block/nbd-client.c

   1 /*
   2  * QEMU Block driver for  NBD
   3  *
   4  * Copyright (C) 2008 Bull S.A.S.
   5  *     Author: Laurent Vivier <Laurent.Vivier@bull.net>
   6  *
   7  * Some parts:
   8  *    Copyright (C) 2007 Anthony Liguori <anthony@codemonkey.ws>
   9  *
  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:
  16  *
  17  * The above copyright notice and this permission notice shall be included in
  18  * all copies or substantial portions of the Software.
  19  *
  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
  26  * THE SOFTWARE.
  27  */
  28
  29 #include "nbd-client.h"
  30 #include "qemu/sockets.h"
  31
  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))
  34
  35 static void nbd_recv_coroutines_enter_all(NbdClientSession *s)
  36 {
  37     int i;
  38
  39     for (i = 0; i < MAX_NBD_REQUESTS; i++) {
  40         if (s->recv_coroutine[i]) {
  41             qemu_coroutine_enter(s->recv_coroutine[i], NULL);
  42         }
  43     }
  44 }
  45
  46 static void nbd_reply_ready(void *opaque)
  47 {
  48     NbdClientSession *s = opaque;
  49     uint64_t i;
  50     int ret;
  51
  52     if (s->reply.handle == 0) {
  53         /* No reply already in flight.  Fetch a header.  It is possible
  54          * that another thread has done the same thing in parallel, so
  55          * the socket is not readable anymore.
  56          */
  57         ret = nbd_receive_reply(s->sock, &s->reply);
  58         if (ret == -EAGAIN) {
  59             return;
  60         }
  61         if (ret < 0) {
  62             s->reply.handle = 0;
  63             goto fail;
  64         }
  65     }
  66
  67     /* There's no need for a mutex on the receive side, because the
  68      * handler acts as a synchronization point and ensures that only
  69      * one coroutine is called until the reply finishes.  */
  70     i = HANDLE_TO_INDEX(s, s->reply.handle);
  71     if (i >= MAX_NBD_REQUESTS) {
  72         goto fail;
  73     }
  74
  75     if (s->recv_coroutine[i]) {
  76         qemu_coroutine_enter(s->recv_coroutine[i], NULL);
  77         return;
  78     }
  79
  80 fail:
  81     nbd_recv_coroutines_enter_all(s);
  82 }
  83
  84 static void nbd_restart_write(void *opaque)
  85 {
  86     NbdClientSession *s = opaque;
  87
  88     qemu_coroutine_enter(s->send_coroutine, NULL);
  89 }
  90
  91 static int nbd_co_send_request(NbdClientSession *s,
  92     struct nbd_request *request,
  93     QEMUIOVector *qiov, int offset)
  94 {
  95     int rc, ret;
  96
  97     qemu_co_mutex_lock(&s->send_mutex);
  98     s->send_coroutine = qemu_coroutine_self();
  99     qemu_aio_set_fd_handler(s->sock, nbd_reply_ready, nbd_restart_write, s);
 100     if (qiov) {
 101         if (!s->is_unix) {
 102             socket_set_cork(s->sock, 1);
 103         }
 104         rc = nbd_send_request(s->sock, request);
 105         if (rc >= 0) {
 106             ret = qemu_co_sendv(s->sock, qiov->iov, qiov->niov,
 107                                 offset, request->len);
 108             if (ret != request->len) {
 109                 rc = -EIO;
 110             }
 111         }
 112         if (!s->is_unix) {
 113             socket_set_cork(s->sock, 0);
 114         }
 115     } else {
 116         rc = nbd_send_request(s->sock, request);
 117     }
 118     qemu_aio_set_fd_handler(s->sock, nbd_reply_ready, NULL, s);
 119     s->send_coroutine = NULL;
 120     qemu_co_mutex_unlock(&s->send_mutex);
 121     return rc;
 122 }
 123
 124 static void nbd_co_receive_reply(NbdClientSession *s,
 125     struct nbd_request *request, struct nbd_reply *reply,
 126     QEMUIOVector *qiov, int offset)
 127 {
 128     int ret;
 129
 130     /* Wait until we're woken up by the read handler.  TODO: perhaps
 131      * peek at the next reply and avoid yielding if it's ours?  */
 132     qemu_coroutine_yield();
 133     *reply = s->reply;
 134     if (reply->handle != request->handle) {
 135         reply->error = EIO;
 136     } else {
 137         if (qiov && reply->error == 0) {
