]>
git.proxmox.com Git - mirror_qemu.git/blob - nbd.c
2 * Copyright (C) 2005 Anthony Liguori <anthony@codemonkey.ws>
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, see <http://www.gnu.org/licenses/>.
25 #include <sys/ioctl.h>
27 #if defined(__sun__) || defined(__HAIKU__)
28 #include <sys/ioccom.h>
37 #include "qemu_socket.h"
42 #define TRACE(msg, ...) do { \
43 LOG(msg, ## __VA_ARGS__); \
46 #define TRACE(msg, ...) \
50 #define LOG(msg, ...) do { \
51 fprintf(stderr, "%s:%s():L%d: " msg "\n", \
52 __FILE__, __FUNCTION__, __LINE__, ## __VA_ARGS__); \
55 /* This is all part of the "official" NBD API */
57 #define NBD_REPLY_SIZE (4 + 4 + 8)
58 #define NBD_REQUEST_MAGIC 0x25609513
59 #define NBD_REPLY_MAGIC 0x67446698
61 #define NBD_SET_SOCK _IO(0xab, 0)
62 #define NBD_SET_BLKSIZE _IO(0xab, 1)
63 #define NBD_SET_SIZE _IO(0xab, 2)
64 #define NBD_DO_IT _IO(0xab, 3)
65 #define NBD_CLEAR_SOCK _IO(0xab, 4)
66 #define NBD_CLEAR_QUE _IO(0xab, 5)
67 #define NBD_PRINT_DEBUG _IO(0xab, 6)
68 #define NBD_SET_SIZE_BLOCKS _IO(0xab, 7)
69 #define NBD_DISCONNECT _IO(0xab, 8)
70 #define NBD_SET_TIMEOUT _IO(0xab, 9)
71 #define NBD_SET_FLAGS _IO(0xab, 10)
73 #define NBD_OPT_EXPORT_NAME (1 << 0)
75 /* That's all folks */
77 #define read_sync(fd, buffer, size) nbd_wr_sync(fd, buffer, size, true)
78 #define write_sync(fd, buffer, size) nbd_wr_sync(fd, buffer, size, false)
80 size_t nbd_wr_sync(int fd
, void *buffer
, size_t size
, bool do_read
)
84 if (qemu_in_coroutine()) {
86 return qemu_co_recv(fd
, buffer
, size
);
88 return qemu_co_send(fd
, buffer
, size
);
92 while (offset
< size
) {
96 len
= qemu_recv(fd
, buffer
+ offset
, size
- offset
, 0);
98 len
= send(fd
, buffer
+ offset
, size
- offset
, 0);
102 errno
= socket_error();
104 /* recoverable error */
105 if (len
== -1 && (errno
== EAGAIN
|| errno
== EINTR
)) {
114 /* unrecoverable error */
125 static void combine_addr(char *buf
, size_t len
, const char* address
,
128 /* If the address-part contains a colon, it's an IPv6 IP so needs [] */
129 if (strstr(address
, ":")) {
130 snprintf(buf
, len
, "[%s]:%u", address
, port
);
132 snprintf(buf
, len
, "%s:%u", address
, port
);
136 int tcp_socket_outgoing(const char *address
, uint16_t port
)
138 char address_and_port
[128];
139 combine_addr(address_and_port
, 128, address
, port
);
140 return tcp_socket_outgoing_spec(address_and_port
);
143 int tcp_socket_outgoing_spec(const char *address_and_port
)
145 return inet_connect(address_and_port
, SOCK_STREAM
);
148 int tcp_socket_incoming(const char *address
, uint16_t port
)
150 char address_and_port
[128];
151 combine_addr(address_and_port
, 128, address
, port
);
152 return tcp_socket_incoming_spec(address_and_port
);
155 int tcp_socket_incoming_spec(const char *address_and_port
)
159 return inet_listen(address_and_port
, ostr
, olen
, SOCK_STREAM
, 0);
162 int unix_socket_incoming(const char *path
)
167 return unix_listen(path
, ostr
, olen
);
170 int unix_socket_outgoing(const char *path
)
172 return unix_connect(path
);
189 int nbd_negotiate(int csock
, off_t size
, uint32_t flags
)
191 char buf
[8 + 8 + 8 + 128];
194 [ 0 .. 7] passwd ("NBDMAGIC")
