]> git.proxmox.com Git - mirror_qemu.git/blame_incremental - nbd.c
CRIS: Dont flush kernel protected pages when modifying $pid.
[mirror_qemu.git] / nbd.c
... / ...
CommitLineData
1/*
2 * Copyright (C) 2005 Anthony Liguori <anthony@codemonkey.ws>
3 *
4 * Network Block Device
5 *
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.
9 *
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.
14 *
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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
18 */
19
20#include "nbd.h"
21
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>
32
33#include "qemu_socket.h"
34
35//#define DEBUG_NBD
36
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
45
46#define LOG(msg, ...) do { \
47 fprintf(stderr, "%s:%s():L%d: " msg "\n", \
48 __FILE__, __FUNCTION__, __LINE__, ## __VA_ARGS__); \
49} while(0)
50
51/* This is all part of the "official" NBD API */
52
53#define NBD_REQUEST_MAGIC 0x25609513
54#define NBD_REPLY_MAGIC 0x67446698
55
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)
65
66/* That's all folks */
67
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)
70
71size_t nbd_wr_sync(int fd, void *buffer, size_t size, bool do_read)
72{
73 size_t offset = 0;
74
75 while (offset < size) {
76 ssize_t len;
77
78 if (do_read) {
79 len = recv(fd, buffer + offset, size - offset, 0);
80 } else {
81 len = send(fd, buffer + offset, size - offset, 0);
82 }
83
84 if (len == -1)
85 errno = socket_error();
86
87 /* recoverable error */
88 if (len == -1 && (errno == EAGAIN || errno == EINTR)) {
89 continue;
90 }
91
92 /* eof */
93 if (len == 0) {
94 break;
95 }
96
97 /* unrecoverable error */
98 if (len == -1) {
99 return 0;
100 }
101
102 offset += len;
103 }
104
105 return offset;
106}
107
108int tcp_socket_outgoing(const char *address, uint16_t port)
109{
110 int s;
111 struct in_addr in;
112 struct sockaddr_in addr;
113
114 s = socket(PF_INET, SOCK_STREAM, 0);
115 if (s == -1) {
116 return -1;
117 }
118
119 if (inet_aton(address, &in) == 0) {
120 struct hostent *ent;
121
122 ent = gethostbyname(address);
123 if (ent == NULL) {
124 goto error;
125 }
126
127 memcpy(&in, ent->h_addr, sizeof(in));
128 }
129
130 addr.sin_family = AF_INET;
131 addr.sin_port = htons(port);
132 memcpy(&addr.sin_addr.s_addr, &in, sizeof(in));
133
134 if (connect(s, (struct sockaddr *)&addr, sizeof(addr)) == -1) {
135 goto error;
136 }
137
138 return s;
139error:
140 closesocket(s);
141 return -1;
142}
143
144int tcp_socket_incoming(const char *address, uint16_t port)
145{
146 int s;
147 struct in_addr in;
148 struct sockaddr_in addr;
149 int opt;
150
151 s = socket(PF_INET, SOCK_STREAM, 0);
152 if (s == -1) {
153 return -1;
154 }
155
156 if (inet_aton(address, &in) == 0) {
157 struct hostent *ent;
158
159 ent = gethostbyname(address);
160 if (ent == NULL) {
161 goto error;
162 }
163
164 memcpy(&in, ent->h_addr, sizeof(in));
165 }
166
167 addr.sin_family = AF_INET;
168 addr.sin_port = htons(port);
169 memcpy(&addr.sin_addr.s_addr, &in, sizeof(in));
170
171 opt = 1;
172 if (setsockopt(s, SOL_SOCKET, SO_REUSEADDR, &opt, sizeof(opt)) == -1) {
173 goto error;
174 }
175
176 if (bind(s, (struct sockaddr *)&addr, sizeof(addr)) == -1) {
177 goto error;
178 }
179
180 if (listen(s, 128) == -1) {
181 goto error;
182 }
183
184 return s;
185error:
186 closesocket(s);
187 return -1;
188}
189
190#ifndef _WIN32
191int unix_socket_incoming(const char *path)
192{
193 int s;
194 struct sockaddr_un addr;
195
196 s = socket(PF_UNIX, SOCK_STREAM, 0);
197 if (s == -1) {
198 return -1;
199 }
200
201 memset(&addr, 0, sizeof(addr));
202 addr.sun_family = AF_UNIX;
203 pstrcpy(addr.sun_path, sizeof(addr.sun_path), path);
204
205 if (bind(s, (struct sockaddr *)&addr, sizeof(addr)) == -1) {
206 goto error;
207 }
208
209 if (listen(s, 128) == -1) {
210 goto error;
211 }
212
213 return s;
214error:
215 closesocket(s);
216 return -1;
217}
218
219int unix_socket_outgoing(const char *path)
220{
221 int s;
222 struct sockaddr_un addr;
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;
238error:
239 closesocket(s);
240 return -1;
241}
242#else
243int unix_socket_incoming(const char *path)
244{
245 errno = ENOTSUP;
246 return -1;
247}
248
249int unix_socket_outgoing(const char *path)
250{
251 errno = ENOTSUP;
252 return -1;
253}
254#endif
255
256
257/* Basic flow
258
259 Server Client
260
261 Negotiate
262 Request
263 Response
264 Request
265 Response
266 ...
