2 * QEMU Block driver for NBD
4 * Copyright (C) 2008 Bull S.A.S.
10 * Permission is hereby granted, free of charge, to any person obtaining a copy
11 * of this software and associated documentation files (the "Software"), to deal
12 * in the Software without restriction, including without limitation the rights
13 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
14 * copies of the Software, and to permit persons to whom the Software is
15 * furnished to do so, subject to the following conditions:
17 * The above copyright notice and this permission notice shall be included in
18 * all copies or substantial portions of the Software.
20 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
21 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
23 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
24 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
25 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
29 #include "nbd-client.h"
30 #include "qemu/sockets.h"
32 #define HANDLE_TO_INDEX(bs, handle) ((handle) ^ ((uint64_t)(intptr_t)bs))
33 #define INDEX_TO_HANDLE(bs, index) ((index) ^ ((uint64_t)(intptr_t)bs))
35 static void nbd_recv_coroutines_enter_all(NbdClientSession *s)
39 for (i = 0; i < MAX_NBD_REQUESTS; i++) {
40 if (s->recv_coroutine[i]) {
41 qemu_coroutine_enter(s->recv_coroutine[i], NULL);
46 static void nbd_teardown_connection(NbdClientSession *client)
48 /* finish any pending coroutines */
49 shutdown(client->sock, 2);
50 nbd_recv_coroutines_enter_all(client);
52 qemu_aio_set_fd_handler(client->sock, NULL, NULL, NULL);
53 closesocket(client->sock);
57 static void nbd_reply_ready(void *opaque)
59 NbdClientSession *s = opaque;
63 if (s->reply.handle == 0) {
64 /* No reply already in flight. Fetch a header. It is possible
65 * that another thread has done the same thing in parallel, so
66 * the socket is not readable anymore.
68 ret = nbd_receive_reply(s->sock, &s->reply);
78 /* There's no need for a mutex on the receive side, because the
79 * handler acts as a synchronization point and ensures that only
80 * one coroutine is called until the reply finishes. */
81 i = HANDLE_TO_INDEX(s, s->reply.handle);
82 if (i >= MAX_NBD_REQUESTS) {
86 if (s->recv_coroutine[i]) {
87 qemu_coroutine_enter(s->recv_coroutine[i], NULL);
92 nbd_teardown_connection(s);
95 static void nbd_restart_write(void *opaque)
97 NbdClientSession *s = opaque;
99 qemu_coroutine_enter(s->send_coroutine, NULL);
102 static int nbd_co_send_request(NbdClientSession *s,
103 struct nbd_request *request,
104 QEMUIOVector *qiov, int offset)
108 qemu_co_mutex_lock(&s->send_mutex);
109 s->send_coroutine = qemu_coroutine_self();
110 qemu_aio_set_fd_handler(s->sock, nbd_reply_ready, nbd_restart_write, s);
113 socket_set_cork(s->sock, 1);
115 rc = nbd_send_request(s->sock, request);
117 ret = qemu_co_sendv(s->sock, qiov->iov, qiov->niov,
118 offset, request->len);
119 if (ret != request->len) {
124 socket_set_cork(s->sock, 0);
127 rc = nbd_send_request(s->sock, request);
129 qemu_aio_set_fd_handler(s->sock, nbd_reply_ready, NULL, s);
130 s->send_coroutine = NULL;
131 qemu_co_mutex_unlock(&s->send_mutex);
135 static void nbd_co_receive_reply(NbdClientSession *s,
136 struct nbd_request *request, struct nbd_reply *reply,
137 QEMUIOVector *qiov, int offset)
141 /* Wait until we're woken up by the read handler. TODO: perhaps
142 * peek at the next reply and avoid yielding if it's ours? */
143 qemu_coroutine_yield();
145 if (reply->handle != request->handle) {
148 if (qiov && reply->error == 0) {
149 ret = qemu_co_recvv(s->sock, qiov->iov, qiov->niov,
150 offset, request->len);
151 if (ret != request->len) {
156 /* Tell the read handler to read another header. */
161 static void nbd_coroutine_start(NbdClientSession *s,
162 struct nbd_request *request)
166 /* Poor man semaphore. The free_sema is locked when no other request
167 * can be accepted, and unlocked after receiving one reply. */
168 if (s->in_flight >= MAX_NBD_REQUESTS - 1) {
169 qemu_co_mutex_lock(&s->free_sema);
170 assert(s->in_flight < MAX_NBD_REQUESTS);
174 for (i = 0; i < MAX_NBD_REQUESTS; i++) {
175 if (s->recv_coroutine[i] == NULL) {
176 s->recv_coroutine[i] = qemu_coroutine_self();
181 assert(i < MAX_NBD_REQUESTS);
182 request->handle = INDEX_TO_HANDLE(s, i);