195 [ 8 .. 15] magic (0x00420281861253)
198 [28 .. 151] reserved (0)
201 TRACE("Beginning negotiation.");
202 memcpy(buf
, "NBDMAGIC", 8);
203 cpu_to_be64w((uint64_t*)(buf
+ 8), 0x00420281861253LL
);
204 cpu_to_be64w((uint64_t*)(buf
+ 16), size
);
205 cpu_to_be32w((uint32_t*)(buf
+ 24),
206 flags
| NBD_FLAG_HAS_FLAGS
| NBD_FLAG_SEND_TRIM
|
207 NBD_FLAG_SEND_FLUSH
| NBD_FLAG_SEND_FUA
);
208 memset(buf
+ 28, 0, 124);
210 if (write_sync(csock
, buf
, sizeof(buf
)) != sizeof(buf
)) {
216 TRACE("Negotiation succeeded.");
221 int nbd_receive_negotiate(int csock
, const char *name
, uint32_t *flags
,
222 off_t
*size
, size_t *blocksize
)
228 TRACE("Receiving negotiation.");
230 if (read_sync(csock
, buf
, 8) != 8) {
237 if (strlen(buf
) == 0) {
238 LOG("server connection closed");
243 TRACE("Magic is %c%c%c%c%c%c%c%c",
244 qemu_isprint(buf
[0]) ? buf
[0] : '.',
245 qemu_isprint(buf
[1]) ? buf
[1] : '.',
246 qemu_isprint(buf
[2]) ? buf
[2] : '.',
247 qemu_isprint(buf
[3]) ? buf
[3] : '.',
248 qemu_isprint(buf
[4]) ? buf
[4] : '.',
249 qemu_isprint(buf
[5]) ? buf
[5] : '.',
250 qemu_isprint(buf
[6]) ? buf
[6] : '.',
251 qemu_isprint(buf
[7]) ? buf
[7] : '.');
253 if (memcmp(buf
, "NBDMAGIC", 8) != 0) {
254 LOG("Invalid magic received");
259 if (read_sync(csock
, &magic
, sizeof(magic
)) != sizeof(magic
)) {
264 magic
= be64_to_cpu(magic
);
265 TRACE("Magic is 0x%" PRIx64
, magic
);
268 uint32_t reserved
= 0;
272 TRACE("Checking magic (opts_magic)");
273 if (magic
!= 0x49484156454F5054LL
) {
274 LOG("Bad magic received");
278 if (read_sync(csock
, &tmp
, sizeof(tmp
)) != sizeof(tmp
)) {
279 LOG("flags read failed");
283 *flags
= be16_to_cpu(tmp
) << 16;
284 /* reserved for future use */
285 if (write_sync(csock
, &reserved
, sizeof(reserved
)) !=
287 LOG("write failed (reserved)");
291 /* write the export name */
292 magic
= cpu_to_be64(magic
);
293 if (write_sync(csock
, &magic
, sizeof(magic
)) != sizeof(magic
)) {
294 LOG("write failed (magic)");
298 opt
= cpu_to_be32(NBD_OPT_EXPORT_NAME
);
299 if (write_sync(csock
, &opt
, sizeof(opt
)) != sizeof(opt
)) {
300 LOG("write failed (opt)");
304 namesize
= cpu_to_be32(strlen(name
));
305 if (write_sync(csock
, &namesize
, sizeof(namesize
)) !=
307 LOG("write failed (namesize)");
311 if (write_sync(csock
, (char*)name
, strlen(name
)) != strlen(name
)) {
312 LOG("write failed (name)");
317 TRACE("Checking magic (cli_magic)");
319 if (magic
!= 0x00420281861253LL
) {
320 LOG("Bad magic received");
326 if (read_sync(csock
, &s
, sizeof(s
)) != sizeof(s
)) {
331 *size
= be64_to_cpu(s
);
333 TRACE("Size is %" PRIu64
, *size
);
336 if (read_sync(csock
, flags
, sizeof(*flags
)) != sizeof(*flags
)) {
337 LOG("read failed (flags)");
341 *flags
= be32_to_cpup(flags
);
343 if (read_sync(csock
, &tmp
, sizeof(tmp
)) != sizeof(tmp
)) {
344 LOG("read failed (tmp)");
348 *flags
|= be32_to_cpu(tmp
);
350 if (read_sync(csock
, &buf
, 124) != 124) {
351 LOG("read failed (buf)");
359 int nbd_init(int fd
, int csock
, uint32_t flags
, off_t size
, size_t blocksize
)
361 TRACE("Setting NBD socket");
363 if (ioctl(fd
, NBD_SET_SOCK
, csock
) == -1) {
365 LOG("Failed to set NBD socket");