267 ...
268 Request (type == 2)
269*/
270
271int nbd_negotiate(int csock, off_t size)
272{
273 char buf[8 + 8 + 8 + 128];
274
275 /* Negotiate
276 [ 0 .. 7] passwd ("NBDMAGIC")
277 [ 8 .. 15] magic (0x00420281861253)
278 [16 .. 23] size
279 [24 .. 151] reserved (0)
280 */
281
282 TRACE("Beginning negotiation.");
283 memcpy(buf, "NBDMAGIC", 8);
284 cpu_to_be64w((uint64_t*)(buf + 8), 0x00420281861253LL);
285 cpu_to_be64w((uint64_t*)(buf + 16), size);
286 memset(buf + 24, 0, 128);
287
288 if (write_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
289 LOG("write failed");
290 errno = EINVAL;
291 return -1;
292 }
293
294 TRACE("Negotation succeeded.");
295
296 return 0;
297}
298
299int nbd_receive_negotiate(int csock, off_t *size, size_t *blocksize)
300{
301 char buf[8 + 8 + 8 + 128];
302 uint64_t magic;
303
304 TRACE("Receiving negotation.");
305
306 if (read_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
307 LOG("read failed");
308 errno = EINVAL;
309 return -1;
310 }
311
312 magic = be64_to_cpup((uint64_t*)(buf + 8));
313 *size = be64_to_cpup((uint64_t*)(buf + 16));
314 *blocksize = 1024;
315
316 TRACE("Magic is %c%c%c%c%c%c%c%c",
317 qemu_isprint(buf[0]) ? buf[0] : '.',
318 qemu_isprint(buf[1]) ? buf[1] : '.',
319 qemu_isprint(buf[2]) ? buf[2] : '.',
320 qemu_isprint(buf[3]) ? buf[3] : '.',
321 qemu_isprint(buf[4]) ? buf[4] : '.',
322 qemu_isprint(buf[5]) ? buf[5] : '.',
323 qemu_isprint(buf[6]) ? buf[6] : '.',
324 qemu_isprint(buf[7]) ? buf[7] : '.');
325 TRACE("Magic is 0x%" PRIx64, magic);
326 TRACE("Size is %" PRIu64, *size);
327
328 if (memcmp(buf, "NBDMAGIC", 8) != 0) {
329 LOG("Invalid magic received");
330 errno = EINVAL;
331 return -1;
332 }
333
334 TRACE("Checking magic");
335
336 if (magic != 0x00420281861253LL) {
337 LOG("Bad magic received");
338 errno = EINVAL;
339 return -1;
340 }
341 return 0;
342}
343
344#ifndef _WIN32
345int nbd_init(int fd, int csock, off_t size, size_t blocksize)
346{
347 TRACE("Setting block size to %lu", (unsigned long)blocksize);
348
349 if (ioctl(fd, NBD_SET_BLKSIZE, blocksize) == -1) {
350 int serrno = errno;
351 LOG("Failed setting NBD block size");
352 errno = serrno;
353 return -1;
354 }
355
356 TRACE("Setting size to %llu block(s)",
357 (unsigned long long)(size / blocksize));
358
359 if (ioctl(fd, NBD_SET_SIZE_BLOCKS, size / blocksize) == -1) {
360 int serrno = errno;
361 LOG("Failed setting size (in blocks)");
362 errno = serrno;
363 return -1;
364 }
365
366 TRACE("Clearing NBD socket");
367
368 if (ioctl(fd, NBD_CLEAR_SOCK) == -1) {
369 int serrno = errno;
370 LOG("Failed clearing NBD socket");
371 errno = serrno;
372 return -1;
373 }
374
375 TRACE("Setting NBD socket");
376
377 if (ioctl(fd, NBD_SET_SOCK, csock) == -1) {
378 int serrno = errno;
379 LOG("Failed to set NBD socket");
380 errno = serrno;
381 return -1;
382 }
383
384 TRACE("Negotiation ended");
385
386 return 0;
387}
388
389int nbd_disconnect(int fd)
390{
391 ioctl(fd, NBD_CLEAR_QUE);
392 ioctl(fd, NBD_DISCONNECT);
393 ioctl(fd, NBD_CLEAR_SOCK);
394 return 0;
395}
396
397int nbd_client(int fd, int csock)
398{
399 int ret;
400 int serrno;
401
