Add vgabios build rules to roms/Makefile

[qemu.git] / nbd.c

diff --git a/nbd.c b/nbd.c
index a9f295f2bbbc20e58802b1ed0d76420490537114..dc0adf90ed1e77dff41ea1b221cb63b967d0c973 100644 (file)
--- a/nbd.c
+++ b/nbd.c
@@ -17,19 +17,27 @@
  */
 
 #include "nbd.h"
  */
 
 #include "nbd.h"

+#include "block.h"
+
+#include "qemu-coroutine.h"

 
 #include <errno.h>
 #include <string.h>
 #ifndef _WIN32
 #include <sys/ioctl.h>
 #endif
 
 #include <errno.h>
 #include <string.h>
 #ifndef _WIN32
 #include <sys/ioctl.h>
 #endif

-#ifdef __sun__
+#if defined(__sun__) || defined(__HAIKU__)

 #include <sys/ioccom.h>
 #endif
 #include <ctype.h>
 #include <inttypes.h>
 
 #include <sys/ioccom.h>
 #endif
 #include <ctype.h>
 #include <inttypes.h>
 

+#ifdef __linux__
+#include <linux/fs.h>
+#endif
+

 #include "qemu_socket.h"
 #include "qemu_socket.h"

+#include "qemu-queue.h"

 
 //#define DEBUG_NBD
 
 
 //#define DEBUG_NBD
 


@@ -49,6 +57,7 @@
 
 /* This is all part of the "official" NBD API */
 
 
 /* This is all part of the "official" NBD API */
 

+#define NBD_REPLY_SIZE          (4 + 4 + 8)

 #define NBD_REQUEST_MAGIC       0x25609513
 #define NBD_REPLY_MAGIC         0x67446698
 
 #define NBD_REQUEST_MAGIC       0x25609513
 #define NBD_REPLY_MAGIC         0x67446698
 


@@ -58,34 +67,48 @@
 #define NBD_DO_IT               _IO(0xab, 3)
 #define NBD_CLEAR_SOCK          _IO(0xab, 4)
 #define NBD_CLEAR_QUE           _IO(0xab, 5)
 #define NBD_DO_IT               _IO(0xab, 3)
 #define NBD_CLEAR_SOCK          _IO(0xab, 4)
 #define NBD_CLEAR_QUE           _IO(0xab, 5)

-#define NBD_PRINT_DEBUG                _IO(0xab, 6)
-#define NBD_SET_SIZE_BLOCKS    _IO(0xab, 7)
+#define NBD_PRINT_DEBUG         _IO(0xab, 6)
+#define NBD_SET_SIZE_BLOCKS     _IO(0xab, 7)

 #define NBD_DISCONNECT          _IO(0xab, 8)
 #define NBD_DISCONNECT          _IO(0xab, 8)

+#define NBD_SET_TIMEOUT         _IO(0xab, 9)
+#define NBD_SET_FLAGS           _IO(0xab, 10)

 
 

-/* That's all folks */
+#define NBD_OPT_EXPORT_NAME     (1 << 0)

 
 

-#define read_sync(fd, buffer, size) nbd_wr_sync(fd, buffer, size, true)
-#define write_sync(fd, buffer, size) nbd_wr_sync(fd, buffer, size, false)
+/* That's all folks */

 
 

-size_t nbd_wr_sync(int fd, void *buffer, size_t size, bool do_read)
+ssize_t nbd_wr_sync(int fd, void *buffer, size_t size, bool do_read)

 {
     size_t offset = 0;
 {
     size_t offset = 0;

+    int err;
+
+    if (qemu_in_coroutine()) {
+        if (do_read) {
+            return qemu_co_recv(fd, buffer, size);
+        } else {
+            return qemu_co_send(fd, buffer, size);
+        }
+    }

 
     while (offset < size) {
         ssize_t len;
 
         if (do_read) {
 
     while (offset < size) {
         ssize_t len;
 
         if (do_read) {

-            len = recv(fd, buffer + offset, size - offset, 0);
+            len = qemu_recv(fd, buffer + offset, size - offset, 0);

