qcow2: Allow different cluster sizes
[qemu-kvm/fedora.git] / nbd.c
blobb397a5f48755e771ef9daf9c6f338c5a5fc00038
1 /*
2 * Copyright (C) 2005 Anthony Liguori <anthony@codemonkey.ws>
4 * Network Block Device
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; under version 2 of the License.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301 USA
20 #include "nbd.h"
22 #include <errno.h>
23 #include <string.h>
24 #ifndef _WIN32
25 #include <sys/ioctl.h>
26 #endif
27 #ifdef __sun__
28 #include <sys/ioccom.h>
29 #endif
30 #include <ctype.h>
31 #include <inttypes.h>
33 #include "qemu_socket.h"
35 //#define DEBUG_NBD
37 #ifdef DEBUG_NBD
38 #define TRACE(msg, ...) do { \
39 LOG(msg, ## __VA_ARGS__); \
40 } while(0)
41 #else
42 #define TRACE(msg, ...) \
43 do { } while (0)
44 #endif
46 #define LOG(msg, ...) do { \
47 fprintf(stderr, "%s:%s():L%d: " msg "\n", \
48 __FILE__, __FUNCTION__, __LINE__, ## __VA_ARGS__); \
49 } while(0)
51 /* This is all part of the "official" NBD API */
53 #define NBD_REQUEST_MAGIC 0x25609513
54 #define NBD_REPLY_MAGIC 0x67446698
56 #define NBD_SET_SOCK _IO(0xab, 0)
57 #define NBD_SET_BLKSIZE _IO(0xab, 1)
58 #define NBD_SET_SIZE _IO(0xab, 2)
59 #define NBD_DO_IT _IO(0xab, 3)
60 #define NBD_CLEAR_SOCK _IO(0xab, 4)
61 #define NBD_CLEAR_QUE _IO(0xab, 5)
62 #define NBD_PRINT_DEBUG _IO(0xab, 6)
63 #define NBD_SET_SIZE_BLOCKS _IO(0xab, 7)
64 #define NBD_DISCONNECT _IO(0xab, 8)
66 /* That's all folks */
68 #define read_sync(fd, buffer, size) nbd_wr_sync(fd, buffer, size, true)
69 #define write_sync(fd, buffer, size) nbd_wr_sync(fd, buffer, size, false)
71 size_t nbd_wr_sync(int fd, void *buffer, size_t size, bool do_read)
73 size_t offset = 0;
75 while (offset < size) {
76 ssize_t len;
78 if (do_read) {
79 len = recv(fd, buffer + offset, size - offset, 0);
80 } else {
81 len = send(fd, buffer + offset, size - offset, 0);
84 if (len == -1)
85 errno = socket_error();
87 /* recoverable error */
88 if (len == -1 && (errno == EAGAIN || errno == EINTR)) {
89 continue;
92 /* eof */
93 if (len == 0) {
94 break;
97 /* unrecoverable error */
98 if (len == -1) {
99 return 0;
102 offset += len;
105 return offset;
108 int tcp_socket_outgoing(const char *address, uint16_t port)
110 int s;
111 struct in_addr in;
112 struct sockaddr_in addr;
114 s = socket(PF_INET, SOCK_STREAM, 0);
115 if (s == -1) {
116 return -1;
119 if (inet_aton(address, &in) == 0) {
120 struct hostent *ent;
122 ent = gethostbyname(address);
123 if (ent == NULL) {
124 goto error;
127 memcpy(&in, ent->h_addr, sizeof(in));
130 addr.sin_family = AF_INET;
131 addr.sin_port = htons(port);
132 memcpy(&addr.sin_addr.s_addr, &in, sizeof(in));
134 if (connect(s, (struct sockaddr *)&addr, sizeof(addr)) == -1) {
135 goto error;
138 return s;
139 error:
140 closesocket(s);
141 return -1;
144 int tcp_socket_incoming(const char *address, uint16_t port)
146 int s;
147 struct in_addr in;
148 struct sockaddr_in addr;
149 int opt;
151 s = socket(PF_INET, SOCK_STREAM, 0);
152 if (s == -1) {
153 return -1;
156 if (inet_aton(address, &in) == 0) {
157 struct hostent *ent;
159 ent = gethostbyname(address);
160 if (ent == NULL) {
161 goto error;
164 memcpy(&in, ent->h_addr, sizeof(in));
167 addr.sin_family = AF_INET;
168 addr.sin_port = htons(port);
169 memcpy(&addr.sin_addr.s_addr, &in, sizeof(in));
171 opt = 1;
172 if (setsockopt(s, SOL_SOCKET, SO_REUSEADDR,
173 (const void *) &opt, sizeof(opt)) == -1) {
