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/>.
24 #include <sys/ioctl.h>
26 #if defined(__sun__) || defined(__HAIKU__)
27 #include <sys/ioccom.h>
32 #include "qemu_socket.h"
37 #define TRACE(msg, ...) do { \
38 LOG(msg, ## __VA_ARGS__); \
41 #define TRACE(msg, ...) \
45 #define LOG(msg, ...) do { \
46 fprintf(stderr, "%s:%s():L%d: " msg "\n", \
47 __FILE__, __FUNCTION__, __LINE__, ## __VA_ARGS__); \
50 /* This is all part of the "official" NBD API */
52 #define NBD_REPLY_SIZE (4 + 4 + 8)
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 #define NBD_OPT_EXPORT_NAME (1 << 0)
68 /* That's all folks */
70 #define read_sync(fd, buffer, size) nbd_wr_sync(fd, buffer, size, true)
71 #define write_sync(fd, buffer, size) nbd_wr_sync(fd, buffer, size, false)
73 size_t nbd_wr_sync(int fd
, void *buffer
, size_t size
, bool do_read
)
77 while (offset
< size
) {
81 len
= recv(fd
, buffer
+ offset
, size
- offset
, 0);
83 len
= send(fd
, buffer
+ offset
, size
- offset
, 0);
87 errno
= socket_error();
89 /* recoverable error */
90 if (len
== -1 && (errno
== EAGAIN
|| errno
== EINTR
)) {
99 /* unrecoverable error */
110 int tcp_socket_outgoing(const char *address
, uint16_t port
)
114 struct sockaddr_in addr
;
116 s
= socket(PF_INET
, SOCK_STREAM
, 0);
121 if (inet_aton(address
, &in
) == 0) {
124 ent
= gethostbyname(address
);
129 memcpy(&in
, ent
->h_addr
, sizeof(in
));
132 addr
.sin_family
= AF_INET
;
133 addr
.sin_port
= htons(port
);
134 memcpy(&addr
.sin_addr
.s_addr
, &in
, sizeof(in
));
136 if (connect(s
, (struct sockaddr
*)&addr
, sizeof(addr
)) == -1) {
146 int tcp_socket_incoming(const char *address
, uint16_t port
)
150 struct sockaddr_in addr
;
153 s
= socket(PF_INET
, SOCK_STREAM
, 0);
158 if (inet_aton(address
, &in
) == 0) {
161 ent
= gethostbyname(address
);
166 memcpy(&in
, ent
->h_addr
, sizeof(in
));
169 addr
.sin_family
= AF_INET
;
170 addr
.sin_port
= htons(port
);
171 memcpy(&addr
.sin_addr
.s_addr
, &in
, sizeof(in
));
174 if (setsockopt(s
, SOL_SOCKET
, SO_REUSEADDR
,
175 (const void *) &opt
, sizeof(opt
)) == -1) {
179 if (bind(s
, (struct sockaddr
*)&addr
, sizeof(addr
)) == -1) {
183 if (listen(s
, 128) == -1) {
194 int unix_socket_incoming(const char *path
)
197 struct sockaddr_un addr
;
199 s
= socket(PF_UNIX
, SOCK_STREAM
, 0);
204 memset(&addr
, 0, sizeof(addr
));
205 addr
.sun_family
= AF_UNIX
;
206 pstrcpy(addr
.sun_path
, sizeof(addr
.sun_path
), path
);
208 if (bind(s
, (struct sockaddr
*)&addr
, sizeof(addr
)) == -1) {
212 if (listen(s
, 128) == -1) {
222 int unix_socket_outgoing(const char *path
)
225 struct sockaddr_un addr
;
227 s
= socket(PF_UNIX
, SOCK_STREAM
, 0);
232 memset(&addr
, 0, sizeof(addr
));
233 addr
.sun_family
= AF_UNIX
;
234 pstrcpy(addr
.sun_path
, sizeof(addr
.sun_path
), path
);
236 if (connect(s
, (struct sockaddr
*)&addr
, sizeof(addr
)) == -1) {
246 int unix_socket_incoming(const char *path
)
252 int unix_socket_outgoing(const char *path