         } else {
             len = send(fd, buffer + offset, size - offset, 0);
         }
 
         } else {
             len = send(fd, buffer + offset, size - offset, 0);
         }
 

-        if (len == -1)
-            errno = socket_error();
+        if (len < 0) {
+            err = socket_error();

 
 

-        /* recoverable error */
-        if (len == -1 && (errno == EAGAIN || errno == EINTR)) {
-            continue;
+            /* recoverable error */
+            if (err == EINTR || (offset > 0 && err == EAGAIN)) {
+                continue;
+            }
+
+            /* unrecoverable error */
+            return -err;

         }
 
         /* eof */
         }
 
         /* eof */


@@ -93,568 +116,873 @@ size_t nbd_wr_sync(int fd, void *buffer, size_t size, bool do_read)
             break;
         }
 
             break;
         }
 

-        /* unrecoverable error */
-        if (len == -1) {
-            return 0;
-        }
-

         offset += len;
     }
 
     return offset;
 }
 
         offset += len;
     }
 
     return offset;
 }
 

-int tcp_socket_outgoing(const char *address, uint16_t port)
+static ssize_t read_sync(int fd, void *buffer, size_t size)
+{
+    /* Sockets are kept in blocking mode in the negotiation phase.  After
+     * that, a non-readable socket simply means that another thread stole
+     * our request/reply.  Synchronization is done with recv_coroutine, so
+     * that this is coroutine-safe.
+     */
+    return nbd_wr_sync(fd, buffer, size, true);
+}
+
+static ssize_t write_sync(int fd, void *buffer, size_t size)

 {
 {

-    int s;
-    struct in_addr in;
-    struct sockaddr_in addr;
+    int ret;
+    do {
+        /* For writes, we do expect the socket to be writable.  */
+        ret = nbd_wr_sync(fd, buffer, size, false);
+    } while (ret == -EAGAIN);
+    return ret;
+}

 
 

-    s = socket(PF_INET, SOCK_STREAM, 0);
-    if (s == -1) {
-        return -1;
+static void combine_addr(char *buf, size_t len, const char* address,
+                         uint16_t port)
+{
+    /* If the address-part contains a colon, it's an IPv6 IP so needs [] */
+    if (strstr(address, ":")) {
+        snprintf(buf, len, "[%s]:%u", address, port);
+    } else {
+        snprintf(buf, len, "%s:%u", address, port);

     }
     }

+}

 
 

-    if (inet_aton(address, &in) == 0) {
-        struct hostent *ent;
+int tcp_socket_outgoing(const char *address, uint16_t port)
+{
+    char address_and_port[128];
+    combine_addr(address_and_port, 128, address, port);
+    return tcp_socket_outgoing_spec(address_and_port);
+}

 
 

-        ent = gethostbyname(address);
-        if (ent == NULL) {
-            goto error;
-        }
+int tcp_socket_outgoing_spec(const char *address_and_port)
+{
+    return inet_connect(address_and_port, true, NULL);
+}

 
 

-        memcpy(&in, ent->h_addr, sizeof(in));
-    }
+int tcp_socket_incoming(const char *address, uint16_t port)
+{
+    char address_and_port[128];
+    combine_addr(address_and_port, 128, address, port);
+    return tcp_socket_incoming_spec(address_and_port);
+}

 
 

-    addr.sin_family = AF_INET;
-    addr.sin_port = htons(port);
-    memcpy(&addr.sin_addr.s_addr, &in, sizeof(in));
+int tcp_socket_incoming_spec(const char *address_and_port)
+{
+    char *ostr  = NULL;
+    int olen = 0;
+    return inet_listen(address_and_port, ostr, olen, SOCK_STREAM, 0, NULL);
+}

 
 

-    if (connect(s, (struct sockaddr *)&addr, sizeof(addr)) == -1) {
-        goto error;
-    }
+int unix_socket_incoming(const char *path)
+{
+    char *ostr = NULL;
+    int olen = 0;