370 TRACE("Setting block size to %lu", (unsigned long)blocksize
);
372 if (ioctl(fd
, NBD_SET_BLKSIZE
, blocksize
) == -1) {
374 LOG("Failed setting NBD block size");
379 TRACE("Setting size to %zd block(s)", (size_t)(size
/ blocksize
));
381 if (ioctl(fd
, NBD_SET_SIZE_BLOCKS
, size
/ blocksize
) == -1) {
383 LOG("Failed setting size (in blocks)");
388 if (flags
& NBD_FLAG_READ_ONLY
) {
390 TRACE("Setting readonly attribute");
392 if (ioctl(fd
, BLKROSET
, (unsigned long) &read_only
) < 0) {
394 LOG("Failed setting read-only attribute");
400 if (ioctl(fd
, NBD_SET_FLAGS
, flags
) < 0
401 && errno
!= ENOTTY
) {
403 LOG("Failed setting flags");
408 TRACE("Negotiation ended");
413 int nbd_disconnect(int fd
)
415 ioctl(fd
, NBD_CLEAR_QUE
);
416 ioctl(fd
, NBD_DISCONNECT
);
417 ioctl(fd
, NBD_CLEAR_SOCK
);
421 int nbd_client(int fd
)
426 TRACE("Doing NBD loop");
428 ret
= ioctl(fd
, NBD_DO_IT
);
429 if (ret
== -1 && errno
== EPIPE
) {
430 /* NBD_DO_IT normally returns EPIPE when someone has disconnected
431 * the socket via NBD_DISCONNECT. We do not want to return 1 in
438 TRACE("NBD loop returned %d: %s", ret
, strerror(serrno
));
440 TRACE("Clearing NBD queue");
441 ioctl(fd
, NBD_CLEAR_QUE
);
443 TRACE("Clearing NBD socket");
444 ioctl(fd
, NBD_CLEAR_SOCK
);
450 int nbd_init(int fd
, int csock
, uint32_t flags
, off_t size
, size_t blocksize
)
456 int nbd_disconnect(int fd
)
462 int nbd_client(int fd
)
469 int nbd_send_request(int csock
, struct nbd_request
*request
)
471 uint8_t buf
[4 + 4 + 8 + 8 + 4];
473 cpu_to_be32w((uint32_t*)buf
, NBD_REQUEST_MAGIC
);
474 cpu_to_be32w((uint32_t*)(buf
+ 4), request
->type
);
475 cpu_to_be64w((uint64_t*)(buf
+ 8), request
->handle
);
476 cpu_to_be64w((uint64_t*)(buf
+ 16), request
->from
);
477 cpu_to_be32w((uint32_t*)(buf
+ 24), request
->len
);
479 TRACE("Sending request to client: "
480 "{ .from = %" PRIu64
", .len = %u, .handle = %" PRIu64
", .type=%i}",
481 request
->from
, request
->len
, request
->handle
, request
->type
);
483 if (write_sync(csock
, buf
, sizeof(buf
)) != sizeof(buf
)) {
484 LOG("writing to socket failed");
491 static int nbd_receive_request(int csock
, struct nbd_request
*request
)
493 uint8_t buf
[4 + 4 + 8 + 8 + 4];
496 if (read_sync(csock
, buf
, sizeof(buf
)) != sizeof(buf
)) {
503 [ 0 .. 3] magic (NBD_REQUEST_MAGIC)
504 [ 4 .. 7] type (0 == READ, 1 == WRITE)
510 magic
= be32_to_cpup((uint32_t*)buf
);
511 request
->type
= be32_to_cpup((uint32_t*)(buf
+ 4));
512 request
->handle
= be64_to_cpup((uint64_t*)(buf
+ 8));
513 request
->from
= be64_to_cpup((uint64_t*)(buf
+ 16));
514 request
->len
= be32_to_cpup((uint32_t*)(buf
+ 24));
516 TRACE("Got request: "
517 "{ magic = 0x%x, .type = %d, from = %" PRIu64
" , len = %u }",
518 magic
, request
->type
, request
->from
, request
->len
);
520 if (magic
!= NBD_REQUEST_MAGIC
) {
521 LOG("invalid magic (got 0x%x)", magic
);
528 int nbd_receive_reply(int csock
, struct nbd_reply
*reply
)
530 uint8_t buf
[NBD_REPLY_SIZE
];
533 memset(buf
, 0xAA, sizeof(buf
));
535 if (read_sync(csock
, buf
, sizeof(buf
)) != sizeof(buf
)) {
542 [ 0 .. 3] magic (NBD_REPLY_MAGIC)
543 [ 4 .. 7] error (0 == no error)
547 magic