402 TRACE("Doing NBD loop");
403
404 ret = ioctl(fd, NBD_DO_IT);
405 serrno = errno;
406
407 TRACE("NBD loop returned %d: %s", ret, strerror(serrno));
408
409 TRACE("Clearing NBD queue");
410 ioctl(fd, NBD_CLEAR_QUE);
411
412 TRACE("Clearing NBD socket");
413 ioctl(fd, NBD_CLEAR_SOCK);
414
415 errno = serrno;
416 return ret;
417}
418#else
419int nbd_init(int fd, int csock, off_t size, size_t blocksize)
420{
421 errno = ENOTSUP;
422 return -1;
423}
424
425int nbd_disconnect(int fd)
426{
427 errno = ENOTSUP;
428 return -1;
429}
430
431int nbd_client(int fd, int csock)
432{
433 errno = ENOTSUP;
434 return -1;
435}
436#endif
437
438int nbd_send_request(int csock, struct nbd_request *request)
439{
440 uint8_t buf[4 + 4 + 8 + 8 + 4];
441
442 cpu_to_be32w((uint32_t*)buf, NBD_REQUEST_MAGIC);
443 cpu_to_be32w((uint32_t*)(buf + 4), request->type);
444 cpu_to_be64w((uint64_t*)(buf + 8), request->handle);
445 cpu_to_be64w((uint64_t*)(buf + 16), request->from);
446 cpu_to_be32w((uint32_t*)(buf + 24), request->len);
447
448 TRACE("Sending request to client");
449
450 if (write_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
451 LOG("writing to socket failed");
452 errno = EINVAL;
453 return -1;
454 }
455 return 0;
456}
457
458
459static int nbd_receive_request(int csock, struct nbd_request *request)
460{
461 uint8_t buf[4 + 4 + 8 + 8 + 4];
462 uint32_t magic;
463
464 if (read_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
465 LOG("read failed");
466 errno = EINVAL;
467 return -1;
468 }
469
470 /* Request
471 [ 0 .. 3] magic (NBD_REQUEST_MAGIC)
472 [ 4 .. 7] type (0 == READ, 1 == WRITE)
473 [ 8 .. 15] handle
474 [16 .. 23] from
475 [24 .. 27] len
476 */
477
478 magic = be32_to_cpup((uint32_t*)buf);
479 request->type = be32_to_cpup((uint32_t*)(buf + 4));
480 request->handle = be64_to_cpup((uint64_t*)(buf + 8));
481 request->from = be64_to_cpup((uint64_t*)(buf + 16));
482 request->len = be32_to_cpup((uint32_t*)(buf + 24));
483
484 TRACE("Got request: "
485 "{ magic = 0x%x, .type = %d, from = %" PRIu64" , len = %u }",
486 magic, request->type, request->from, request->len);
487
488 if (magic != NBD_REQUEST_MAGIC) {
489 LOG("invalid magic (got 0x%x)", magic);
490 errno = EINVAL;
491 return -1;
492 }
493 return 0;
494}
495
496int nbd_receive_reply(int csock, struct nbd_reply *reply)
497{
498 uint8_t buf[4 + 4 + 8];
499 uint32_t magic;
500
501 memset(buf, 0xAA, sizeof(buf));
502
503 if (read_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
504 LOG("read failed");
505 errno = EINVAL;
506 return -1;
507 }
508
509 /* Reply
510 [ 0 .. 3] magic (NBD_REPLY_MAGIC)
511 [ 4 .. 7] error (0 == no error)
512 [ 7 .. 15] handle
513 */
514
515 magic = be32_to_cpup((uint32_t*)buf);
516 reply->error = be32_to_cpup((uint32_t*)(buf + 4));
517 reply->handle = be64_to_cpup((uint64_t*)(buf + 8));
518
519 TRACE("Got reply: "
520 "{ magic = 0x%x, .error = %d, handle = %" PRIu64" }",
521 magic, reply->error, reply->handle);
522
523 if (magic != NBD_REPLY_MAGIC) {
524 LOG("invalid magic (got 0x%x)", magic);
525 errno = EINVAL;
526 return -1;
527 }
528 return 0;
529}
530
531static int nbd_send_reply(int csock, struct nbd_reply *reply)