 
 

-    return s;
-error:
-    closesocket(s);
-    return -1;
+    return unix_listen(path, ostr, olen);

 }
 
 }
 

-int tcp_socket_incoming(const char *address, uint16_t port)
+int unix_socket_outgoing(const char *path)

 {
 {

-    int s;
-    struct in_addr in;
-    struct sockaddr_in addr;
-    int opt;
+    return unix_connect(path);
+}
+
+/* Basic flow
+
+   Server         Client
+
+   Negotiate
+                  Request
+   Response
+                  Request
+   Response
+                  ...
+   ...
+                  Request (type == 2)
+*/

 
 

-    s = socket(PF_INET, SOCK_STREAM, 0);
-    if (s == -1) {
-        return -1;
+static int nbd_send_negotiate(int csock, off_t size, uint32_t flags)
+{
+    char buf[8 + 8 + 8 + 128];
+    int rc;
+
+    /* Negotiate
+        [ 0 ..   7]   passwd   ("NBDMAGIC")
+        [ 8 ..  15]   magic    (0x00420281861253)
+        [16 ..  23]   size
+        [24 ..  27]   flags
+        [28 .. 151]   reserved (0)
+     */
+
+    socket_set_block(csock);
+    rc = -EINVAL;
+
+    TRACE("Beginning negotiation.");
+    memcpy(buf, "NBDMAGIC", 8);
+    cpu_to_be64w((uint64_t*)(buf + 8), 0x00420281861253LL);
+    cpu_to_be64w((uint64_t*)(buf + 16), size);
+    cpu_to_be32w((uint32_t*)(buf + 24),
+                 flags | NBD_FLAG_HAS_FLAGS | NBD_FLAG_SEND_TRIM |
+                 NBD_FLAG_SEND_FLUSH | NBD_FLAG_SEND_FUA);
+    memset(buf + 28, 0, 124);
+
+    if (write_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
+        LOG("write failed");
+        goto fail;

     }
 
     }
 

-    if (inet_aton(address, &in) == 0) {
-        struct hostent *ent;
+    TRACE("Negotiation succeeded.");
+    rc = 0;
+fail:
+    socket_set_nonblock(csock);
+    return rc;
+}

 
 

-        ent = gethostbyname(address);
-        if (ent == NULL) {
-            goto error;
-        }
+int nbd_receive_negotiate(int csock, const char *name, uint32_t *flags,
+                          off_t *size, size_t *blocksize)
+{
+    char buf[256];
+    uint64_t magic, s;
+    uint16_t tmp;
+    int rc;

 
 

-        memcpy(&in, ent->h_addr, sizeof(in));
+    TRACE("Receiving negotiation.");
+
+    socket_set_block(csock);
+    rc = -EINVAL;
+
+    if (read_sync(csock, buf, 8) != 8) {
+        LOG("read failed");
+        goto fail;

     }
 
     }
 

-    addr.sin_family = AF_INET;
-    addr.sin_port = htons(port);
-    memcpy(&addr.sin_addr.s_addr, &in, sizeof(in));
+    buf[8] = '\0';
+    if (strlen(buf) == 0) {
+        LOG("server connection closed");
+        goto fail;
+    }

 
 

-    opt = 1;
-    if (setsockopt(s, SOL_SOCKET, SO_REUSEADDR,
-                   (const void *) &opt, sizeof(opt)) == -1) {
-        goto error;
+    TRACE("Magic is %c%c%c%c%c%c%c%c",
+          qemu_isprint(buf[0]) ? buf[0] : '.',
+          qemu_isprint(buf[1]) ? buf[1] : '.',
+          qemu_isprint(buf[2]) ? buf[2] : '.',
+          qemu_isprint(buf[3]) ? buf[3] : '.',
+          qemu_isprint(buf[4]) ? buf[4] : '.',
+          qemu_isprint(buf[5]) ? buf[5] : '.',
+          qemu_isprint(buf[6]) ? buf[6] : '.',
+          qemu_isprint(buf[7]) ? buf[7] : '.');
+
+    if (memcmp(buf, "NBDMAGIC", 8) != 0) {
+        LOG("Invalid magic received");
+        goto fail;