174 goto error;
177 if (bind(s, (struct sockaddr *)&addr, sizeof(addr)) == -1) {
178 goto error;
181 if (listen(s, 128) == -1) {
182 goto error;
185 return s;
186 error:
187 closesocket(s);
188 return -1;
191 #ifndef _WIN32
192 int unix_socket_incoming(const char *path)
194 int s;
195 struct sockaddr_un addr;
197 s = socket(PF_UNIX, SOCK_STREAM, 0);
198 if (s == -1) {
199 return -1;
202 memset(&addr, 0, sizeof(addr));
203 addr.sun_family = AF_UNIX;
204 pstrcpy(addr.sun_path, sizeof(addr.sun_path), path);
206 if (bind(s, (struct sockaddr *)&addr, sizeof(addr)) == -1) {
207 goto error;
210 if (listen(s, 128) == -1) {
211 goto error;
214 return s;
215 error:
216 closesocket(s);
217 return -1;
220 int unix_socket_outgoing(const char *path)
222 int s;
223 struct sockaddr_un addr;
225 s = socket(PF_UNIX, SOCK_STREAM, 0);
226 if (s == -1) {
227 return -1;
230 memset(&addr, 0, sizeof(addr));
231 addr.sun_family = AF_UNIX;
232 pstrcpy(addr.sun_path, sizeof(addr.sun_path), path);
234 if (connect(s, (struct sockaddr *)&addr, sizeof(addr)) == -1) {
235 goto error;
238 return s;
239 error:
240 closesocket(s);
241 return -1;
243 #else
244 int unix_socket_incoming(const char *path)
246 errno = ENOTSUP;
247 return -1;
250 int unix_socket_outgoing(const char *path)
252 errno = ENOTSUP;
253 return -1;
255 #endif
258 /* Basic flow
260 Server Client
262 Negotiate
263 Request
264 Response
265 Request
266 Response
269 Request (type == 2)
272 int nbd_negotiate(int csock, off_t size)
274 char buf[8 + 8 + 8 + 128];
276 /* Negotiate
277 [ 0 .. 7] passwd ("NBDMAGIC")
278 [ 8 .. 15] magic (0x00420281861253)
279 [16 .. 23] size
280 [24 .. 151] reserved (0)
283 TRACE("Beginning negotiation.");
284 memcpy(buf, "NBDMAGIC", 8);
285 cpu_to_be64w((uint64_t*)(buf + 8), 0x00420281861253LL);
286 cpu_to_be64w((uint64_t*)(buf + 16), size);
287 memset(buf + 24, 0, 128);
289 if (write_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
290 LOG("write failed");
291 errno = EINVAL;
292 return -1;
295 TRACE("Negotation succeeded.");
297 return 0;
300 int nbd_receive_negotiate(int csock, off_t *size, size_t *blocksize)
302 char buf[8 + 8 + 8 + 128];
303 uint64_t magic;
305 TRACE("Receiving negotation.");
307 if (read_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
308 LOG("read failed");
309 errno = EINVAL;
310 return -1;
313 magic = be64_to_cpup((uint64_t*)(buf + 8));
314 *size = be64_to_cpup((uint64_t*)(buf + 16));
315 *blocksize = 1024;
317 TRACE("Magic is %c%c%c%c%c%c%c%c",
318 qemu_isprint(buf[0]) ? buf[0] : '.',
319 qemu_isprint(buf[1]) ? buf[1] : '.',
320 qemu_isprint(buf[2]) ? buf[2] : '.',
321 qemu_isprint(buf[3]) ? buf[3] : '.',
322 qemu_isprint(buf[4]) ? buf[4] : '.',
323 qemu_isprint(buf[5]) ? buf[5] : '.',
324 qemu_isprint(buf[6]) ? buf[6] : '.',
325 qemu_isprint(buf[7]) ? buf[7] : '.');
326 TRACE("Magic is 0x%" PRIx64, magic);
327 TRACE("Size is %" PRIu64, *size);
329 if (memcmp(buf, "NBDMAGIC", 8) != 0) {
330 LOG("Invalid magic received");
331 errno = EINVAL;
332 return -1;
335 TRACE("Checking magic");
337 if (magic != 0x00420281861253LL) {
338 LOG("Bad magic received");
339 errno = EINVAL;
340 return -1;
342 return 0;
345 #ifndef _WIN32
346 int nbd_init(int fd, int csock, off_t size, size_t blocksize)
348 TRACE("Setting block size to %lu", (unsigned long)blocksize);
350 if (ioctl(fd, NBD_SET_BLKSIZE, blocksize) == -1) {
351 int serrno = errno;