185 static void nbd_coroutine_end(NbdClientSession *s,
186 struct nbd_request *request)
188 int i = HANDLE_TO_INDEX(s, request->handle);
189 s->recv_coroutine[i] = NULL;
190 if (s->in_flight-- == MAX_NBD_REQUESTS) {
191 qemu_co_mutex_unlock(&s->free_sema);
195 static int nbd_co_readv_1(NbdClientSession *client, int64_t sector_num,
196 int nb_sectors, QEMUIOVector *qiov,
199 struct nbd_request request = { .type = NBD_CMD_READ };
200 struct nbd_reply reply;
203 request.from = sector_num * 512;
204 request.len = nb_sectors * 512;
206 nbd_coroutine_start(client, &request);
207 ret = nbd_co_send_request(client, &request, NULL, 0);
211 nbd_co_receive_reply(client, &request, &reply, qiov, offset);
213 nbd_coroutine_end(client, &request);
218 static int nbd_co_writev_1(NbdClientSession *client, int64_t sector_num,
219 int nb_sectors, QEMUIOVector *qiov,
222 struct nbd_request request = { .type = NBD_CMD_WRITE };
223 struct nbd_reply reply;
226 if (!bdrv_enable_write_cache(client->bs) &&
227 (client->nbdflags & NBD_FLAG_SEND_FUA)) {
228 request.type |= NBD_CMD_FLAG_FUA;
231 request.from = sector_num * 512;
232 request.len = nb_sectors * 512;
234 nbd_coroutine_start(client, &request);
235 ret = nbd_co_send_request(client, &request, qiov, offset);
239 nbd_co_receive_reply(client, &request, &reply, NULL, 0);
241 nbd_coroutine_end(client, &request);
245 /* qemu-nbd has a limit of slightly less than 1M per request. Try to
246 * remain aligned to 4K. */
247 #define NBD_MAX_SECTORS 2040
249 int nbd_client_session_co_readv(NbdClientSession *client, int64_t sector_num,
250 int nb_sectors, QEMUIOVector *qiov)
254 while (nb_sectors > NBD_MAX_SECTORS) {
255 ret = nbd_co_readv_1(client, sector_num,
256 NBD_MAX_SECTORS, qiov, offset);
260 offset += NBD_MAX_SECTORS * 512;
261 sector_num += NBD_MAX_SECTORS;
262 nb_sectors -= NBD_MAX_SECTORS;
264 return nbd_co_readv_1(client, sector_num, nb_sectors, qiov, offset);
267 int nbd_client_session_co_writev(NbdClientSession *client, int64_t sector_num,
268 int nb_sectors, QEMUIOVector *qiov)
272 while (nb_sectors > NBD_MAX_SECTORS) {
273 ret = nbd_co_writev_1(client, sector_num,
274 NBD_MAX_SECTORS, qiov, offset);
278 offset += NBD_MAX_SECTORS * 512;
279 sector_num += NBD_MAX_SECTORS;
280 nb_sectors -= NBD_MAX_SECTORS;
282 return nbd_co_writev_1(client, sector_num, nb_sectors, qiov, offset);
285 int nbd_client_session_co_flush(NbdClientSession *client)
287 struct nbd_request request = { .type = NBD_CMD_FLUSH };
288 struct nbd_reply reply;
291 if (!(client->nbdflags & NBD_FLAG_SEND_FLUSH)) {
295 if (client->nbdflags & NBD_FLAG_SEND_FUA) {
296 request.type |= NBD_CMD_FLAG_FUA;
302 nbd_coroutine_start(client, &request);
303 ret = nbd_co_send_request(client, &request, NULL, 0);
307 nbd_co_receive_reply(client, &request, &reply, NULL, 0);
309 nbd_coroutine_end(client, &request);
313 int nbd_client_session_co_discard(NbdClientSession *client, int64_t sector_num,
316 struct nbd_request request = { .type = NBD_CMD_TRIM };
317 struct nbd_reply reply;
320 if (!(client->nbdflags & NBD_FLAG_SEND_TRIM)) {
323 request.from = sector_num * 512;
324 request.len = nb_sectors * 512;
326 nbd_coroutine_start(client, &request);
327 ret = nbd_co_send_request(client, &request, NULL, 0);
331 nbd_co_receive_reply(client, &request, &reply, NULL, 0);
333 nbd_coroutine_end(client, &request);
338 void nbd_client_session_close(NbdClientSession *client)
340 struct nbd_request request = {
341 .type = NBD_CMD_DISC,
349 if (client->sock == -1) {
353 nbd_send_request(client->sock, &request);
355 nbd_teardown_connection(client);
359 int nbd_client_session_init(NbdClientSession *client, BlockDriverState *bs,
360 int sock, const char *export)
365 logout("session init %s\n", export);
366 qemu_set_block(sock);
367 ret = nbd_receive_negotiate(sock, export,
368 &client->nbdflags, &client->size,
371 logout("Failed to negotiate with the NBD server\n");
376 qemu_co_mutex_init(&client->send_mutex);
377 qemu_co_mutex_init(&client->free_sema);
381 /* Now that we're connected, set the socket to be non-blocking and
382 * kick the reply mechanism. */
383 qemu_set_nonblock(sock);
384 qemu_aio_set_fd_handler(sock, nbd_reply_ready, NULL, client);
386 logout("Established connection with NBD server\n");