 138             ret = qemu_co_recvv(s->sock, qiov->iov, qiov->niov,
 139                                 offset, request->len);
 140             if (ret != request->len) {
 141                 reply->error = EIO;
 142             }
 143         }
 144
 145         /* Tell the read handler to read another header.  */
 146         s->reply.handle = 0;
 147     }
 148 }
 149
 150 static void nbd_coroutine_start(NbdClientSession *s,
 151    struct nbd_request *request)
 152 {
 153     int i;
 154
 155     /* Poor man semaphore.  The free_sema is locked when no other request
 156      * can be accepted, and unlocked after receiving one reply.  */
 157     if (s->in_flight >= MAX_NBD_REQUESTS - 1) {
 158         qemu_co_mutex_lock(&s->free_sema);
 159         assert(s->in_flight < MAX_NBD_REQUESTS);
 160     }
 161     s->in_flight++;
 162
 163     for (i = 0; i < MAX_NBD_REQUESTS; i++) {
 164         if (s->recv_coroutine[i] == NULL) {
 165             s->recv_coroutine[i] = qemu_coroutine_self();
 166             break;
 167         }
 168     }
 169
 170     assert(i < MAX_NBD_REQUESTS);
 171     request->handle = INDEX_TO_HANDLE(s, i);
 172 }
 173
 174 static void nbd_coroutine_end(NbdClientSession *s,
 175     struct nbd_request *request)
 176 {
 177     int i = HANDLE_TO_INDEX(s, request->handle);
 178     s->recv_coroutine[i] = NULL;
 179     if (s->in_flight-- == MAX_NBD_REQUESTS) {
 180         qemu_co_mutex_unlock(&s->free_sema);
 181     }
 182 }
 183
 184 static int nbd_co_readv_1(NbdClientSession *client, int64_t sector_num,
 185                           int nb_sectors, QEMUIOVector *qiov,
 186                           int offset)
 187 {
 188     struct nbd_request request = { .type = NBD_CMD_READ };
 189     struct nbd_reply reply;
 190     ssize_t ret;
 191
 192     request.from = sector_num * 512;
 193     request.len = nb_sectors * 512;
 194
 195     nbd_coroutine_start(client, &request);
 196     ret = nbd_co_send_request(client, &request, NULL, 0);
 197     if (ret < 0) {
 198         reply.error = -ret;
 199     } else {
 200         nbd_co_receive_reply(client, &request, &reply, qiov, offset);
 201     }
 202     nbd_coroutine_end(client, &request);
 203     return -reply.error;
 204
 205 }
 206
 207 static int nbd_co_writev_1(NbdClientSession *client, int64_t sector_num,
 208                            int nb_sectors, QEMUIOVector *qiov,
 209                            int offset)
 210 {
 211     struct nbd_request request = { .type = NBD_CMD_WRITE };
 212     struct nbd_reply reply;
 213     ssize_t ret;
 214
 215     if (!bdrv_enable_write_cache(client->bs) &&
 216         (client->nbdflags & NBD_FLAG_SEND_FUA)) {
 217         request.type |= NBD_CMD_FLAG_FUA;
 218     }
 219
 220     request.from = sector_num * 512;
 221     request.len = nb_sectors * 512;
 222
 223     nbd_coroutine_start(client, &request);
 224     ret = nbd_co_send_request(client, &request, qiov, offset);
 225     if (ret < 0) {
 226         reply.error = -ret;
 227     } else {
 228         nbd_co_receive_reply(client, &request, &reply, NULL, 0);
 229     }
 230     nbd_coroutine_end(client, &request);
 231     return -reply.error;
 232 }
 233
 234 /* qemu-nbd has a limit of slightly less than 1M per request.  Try to
 235  * remain aligned to 4K. */
 236 #define NBD_MAX_SECTORS 2040
 237
 238 int nbd_client_session_co_readv(NbdClientSession *client, int64_t sector_num,
 239     int nb_sectors, QEMUIOVector *qiov)
 240 {
 241     int offset = 0;
 242     int ret;
 243     while (nb_sectors > NBD_MAX_SECTORS) {
 244         ret = nbd_co_readv_1(client, sector_num,
 245                              NBD_MAX_SECTORS, qiov, offset);
 246         if (ret < 0) {
 247             return ret;
 248         }
 249         offset += NBD_MAX_SECTORS * 512;
 250         sector_num += NBD_MAX_SECTORS;
 251         nb_sectors -= NBD_MAX_SECTORS;
 252     }
 253     return nbd_co_readv_1(client, sector_num, nb_sectors, qiov, offset);
 254 }
 255
 256 int nbd_client_session_co_writev(NbdClientSession *client, int64_t sector_num,