532{
533 uint8_t buf[4 + 4 + 8];
534
535 /* Reply
536 [ 0 .. 3] magic (NBD_REPLY_MAGIC)
537 [ 4 .. 7] error (0 == no error)
538 [ 7 .. 15] handle
539 */
540 cpu_to_be32w((uint32_t*)buf, NBD_REPLY_MAGIC);
541 cpu_to_be32w((uint32_t*)(buf + 4), reply->error);
542 cpu_to_be64w((uint64_t*)(buf + 8), reply->handle);
543
544 TRACE("Sending response to client");
545
546 if (write_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
547 LOG("writing to socket failed");
548 errno = EINVAL;
549 return -1;
550 }
551 return 0;
552}
553
554int nbd_trip(BlockDriverState *bs, int csock, off_t size, uint64_t dev_offset,
555 off_t *offset, bool readonly, uint8_t *data, int data_size)
556{
557 struct nbd_request request;
558 struct nbd_reply reply;
559
560 TRACE("Reading request.");
561
562 if (nbd_receive_request(csock, &request) == -1)
563 return -1;
564
565 if (request.len > data_size) {
566 LOG("len (%u) is larger than max len (%u)",
567 request.len, data_size);
568 errno = EINVAL;
569 return -1;
570 }
571
572 if ((request.from + request.len) < request.from) {
573 LOG("integer overflow detected! "
574 "you're probably being attacked");
575 errno = EINVAL;
576 return -1;
577 }
578
579 if ((request.from + request.len) > size) {
580 LOG("From: %" PRIu64 ", Len: %u, Size: %" PRIu64
581 ", Offset: %" PRIu64 "\n",
582 request.from, request.len, size, dev_offset);
583 LOG("requested operation past EOF--bad client?");
584 errno = EINVAL;
585 return -1;
586 }
587
588 TRACE("Decoding type");
589
590 reply.handle = request.handle;
591 reply.error = 0;
592
593 switch (request.type) {
594 case NBD_CMD_READ:
595 TRACE("Request type is READ");
596
597 if (bdrv_read(bs, (request.from + dev_offset) / 512, data,
598 request.len / 512) == -1) {
599 LOG("reading from file failed");
600 errno = EINVAL;
601 return -1;
602 }
603 *offset += request.len;
604
605 TRACE("Read %u byte(s)", request.len);
606
607 if (nbd_send_reply(csock, &reply) == -1)
608 return -1;
609
610 TRACE("Sending data to client");
611
612 if (write_sync(csock, data, request.len) != request.len) {
613 LOG("writing to socket failed");
614 errno = EINVAL;
615 return -1;
616 }
617 break;
618 case NBD_CMD_WRITE:
619 TRACE("Request type is WRITE");
620
621 TRACE("Reading %u byte(s)", request.len);
622
623 if (read_sync(csock, data, request.len) != request.len) {
624 LOG("reading from socket failed");
625 errno = EINVAL;
626 return -1;
627 }
628
629 if (readonly) {
630 TRACE("Server is read-only, return error");
631 reply.error = 1;
632 } else {
633 TRACE("Writing to device");
634
635 if (bdrv_write(bs, (request.from + dev_offset) / 512,
636 data, request.len / 512) == -1) {
637 LOG("writing to file failed");
638 errno = EINVAL;
639 return -1;
640 }
641
642 *offset += request.len;
643 }
644
645 if (nbd_send_reply(csock, &reply) == -1)
646 return -1;
647 break;
648 case NBD_CMD_DISC:
649 TRACE("Request type is DISCONNECT");
650 errno = 0;
651 return 1;
652 default:
653 LOG("invalid request type (%u) received", request.type);
654 errno = EINVAL;
655 return -1;
656 }
657
658 TRACE("Request/Reply complete");
659
660 return 0;
661}