)
274 int nbd_negotiate(int csock
, off_t size
)
276 char buf
[8 + 8 + 8 + 128];
279 [ 0 .. 7] passwd ("NBDMAGIC")
280 [ 8 .. 15] magic (0x00420281861253)
282 [24 .. 151] reserved (0)
285 TRACE("Beginning negotiation.");
286 memcpy(buf
, "NBDMAGIC", 8);
287 cpu_to_be64w((uint64_t*)(buf
+ 8), 0x00420281861253LL
);
288 cpu_to_be64w((uint64_t*)(buf
+ 16), size
);
289 memset(buf
+ 24, 0, 128);
291 if (write_sync(csock
, buf
, sizeof(buf
)) != sizeof(buf
)) {
297 TRACE("Negotation succeeded.");
302 int nbd_receive_negotiate(int csock
, const char *name
, uint32_t *flags
,
303 off_t
*size
, size_t *blocksize
)
309 TRACE("Receiving negotation.");
311 if (read_sync(csock
, buf
, 8) != 8) {
318 if (strlen(buf
) == 0) {
319 LOG("server connection closed");
324 TRACE("Magic is %c%c%c%c%c%c%c%c",
325 qemu_isprint(buf
[0]) ? buf
[0] : '.',
326 qemu_isprint(buf
[1]) ? buf
[1] : '.',
327 qemu_isprint(buf
[2]) ? buf
[2] : '.',
328 qemu_isprint(buf
[3]) ? buf
[3] : '.',
329 qemu_isprint(buf
[4]) ? buf
[4] : '.',
330 qemu_isprint(buf
[5]) ? buf
[5] : '.',
331 qemu_isprint(buf
[6]) ? buf
[6] : '.',
332 qemu_isprint(buf
[7]) ? buf
[7] : '.');
334 if (memcmp(buf
, "NBDMAGIC", 8) != 0) {
335 LOG("Invalid magic received");
340 if (read_sync(csock
, &magic
, sizeof(magic
)) != sizeof(magic
)) {
345 magic
= be64_to_cpu(magic
);
346 TRACE("Magic is 0x%" PRIx64
, magic
);
349 uint32_t reserved
= 0;
353 TRACE("Checking magic (opts_magic)");
354 if (magic
!= 0x49484156454F5054LL
) {
355 LOG("Bad magic received");
359 if (read_sync(csock
, &tmp
, sizeof(tmp
)) != sizeof(tmp
)) {
360 LOG("flags read failed");
364 *flags
= be16_to_cpu(tmp
) << 16;
365 /* reserved for future use */
366 if (write_sync(csock
, &reserved
, sizeof(reserved
)) !=
368 LOG("write failed (reserved)");
372 /* write the export name */
373 magic
= cpu_to_be64(magic
);
374 if (write_sync(csock
, &magic
, sizeof(magic
)) != sizeof(magic
)) {
375 LOG("write failed (magic)");
379 opt
= cpu_to_be32(NBD_OPT_EXPORT_NAME
);
380 if (write_sync(csock
, &opt
, sizeof(opt
)) != sizeof(opt
)) {
381 LOG("write failed (opt)");
385 namesize
= cpu_to_be32(strlen(name
));
386 if (write_sync(csock
, &namesize
, sizeof(namesize
)) !=
388 LOG("write failed (namesize)");
392 if (write_sync(csock
, (char*)name
, strlen(name
)) != strlen(name
)) {
393 LOG("write failed (name)");
398 TRACE("Checking magic (cli_magic)");
400 if (magic
!= 0x00420281861253LL
) {
401 LOG("Bad magic received");
407 if (read_sync(csock
, &s
, sizeof(s
)) != sizeof(s
)) {
412 *size
= be64_to_cpu(s
);
414 TRACE("Size is %" PRIu64
, *size
);
417 if (read_sync(csock
, flags
, sizeof(*flags
)) != sizeof(*flags
)) {
418 LOG("read failed (flags)");
422 *flags
= be32_to_cpup(flags
);
424 if (read_sync(csock
, &tmp
, sizeof(tmp
)) != sizeof(tmp
)) {
425 LOG("read failed (tmp)");
429 *flags
|= be32_to_cpu(tmp
);
431 if (read_sync(csock
, &buf
, 124) != 124) {
432 LOG("read failed (buf)");
440 int nbd_init(int fd