     }
 
     }
 

-    if (bind(s, (struct sockaddr *)&addr, sizeof(addr)) == -1) {
-        goto error;
+    if (read_sync(csock, &magic, sizeof(magic)) != sizeof(magic)) {
+        LOG("read failed");
+        goto fail;
+    }
+    magic = be64_to_cpu(magic);
+    TRACE("Magic is 0x%" PRIx64, magic);
+
+    if (name) {
+        uint32_t reserved = 0;
+        uint32_t opt;
+        uint32_t namesize;
+
+        TRACE("Checking magic (opts_magic)");
+        if (magic != 0x49484156454F5054LL) {
+            LOG("Bad magic received");
+            goto fail;
+        }
+        if (read_sync(csock, &tmp, sizeof(tmp)) != sizeof(tmp)) {
+            LOG("flags read failed");
+            goto fail;
+        }
+        *flags = be16_to_cpu(tmp) << 16;
+        /* reserved for future use */
+        if (write_sync(csock, &reserved, sizeof(reserved)) !=
+            sizeof(reserved)) {
+            LOG("write failed (reserved)");
+            goto fail;
+        }
+        /* write the export name */
+        magic = cpu_to_be64(magic);
+        if (write_sync(csock, &magic, sizeof(magic)) != sizeof(magic)) {
+            LOG("write failed (magic)");
+            goto fail;
+        }
+        opt = cpu_to_be32(NBD_OPT_EXPORT_NAME);
+        if (write_sync(csock, &opt, sizeof(opt)) != sizeof(opt)) {
+            LOG("write failed (opt)");
+            goto fail;
+        }
+        namesize = cpu_to_be32(strlen(name));
+        if (write_sync(csock, &namesize, sizeof(namesize)) !=
+            sizeof(namesize)) {
+            LOG("write failed (namesize)");
+            goto fail;
+        }
+        if (write_sync(csock, (char*)name, strlen(name)) != strlen(name)) {
+            LOG("write failed (name)");
+            goto fail;
+        }
+    } else {
+        TRACE("Checking magic (cli_magic)");
+
+        if (magic != 0x00420281861253LL) {
+            LOG("Bad magic received");
+            goto fail;
+        }

     }
 
     }
 

-    if (listen(s, 128) == -1) {
-        goto error;
+    if (read_sync(csock, &s, sizeof(s)) != sizeof(s)) {
+        LOG("read failed");
+        goto fail;
+    }
+    *size = be64_to_cpu(s);
+    *blocksize = 1024;
+    TRACE("Size is %" PRIu64, *size);
+
+    if (!name) {
+        if (read_sync(csock, flags, sizeof(*flags)) != sizeof(*flags)) {
+            LOG("read failed (flags)");
+            goto fail;
+        }
+        *flags = be32_to_cpup(flags);
+    } else {
+        if (read_sync(csock, &tmp, sizeof(tmp)) != sizeof(tmp)) {
+            LOG("read failed (tmp)");
+            goto fail;
+        }
+        *flags |= be32_to_cpu(tmp);
+    }
+    if (read_sync(csock, &buf, 124) != 124) {
+        LOG("read failed (buf)");
+        goto fail;

     }
     }

+    rc = 0;

 
 

-    return s;
-error:
-    closesocket(s);
-    return -1;
+fail:
+    socket_set_nonblock(csock);
+    return rc;

 }
 
 }
 

-#ifndef _WIN32
-int unix_socket_incoming(const char *path)
+#ifdef __linux__
+int nbd_init(int fd, int csock, uint32_t flags, off_t size, size_t blocksize)