 257                                  int nb_sectors, QEMUIOVector *qiov)
 258 {
 259     int offset = 0;
 260     int ret;
 261     while (nb_sectors > NBD_MAX_SECTORS) {
 262         ret = nbd_co_writev_1(client, sector_num,
 263                               NBD_MAX_SECTORS, qiov, offset);
 264         if (ret < 0) {
 265             return ret;
 266         }
 267         offset += NBD_MAX_SECTORS * 512;
 268         sector_num += NBD_MAX_SECTORS;
 269         nb_sectors -= NBD_MAX_SECTORS;
 270     }
 271     return nbd_co_writev_1(client, sector_num, nb_sectors, qiov, offset);
 272 }
 273
 274 int nbd_client_session_co_flush(NbdClientSession *client)
 275 {
 276     struct nbd_request request = { .type = NBD_CMD_FLUSH };
 277     struct nbd_reply reply;
 278     ssize_t ret;
 279
 280     if (!(client->nbdflags & NBD_FLAG_SEND_FLUSH)) {
 281         return 0;
 282     }
 283
 284     if (client->nbdflags & NBD_FLAG_SEND_FUA) {
 285         request.type |= NBD_CMD_FLAG_FUA;
 286     }
 287
 288     request.from = 0;
 289     request.len = 0;
 290
 291     nbd_coroutine_start(client, &request);
 292     ret = nbd_co_send_request(client, &request, NULL, 0);
 293     if (ret < 0) {
 294         reply.error = -ret;
 295     } else {
 296         nbd_co_receive_reply(client, &request, &reply, NULL, 0);
 297     }
 298     nbd_coroutine_end(client, &request);
 299     return -reply.error;
 300 }
 301
 302 int nbd_client_session_co_discard(NbdClientSession *client, int64_t sector_num,
 303     int nb_sectors)
 304 {
 305     struct nbd_request request = { .type = NBD_CMD_TRIM };
 306     struct nbd_reply reply;
 307     ssize_t ret;
 308
 309     if (!(client->nbdflags & NBD_FLAG_SEND_TRIM)) {
 310         return 0;
 311     }
 312     request.from = sector_num * 512;
 313     request.len = nb_sectors * 512;
 314
 315     nbd_coroutine_start(client, &request);
 316     ret = nbd_co_send_request(client, &request, NULL, 0);
 317     if (ret < 0) {
 318         reply.error = -ret;
 319     } else {
 320         nbd_co_receive_reply(client, &request, &reply, NULL, 0);
 321     }
 322     nbd_coroutine_end(client, &request);
 323     return -reply.error;
 324
 325 }
 326
 327 static void nbd_teardown_connection(NbdClientSession *client)
 328 {
 329     struct nbd_request request = {
 330         .type = NBD_CMD_DISC,
 331         .from = 0,
 332         .len = 0
 333     };
 334
 335     nbd_send_request(client->sock, &request);
 336
 337     /* finish any pending coroutines */
 338     shutdown(client->sock, 2);
 339     nbd_recv_coroutines_enter_all(client);
 340
 341     qemu_aio_set_fd_handler(client->sock, NULL, NULL, NULL);
 342     closesocket(client->sock);
 343     client->sock = -1;
 344 }
 345
 346 void nbd_client_session_close(NbdClientSession *client)
 347 {
 348     if (!client->bs) {
 349         return;
 350     }
 351
 352     nbd_teardown_connection(client);
 353     client->bs = NULL;
 354 }
 355
 356 int nbd_client_session_init(NbdClientSession *client, BlockDriverState *bs,
 357     int sock, const char *export)
 358 {
 359     int ret;
 360
 361     /* NBD handshake */
 362     logout("session init %s\n", export);
 363     qemu_set_block(sock);
 364     ret = nbd_receive_negotiate(sock, export,
 365                                 &client->nbdflags, &client->size,
 366                                 &client->blocksize);
 367     if (ret < 0) {
 368         logout("Failed to negotiate with the NBD server\n");
 369         closesocket(sock);
 370         return ret;
 371     }
 372
 373     qemu_co_mutex_init(&client->send_mutex);
 374     qemu_co_mutex_init(&client->free_sema);
 375     client->bs = bs;
 376     client->sock = sock;
 377
 378     /* Now that we're connected, set the socket to be non-blocking and
 379      * kick the reply mechanism.  */
 380     qemu_set_nonblock(sock);
 381     qemu_aio_set_fd_handler(sock, nbd_reply_ready, NULL, client);
 382
 383     logout("Established connection with NBD server\n");
 384     return 0;
 385 }