= be32_to_cpup((uint32_t*)buf
);
548 reply
->error
= be32_to_cpup((uint32_t*)(buf
+ 4));
549 reply
->handle
= be64_to_cpup((uint64_t*)(buf
+ 8));
552 "{ magic = 0x%x, .error = %d, handle = %" PRIu64
" }",
553 magic
, reply
->error
, reply
->handle
);
555 if (magic
!= NBD_REPLY_MAGIC
) {
556 LOG("invalid magic (got 0x%x)", magic
);
563 static int nbd_send_reply(int csock
, struct nbd_reply
*reply
)
565 uint8_t buf
[4 + 4 + 8];
568 [ 0 .. 3] magic (NBD_REPLY_MAGIC)
569 [ 4 .. 7] error (0 == no error)
572 cpu_to_be32w((uint32_t*)buf
, NBD_REPLY_MAGIC
);
573 cpu_to_be32w((uint32_t*)(buf
+ 4), reply
->error
);
574 cpu_to_be64w((uint64_t*)(buf
+ 8), reply
->handle
);
576 TRACE("Sending response to client");
578 if (write_sync(csock
, buf
, sizeof(buf
)) != sizeof(buf
)) {
579 LOG("writing to socket failed");
586 int nbd_trip(BlockDriverState
*bs
, int csock
, off_t size
,
587 uint64_t dev_offset
, uint32_t nbdflags
,
590 struct nbd_request request
;
591 struct nbd_reply reply
;
594 TRACE("Reading request.");
596 if (nbd_receive_request(csock
, &request
) == -1)
599 if (request
.len
> NBD_BUFFER_SIZE
) {
600 LOG("len (%u) is larger than max len (%u)",
601 request
.len
, NBD_BUFFER_SIZE
);
606 if ((request
.from
+ request
.len
) < request
.from
) {
607 LOG("integer overflow detected! "
608 "you're probably being attacked");
613 if ((request
.from
+ request
.len
) > size
) {
614 LOG("From: %" PRIu64
", Len: %u, Size: %" PRIu64
615 ", Offset: %" PRIu64
"\n",
616 request
.from
, request
.len
, (uint64_t)size
, dev_offset
);
617 LOG("requested operation past EOF--bad client?");
622 TRACE("Decoding type");
624 reply
.handle
= request
.handle
;
627 switch (request
.type
& NBD_CMD_MASK_COMMAND
) {
629 TRACE("Request type is READ");
631 ret
= bdrv_read(bs
, (request
.from
+ dev_offset
) / 512,
632 data
, request
.len
/ 512);
634 LOG("reading from file failed");
639 TRACE("Read %u byte(s)", request
.len
);
640 socket_set_cork(csock
, 1);
641 if (nbd_send_reply(csock
, &reply
) == -1)
644 TRACE("Sending data to client");
646 if (write_sync(csock
, data
, request
.len
) != request
.len
) {
647 LOG("writing to socket failed");
651 socket_set_cork(csock
, 0);
654 TRACE("Request type is WRITE");
656 TRACE("Reading %u byte(s)", request
.len
);
658 if (read_sync(csock
, data
, request
.len
) != request
.len
) {
659 LOG("reading from socket failed");
664 if (nbdflags
& NBD_FLAG_READ_ONLY
) {
665 TRACE("Server is read-only, return error");
668 TRACE("Writing to device");
670 ret
= bdrv_write(bs
, (request
.from
+ dev_offset
) / 512,
671 data
, request
.len
/ 512);
673 LOG("writing to file failed");
678 if (request
.type
& NBD_CMD_FLAG_FUA
) {
679 ret
= bdrv_flush(bs
);
687 if (nbd_send_reply(csock
, &reply
) == -1)
691 TRACE("Request type is DISCONNECT");
695 TRACE("Request type is FLUSH");
697 ret
= bdrv_flush(bs
);
703 if (nbd_send_reply(csock
, &reply
) == -1)
707 TRACE("Request type is TRIM");
708 ret
= bdrv_discard(bs
, (request
.from
+ dev_offset
) / 512,
711 LOG("discard failed");
714 if (nbd_send_reply(csock
, &reply
) == -1)
718 LOG("invalid request type (%u) received", request
.type
);
723 TRACE("Request/Reply complete");