 {
 {

-    int s;
-    struct sockaddr_un addr;
+    TRACE("Setting NBD socket");
+
+    if (ioctl(fd, NBD_SET_SOCK, csock) < 0) {
+        int serrno = errno;
+        LOG("Failed to set NBD socket");
+        return -serrno;
+    }
+
+    TRACE("Setting block size to %lu", (unsigned long)blocksize);
+
+    if (ioctl(fd, NBD_SET_BLKSIZE, blocksize) < 0) {
+        int serrno = errno;
+        LOG("Failed setting NBD block size");
+        return -serrno;
+    }

 
 

-    s = socket(PF_UNIX, SOCK_STREAM, 0);
-    if (s == -1) {
-        return -1;
+        TRACE("Setting size to %zd block(s)", (size_t)(size / blocksize));
+
+    if (ioctl(fd, NBD_SET_SIZE_BLOCKS, size / blocksize) < 0) {
+        int serrno = errno;
+        LOG("Failed setting size (in blocks)");
+        return -serrno;

     }
 
     }
 

-    memset(&addr, 0, sizeof(addr));
-    addr.sun_family = AF_UNIX;
-    pstrcpy(addr.sun_path, sizeof(addr.sun_path), path);
+    if (flags & NBD_FLAG_READ_ONLY) {
+        int read_only = 1;
+        TRACE("Setting readonly attribute");

 
 

-    if (bind(s, (struct sockaddr *)&addr, sizeof(addr)) == -1) {
-        goto error;
+        if (ioctl(fd, BLKROSET, (unsigned long) &read_only) < 0) {
+            int serrno = errno;
+            LOG("Failed setting read-only attribute");
+            return -serrno;
+        }

     }
 
     }
 

-    if (listen(s, 128) == -1) {
-        goto error;
+    if (ioctl(fd, NBD_SET_FLAGS, flags) < 0
+        && errno != ENOTTY) {
+        int serrno = errno;
+        LOG("Failed setting flags");
+        return -serrno;

     }
 
     }
 

-    return s;
-error:
-    closesocket(s);
-    return -1;
+    TRACE("Negotiation ended");
+
+    return 0;

 }
 
 }
 

-int unix_socket_outgoing(const char *path)
+int nbd_disconnect(int fd)

 {
 {

-    int s;
-    struct sockaddr_un addr;
+    ioctl(fd, NBD_CLEAR_QUE);
+    ioctl(fd, NBD_DISCONNECT);
+    ioctl(fd, NBD_CLEAR_SOCK);
+    return 0;
+}

 
 

-    s = socket(PF_UNIX, SOCK_STREAM, 0);
-    if (s == -1) {
-        return -1;
+int nbd_client(int fd)
+{
+    int ret;
+    int serrno;
+
+    TRACE("Doing NBD loop");
+
+    ret = ioctl(fd, NBD_DO_IT);
+    if (ret < 0 && errno == EPIPE) {
+        /* NBD_DO_IT normally returns EPIPE when someone has disconnected
+         * the socket via NBD_DISCONNECT.  We do not want to return 1 in
+         * that case.
+         */
+        ret = 0;

     }
     }

+    serrno = errno;

 
 

-    memset(&addr, 0, sizeof(addr));
-    addr.sun_family = AF_UNIX;
-    pstrcpy(addr.sun_path, sizeof(addr.sun_path), path);
+    TRACE("NBD loop returned %d: %s", ret, strerror(serrno));

 
 

-    if (connect(s, (struct sockaddr *)&addr, sizeof(addr)) == -1) {
-        goto error;
-    }
+    TRACE("Clearing NBD queue");
+    ioctl(fd, NBD_CLEAR_QUE);

 
 

-    return s;
-error:
-    closesocket(s);
-    return -1;
+    TRACE("Clearing NBD socket");
+    ioctl(fd, NBD_CLEAR_SOCK);
+
+    errno = serrno;
+    return ret;

 }
 #else
 }
 #else

-int unix_socket_incoming(const char *path)
+int nbd_init(int fd, int csock, uint32_t flags, off_t size, size_t blocksize)

 {
 {

-    errno = ENOTSUP;
-    return -1;
+    return -ENOTSUP;

 }
 
 }
 

-int unix_socket_outgoing(const char *path)
+int nbd_disconnect(int fd)

