175,6 +207,8 @@ int main(int argc, char **argv) | @@ -175,6 +207,8 @@ int main(int argc, char **argv) | ||
175 | char *end; | 207 | char *end; |
176 | bool snapshot = false; | 208 | bool snapshot = false; |
177 | int partition = -1; | 209 | int partition = -1; |
210 | + int fd; | ||
211 | + int ret; | ||
178 | 212 | ||
179 | while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) { | 213 | while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) { |
180 | switch (ch) { | 214 | switch (ch) { |
@@ -213,6 +247,17 @@ int main(int argc, char **argv) | @@ -213,6 +247,17 @@ int main(int argc, char **argv) | ||
213 | if (partition < 1 || partition > 8) | 247 | if (partition < 1 || partition > 8) |
214 | errx(EINVAL, "Invalid partition %d", partition); | 248 | errx(EINVAL, "Invalid partition %d", partition); |
215 | break; | 249 | break; |
250 | + case 'k': | ||
251 | + socket = optarg; | ||
252 | + if (socket[0] != '/') | ||
253 | + errx(EINVAL, "socket path must be absolute\n"); | ||
254 | + break; | ||
255 | + case 'd': | ||
256 | + disconnect = true; | ||
257 | + break; | ||
258 | + case 'c': | ||
259 | + device = optarg; | ||
260 | + break; | ||
216 | case 'v': | 261 | case 'v': |
217 | verbose = 1; | 262 | verbose = 1; |
218 | break; | 263 | break; |
@@ -236,6 +281,20 @@ int main(int argc, char **argv) | @@ -236,6 +281,20 @@ int main(int argc, char **argv) | ||
236 | argv[0]); | 281 | argv[0]); |
237 | } | 282 | } |
238 | 283 | ||
284 | + if (disconnect) { | ||
285 | + fd = open(argv[optind], O_RDWR); | ||
286 | + if (fd == -1) | ||
287 | + errx(errno, "Cannot open %s", argv[optind]); | ||
288 | + | ||
289 | + nbd_disconnect(fd); | ||
290 | + | ||
291 | + close(fd); | ||
292 | + | ||
293 | + printf("%s disconnected\n", argv[optind]); | ||
294 | + | ||
295 | + return 0; | ||
296 | + } | ||
297 | + | ||
239 | bdrv_init(); | 298 | bdrv_init(); |
240 | 299 | ||
241 | bs = bdrv_new("hda"); | 300 | bs = bdrv_new("hda"); |
@@ -251,7 +310,77 @@ int main(int argc, char **argv) | @@ -251,7 +310,77 @@ int main(int argc, char **argv) | ||
251 | find_partition(bs, partition, &dev_offset, &fd_size)) | 310 | find_partition(bs, partition, &dev_offset, &fd_size)) |
252 | errx(errno, "Could not find partition %d", partition); | 311 | errx(errno, "Could not find partition %d", partition); |
253 | 312 | ||
254 | - sock = tcp_socket_incoming(bindto, port); | 313 | + if (device) { |
314 | + pid_t pid; | ||
315 | + if (!verbose) | ||
316 | + daemon(0, 0); /* detach client and server */ | ||
317 | + | ||
318 | + if (socket == NULL) { | ||
319 | + sprintf(sockpath, SOCKET_PATH, basename(device)); | ||
320 | + socket = sockpath; | ||
321 | + } | ||
322 | + | ||
323 | + pid = fork(); | ||
324 | + if (pid < 0) | ||
325 | + return 1; | ||
326 | + if (pid != 0) { | ||
327 | + off_t size; | ||
328 | + size_t blocksize; | ||
329 | + | ||
330 | + ret = 0; | ||
331 | + bdrv_close(bs); | ||
332 | + | ||
333 | + do { | ||
334 | + sock = unix_socket_outgoing(socket); | ||
335 | + if (sock == -1) { | ||
336 | + if (errno != ENOENT && errno != ECONNREFUSED) | ||
337 | + goto out; | ||
338 | + sleep(1); /* wait children */ | ||
339 | + } | ||
340 | + } while (sock == -1); | ||
341 | + | ||
342 | + fd = open(device, O_RDWR); | ||
343 | + if (fd == -1) { | ||
344 | + ret = 1; | ||
345 | + goto out; | ||
346 | + } | ||
347 | + | ||
348 | + ret = nbd_receive_negotiate(sock, &size, &blocksize); | ||
349 | + if (ret == -1) { | ||
350 | + ret = 1; | ||
351 | + goto out; | ||
352 | + } | ||
353 | + | ||
354 | + ret = nbd_init(fd, sock, size, blocksize); | ||
355 | + if (ret == -1) { | ||
356 | + ret = 1; | ||
357 | + goto out; | ||
358 | + } | ||
359 | + | ||
360 | + printf("NBD device %s is now connected to file %s\n", | ||
361 | + device, argv[optind]); | ||
362 | + | ||
363 | + /* update partition table */ | ||
364 | + | ||
365 | + show_parts(device); | ||
366 | + | ||
367 | + nbd_client(fd, sock); | ||
368 | + close(fd); | ||
369 | + out: | ||
370 | + kill(pid, SIGTERM); | ||
371 | + unlink(socket); | ||
372 | + | ||
373 | + return ret; | ||
374 | + } | ||
375 | + /* children */ | ||
376 | + } | ||
377 | + | ||
378 | + if (socket) { | ||
379 | + sock = unix_socket_incoming(socket); | ||
380 | + } else { | ||
381 | + sock = tcp_socket_incoming(bindto, port); | ||
382 | + } | ||
383 | + | ||
255 | if (sock == -1) | 384 | if (sock == -1) |
256 | return 1; | 385 | return 1; |
257 | 386 | ||
@@ -270,6 +399,8 @@ int main(int argc, char **argv) | @@ -270,6 +399,8 @@ int main(int argc, char **argv) | ||
270 | close(csock); | 399 | close(csock); |
271 | close(sock); | 400 | close(sock); |
272 | bdrv_close(bs); | 401 | bdrv_close(bs); |
402 | + if (socket) | ||
403 | + unlink(socket); | ||
273 | 404 | ||
274 | return 0; | 405 | return 0; |
275 | } | 406 | } |
qemu-nbd.texi
@@ -20,10 +20,16 @@ Export Qemu disk image using NBD protocol. | @@ -20,10 +20,16 @@ Export Qemu disk image using NBD protocol. | ||
20 | offset into the image | 20 | offset into the image |
21 | @item -b, --bind=IFACE | 21 | @item -b, --bind=IFACE |
22 | interface to bind to (default `0.0.0.0') | 22 | interface to bind to (default `0.0.0.0') |
23 | +@item -k, --socket=PATH | ||
24 | + Use a unix socket with path PATH | ||
23 | @item -r, --read-only | 25 | @item -r, --read-only |
24 | export read-only | 26 | export read-only |
25 | @item -P, --partition=NUM | 27 | @item -P, --partition=NUM |
26 | only expose partition NUM | 28 | only expose partition NUM |
29 | +@item -c, --connect | ||
30 | + connect FILE to NBD device DEV | ||
31 | +@item -d, --disconnect | ||
32 | + disconnect the specified device | ||
27 | @item -v, --verbose | 33 | @item -v, --verbose |
28 | display extra debugging information | 34 | display extra debugging information |
29 | @item -h, --help | 35 | @item -h, --help |