352 LOG("Failed setting NBD block size");
353 errno = serrno;
354 return -1;
357 TRACE("Setting size to %llu block(s)",
358 (unsigned long long)(size / blocksize));
360 if (ioctl(fd, NBD_SET_SIZE_BLOCKS, size / blocksize) == -1) {
361 int serrno = errno;
362 LOG("Failed setting size (in blocks)");
363 errno = serrno;
364 return -1;
367 TRACE("Clearing NBD socket");
369 if (ioctl(fd, NBD_CLEAR_SOCK) == -1) {
370 int serrno = errno;
371 LOG("Failed clearing NBD socket");
372 errno = serrno;
373 return -1;
376 TRACE("Setting NBD socket");
378 if (ioctl(fd, NBD_SET_SOCK, csock) == -1) {
379 int serrno = errno;
380 LOG("Failed to set NBD socket");
381 errno = serrno;
382 return -1;
385 TRACE("Negotiation ended");
387 return 0;
390 int nbd_disconnect(int fd)
392 ioctl(fd, NBD_CLEAR_QUE);
393 ioctl(fd, NBD_DISCONNECT);
394 ioctl(fd, NBD_CLEAR_SOCK);
395 return 0;
398 int nbd_client(int fd, int csock)
400 int ret;
401 int serrno;
403 TRACE("Doing NBD loop");
405 ret = ioctl(fd, NBD_DO_IT);
406 serrno = errno;
408 TRACE("NBD loop returned %d: %s", ret, strerror(serrno));
410 TRACE("Clearing NBD queue");
411 ioctl(fd, NBD_CLEAR_QUE);
413 TRACE("Clearing NBD socket");
414 ioctl(fd, NBD_CLEAR_SOCK);
416 errno = serrno;
417 return ret;
419 #else
420 int nbd_init(int fd, int csock, off_t size, size_t blocksize)
422 errno = ENOTSUP;
423 return -1;
426 int nbd_disconnect(int fd)
428 errno = ENOTSUP;
429 return -1;
432 int nbd_client(int fd, int csock)
434 errno = ENOTSUP;
435 return -1;
437 #endif
439 int nbd_send_request(int csock, struct nbd_request *request)
441 uint8_t buf[4 + 4 + 8 + 8 + 4];
443 cpu_to_be32w((uint32_t*)buf, NBD_REQUEST_MAGIC);
444 cpu_to_be32w((uint32_t*)(buf + 4), request->type);
445 cpu_to_be64w((uint64_t*)(buf + 8), request->handle);
446 cpu_to_be64w((uint64_t*)(buf + 16), request->from);
447 cpu_to_be32w((uint32_t*)(buf + 24), request->len);
449 TRACE("Sending request to client");
451 if (write_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
452 LOG("writing to socket failed");
453 errno = EINVAL;
454 return -1;
456 return 0;
460 static int nbd_receive_request(int csock, struct nbd_request *request)
462 uint8_t buf[4 + 4 + 8 + 8 + 4];
463 uint32_t magic;
465 if (read_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
466 LOG("read failed");
467 errno = EINVAL;
468 return -1;
471 /* Request
472 [ 0 .. 3] magic (NBD_REQUEST_MAGIC)
473 [ 4 .. 7] type (0 == READ, 1 == WRITE)
474 [ 8 .. 15] handle
475 [16 .. 23] from
476 [24 .. 27] len
479 magic = be32_to_cpup((uint32_t*)buf);
480 request->type = be32_to_cpup((uint32_t*)(buf + 4));
481 request->handle = be64_to_cpup((uint64_t*)(buf + 8));
482 request->from = be64_to_cpup((uint64_t*)(buf + 16));
483 request->len = be32_to_cpup((uint32_t*)(buf + 24));
485 TRACE("Got request: "
486 "{ magic = 0x%x, .type = %d, from = %" PRIu64" , len = %u }",
487 magic, request->type, request->from, request->len);
489 if (magic != NBD_REQUEST_MAGIC) {
490 LOG("invalid magic (got 0x%x)", magic);
491 errno = EINVAL;
492 return -1;
494 return 0;
497 int nbd_receive_reply(int csock, struct nbd_reply *reply)
499 uint8_t buf[4 + 4 + 8];
500 uint32_t magic;
502 memset(buf, 0xAA, sizeof(buf));
504 if (read_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
505 LOG("read failed");
506 errno = EINVAL;
507 return -1;
510 /* Reply
511 [ 0 .. 3] magic (NBD_REPLY_MAGIC)
512 [ 4 .. 7] error (0 == no error)
513 [ 7 .. 15] handle