, int csock
, off_t size
, size_t blocksize
)
442 TRACE("Setting block size to %lu", (unsigned long)blocksize
);
444 if (ioctl(fd
, NBD_SET_BLKSIZE
, blocksize
) == -1) {
446 LOG("Failed setting NBD block size");
451 TRACE("Setting size to %zd block(s)", (size_t)(size
/ blocksize
));
453 if (ioctl(fd
, NBD_SET_SIZE_BLOCKS
, size
/ blocksize
) == -1) {
455 LOG("Failed setting size (in blocks)");
460 TRACE("Clearing NBD socket");
462 if (ioctl(fd
, NBD_CLEAR_SOCK
) == -1) {
464 LOG("Failed clearing NBD socket");
469 TRACE("Setting NBD socket");
471 if (ioctl(fd
, NBD_SET_SOCK
, csock
) == -1) {
473 LOG("Failed to set NBD socket");
478 TRACE("Negotiation ended");
483 int nbd_disconnect(int fd
)
485 ioctl(fd
, NBD_CLEAR_QUE
);
486 ioctl(fd
, NBD_DISCONNECT
);
487 ioctl(fd
, NBD_CLEAR_SOCK
);
491 int nbd_client(int fd
)
496 TRACE("Doing NBD loop");
498 ret
= ioctl(fd
, NBD_DO_IT
);
501 TRACE("NBD loop returned %d: %s", ret
, strerror(serrno
));
503 TRACE("Clearing NBD queue");
504 ioctl(fd
, NBD_CLEAR_QUE
);
506 TRACE("Clearing NBD socket");
507 ioctl(fd
, NBD_CLEAR_SOCK
);
513 int nbd_init(int fd
, int csock
, off_t size
, size_t blocksize
)
519 int nbd_disconnect(int fd
)
525 int nbd_client(int fd
)
532 int nbd_send_request(int csock
, struct nbd_request
*request
)
534 uint8_t buf
[4 + 4 + 8 + 8 + 4];
536 cpu_to_be32w((uint32_t*)buf
, NBD_REQUEST_MAGIC
);
537 cpu_to_be32w((uint32_t*)(buf
+ 4), request
->type
);
538 cpu_to_be64w((uint64_t*)(buf
+ 8), request
->handle
);
539 cpu_to_be64w((uint64_t*)(buf
+ 16), request
->from
);
540 cpu_to_be32w((uint32_t*)(buf
+ 24), request
->len
);
542 TRACE("Sending request to client");
544 if (write_sync(csock
, buf
, sizeof(buf
)) != sizeof(buf
)) {
545 LOG("writing to socket failed");
553 static int nbd_receive_request(int csock
, struct nbd_request
*request
)
555 uint8_t buf
[4 + 4 + 8 + 8 + 4];
558 if (read_sync(csock
, buf
, sizeof(buf
)) != sizeof(buf
)) {
565 [ 0 .. 3] magic (NBD_REQUEST_MAGIC)
566 [ 4 .. 7] type (0 == READ, 1 == WRITE)
572 magic
= be32_to_cpup((uint32_t*)buf
);
573 request
->type
= be32_to_cpup((uint32_t*)(buf
+ 4));
574 request
->handle
= be64_to_cpup((uint64_t*)(buf
+ 8));
575 request
->from
= be64_to_cpup((uint64_t*)(buf
+ 16));
576 request
->len
= be32_to_cpup((uint32_t*)(buf
+ 24));
578 TRACE("Got request: "
579 "{ magic = 0x%x, .type = %d, from = %" PRIu64
" , len = %u }",
580 magic
, request
->type
, request
->from
, request
->len
);
582 if (magic
!= NBD_REQUEST_MAGIC
) {
583 LOG("invalid magic (got 0x%x)", magic
);
590 int nbd_receive_reply(int csock
, struct nbd_reply
*reply
)
592 uint8_t buf
[NBD_REPLY_SIZE
];
595 memset(buf
, 0xAA, sizeof(buf
));
597 if (read_sync(csock
, buf
, sizeof(buf
)) != sizeof(buf
)) {
604 [ 0 .. 3] magic (NBD_REPLY_MAGIC)
605 [ 4 .. 7] error (0 == no error)
609 magic
= be32_to_cpup((uint32_t*)buf
);
610 reply
->error
= be32_to_cpup((uint32_t*)(buf
+ 4));
611 reply
->handle
= be64_to_cpup((uint64_t*)(buf
+ 8));
614 "{ magic = 0x%x, .error = %d, handle = %" PRIu64
" }",
615 magic
, reply
->error
, reply
->handle
);
617 if (magic
!= NBD_REPLY_MAGIC
) {
618 LOG("invalid magic (got 0x%x)", magic
);
625 static int nbd_send_reply(int csock
, struct nbd_reply
*reply
)
627 uint8_t buf
[4 + 4 + 8];
630 [ 0 .. 3] magic (NBD_REPLY_MAGIC)
631 [ 4 .. 7] error (0 == no error)
634 cpu_to_be32w((uint32_t*)buf
, NBD_REPLY_MAGIC
);
635 cpu_to_be32w((uint32_t*)(buf
+ 4), reply
->error
);
636 cpu_to_be64w((uint64_t*)(buf
+ 8), reply
->handle
);
638 TRACE("Sending response to client");
640 if (write_sync(csock
, buf
, sizeof(buf
)) != sizeof(buf
)) {
641 LOG("writing to socket failed");
648 int nbd_trip(BlockDriverState
*bs
, int csock
, off_t size
, uint64_t dev_offset
,
649 off_t
*offset
, bool readonly
, uint8_t *data
, int data_size
)
651 struct nbd_request request
;
652 struct nbd_reply reply
;
654 TRACE("Reading request.");
656 if (nbd_receive_request(csock
, &request
) == -1)
659 if (request
.len
+ NBD_REPLY_SIZE
> data_size
) {
660 LOG("len (%u) is larger than max len (%u)",
661 request
.len
+ NBD_REPLY_SIZE
, data_size
);
666 if ((request
.from
+ request
.len
) < request
.from
) {
667 LOG("integer overflow detected! "
668 "you're probably being attacked");
673 if ((request
.from
+ request
.len
) > size
) {
674 LOG("From: %" PRIu64
", Len: %u, Size: %" PRIu64
675 ", Offset: %" PRIu64
"\n",
676 request
.from
, request
.len
, (uint64_t)size
, dev_offset
);
677 LOG("requested operation past EOF--bad client?");
682 TRACE("Decoding type");
684 reply
.handle
= request
.handle
;
687 switch (request
.type
) {
689 TRACE("Request type is READ");
691 if (bdrv_read(bs
, (request
.from
+ dev_offset
) / 512,
692 data
+ NBD_REPLY_SIZE
,
693 request
.len
/ 512) == -1) {
694 LOG("reading from file failed");
698 *offset
+= request
.len
;
700 TRACE("Read %u byte(s)", request
.len
);
703 [ 0 .. 3] magic (NBD_REPLY_MAGIC)
704 [ 4 .. 7] error (0 == no error)
708 cpu_to_be32w((uint32_t*)data
, NBD_REPLY_MAGIC
);
709 cpu_to_be32w((uint32_t*)(data
+ 4), reply
.error
);
710 cpu_to_be64w((uint64_t*)(data
+ 8), reply
.handle
);
712 TRACE("Sending data to client");
714 if (write_sync(csock
, data
,
715 request
.len
+ NBD_REPLY_SIZE
) !=
716 request
.len
+ NBD_REPLY_SIZE
) {
717 LOG("writing to socket failed");
723 TRACE("Request type is WRITE");
725 TRACE("Reading %u byte(s)", request
.len
);
727 if (read_sync(csock
, data
, request
.len
) != request
.len
) {
728 LOG("reading from socket failed");
734 TRACE("Server is read-only, return error");
737 TRACE("Writing to device");
739 if (bdrv_write(bs
, (request
.from
+ dev_offset
) / 512,
740 data
, request
.len
/ 512) == -1) {
741 LOG("writing to file failed");
746 *offset
+= request
.len
;
749 if (nbd_send_reply(csock
, &reply
) == -1)
753 TRACE("Request type is DISCONNECT");
757 LOG("invalid request type (%u) received", request
.type
);
762 TRACE("Request/Reply complete");