516 magic = be32_to_cpup((uint32_t*)buf);
517 reply->error = be32_to_cpup((uint32_t*)(buf + 4));
518 reply->handle = be64_to_cpup((uint64_t*)(buf + 8));
520 TRACE("Got reply: "
521 "{ magic = 0x%x, .error = %d, handle = %" PRIu64" }",
522 magic, reply->error, reply->handle);
524 if (magic != NBD_REPLY_MAGIC) {
525 LOG("invalid magic (got 0x%x)", magic);
526 errno = EINVAL;
527 return -1;
529 return 0;
532 static int nbd_send_reply(int csock, struct nbd_reply *reply)
534 uint8_t buf[4 + 4 + 8];
536 /* Reply
537 [ 0 .. 3] magic (NBD_REPLY_MAGIC)
538 [ 4 .. 7] error (0 == no error)
539 [ 7 .. 15] handle
541 cpu_to_be32w((uint32_t*)buf, NBD_REPLY_MAGIC);
542 cpu_to_be32w((uint32_t*)(buf + 4), reply->error);
543 cpu_to_be64w((uint64_t*)(buf + 8), reply->handle);
545 TRACE("Sending response to client");
547 if (write_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
548 LOG("writing to socket failed");
549 errno = EINVAL;
550 return -1;
552 return 0;
555 int nbd_trip(BlockDriverState *bs, int csock, off_t size, uint64_t dev_offset,
556 off_t *offset, bool readonly, uint8_t *data, int data_size)
558 struct nbd_request request;
559 struct nbd_reply reply;
561 TRACE("Reading request.");
563 if (nbd_receive_request(csock, &request) == -1)
564 return -1;
566 if (request.len > data_size) {
567 LOG("len (%u) is larger than max len (%u)",
568 request.len, data_size);
569 errno = EINVAL;
570 return -1;
573 if ((request.from + request.len) < request.from) {
574 LOG("integer overflow detected! "
575 "you're probably being attacked");
576 errno = EINVAL;
577 return -1;
580 if ((request.from + request.len) > size) {
581 LOG("From: %" PRIu64 ", Len: %u, Size: %" PRIu64
582 ", Offset: %" PRIu64 "\n",
583 request.from, request.len, (uint64_t)size, dev_offset);
584 LOG("requested operation past EOF--bad client?");
585 errno = EINVAL;
586 return -1;
589 TRACE("Decoding type");
591 reply.handle = request.handle;
592 reply.error = 0;
594 switch (request.type) {
595 case NBD_CMD_READ:
596 TRACE("Request type is READ");
598 if (bdrv_read(bs, (request.from + dev_offset) / 512, data,
599 request.len / 512) == -1) {
600 LOG("reading from file failed");
601 errno = EINVAL;
602 return -1;
604 *offset += request.len;
606 TRACE("Read %u byte(s)", request.len);
608 if (nbd_send_reply(csock, &reply) == -1)
609 return -1;
611 TRACE("Sending data to client");
613 if (write_sync(csock, data, request.len) != request.len) {
614 LOG("writing to socket failed");
615 errno = EINVAL;
616 return -1;
618 break;
619 case NBD_CMD_WRITE:
620 TRACE("Request type is WRITE");
622 TRACE("Reading %u byte(s)", request.len);
624 if (read_sync(csock, data, request.len) != request.len) {
625 LOG("reading from socket failed");
626 errno = EINVAL;
627 return -1;
630 if (readonly) {
631 TRACE("Server is read-only, return error");
632 reply.error = 1;
633 } else {
634 TRACE("Writing to device");
636 if (bdrv_write(bs, (request.from + dev_offset) / 512,
637 data, request.len / 512) == -1) {
638 LOG("writing to file failed");
639 errno = EINVAL;
640 return -1;
643 *offset += request.len;
646 if (nbd_send_reply(csock, &reply) == -1)
647 return -1;
648 break;
649 case NBD_CMD_DISC:
650 TRACE("Request type is DISCONNECT");
651 errno = 0;
652 return 1;
653 default:
654 LOG("invalid request type (%u) received", request.type);
655 errno = EINVAL;
656 return -1;
659 TRACE("Request/Reply complete");
661 return 0;