 {
 {

-    errno = ENOTSUP;
-    return -1;
+    return -ENOTSUP;

 }
 }

-#endif
-
-
-/* Basic flow
-
-   Server         Client

 
 

-   Negotiate
-                  Request
-   Response
-                  Request
-   Response
-                  ...
-   ...
-                  Request (type == 2)
-*/
+int nbd_client(int fd)
+{
+    return -ENOTSUP;
+}
+#endif

 
 

-int nbd_negotiate(int csock, off_t size)
+ssize_t nbd_send_request(int csock, struct nbd_request *request)

 {
 {

-       char buf[8 + 8 + 8 + 128];
+    uint8_t buf[4 + 4 + 8 + 8 + 4];
+    ssize_t ret;
+
+    cpu_to_be32w((uint32_t*)buf, NBD_REQUEST_MAGIC);
+    cpu_to_be32w((uint32_t*)(buf + 4), request->type);
+    cpu_to_be64w((uint64_t*)(buf + 8), request->handle);
+    cpu_to_be64w((uint64_t*)(buf + 16), request->from);
+    cpu_to_be32w((uint32_t*)(buf + 24), request->len);
+
+    TRACE("Sending request to client: "
+          "{ .from = %" PRIu64", .len = %u, .handle = %" PRIu64", .type=%i}",
+          request->from, request->len, request->handle, request->type);
+
+    ret = write_sync(csock, buf, sizeof(buf));
+    if (ret < 0) {
+        return ret;
+    }

 
 

-       /* Negotiate
-          [ 0 ..   7]   passwd   ("NBDMAGIC")
-          [ 8 ..  15]   magic    (0x00420281861253)
-          [16 ..  23]   size
-          [24 .. 151]   reserved (0)
-        */
+    if (ret != sizeof(buf)) {
+        LOG("writing to socket failed");
+        return -EINVAL;
+    }
+    return 0;
+}

 
 

-       TRACE("Beginning negotiation.");
-       memcpy(buf, "NBDMAGIC", 8);
-       cpu_to_be64w((uint64_t*)(buf + 8), 0x00420281861253LL);
-       cpu_to_be64w((uint64_t*)(buf + 16), size);
-       memset(buf + 24, 0, 128);
+static ssize_t nbd_receive_request(int csock, struct nbd_request *request)
+{
+    uint8_t buf[4 + 4 + 8 + 8 + 4];
+    uint32_t magic;
+    ssize_t ret;

 
 

-       if (write_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
-               LOG("write failed");
-               errno = EINVAL;
-               return -1;
-       }
+    ret = read_sync(csock, buf, sizeof(buf));
+    if (ret < 0) {
+        return ret;
+    }

 
 

-       TRACE("Negotation succeeded.");
+    if (ret != sizeof(buf)) {
+        LOG("read failed");
+        return -EINVAL;
+    }

 
 

-       return 0;
+    /* Request
+       [ 0 ..  3]   magic   (NBD_REQUEST_MAGIC)
+       [ 4 ..  7]   type    (0 == READ, 1 == WRITE)
+       [ 8 .. 15]   handle
+       [16 .. 23]   from
+       [24 .. 27]   len
+     */
+
+    magic = be32_to_cpup((uint32_t*)buf);
+    request->type  = be32_to_cpup((uint32_t*)(buf + 4));
+    request->handle = be64_to_cpup((uint64_t*)(buf + 8));
+    request->from  = be64_to_cpup((uint64_t*)(buf + 16));
+    request->len   = be32_to_cpup((uint32_t*)(buf + 24));
+
+    TRACE("Got request: "
+          "{ magic = 0x%x, .type = %d, from = %" PRIu64" , len = %u }",
+          magic, request->type, request->from, request->len);
+
+    if (magic != NBD_REQUEST_MAGIC) {
+        LOG("invalid magic (got 0x%x)", magic);
+        return -EINVAL;
+    }
+    return 0;

 }
 
 }
 

-int nbd_receive_negotiate(int csock, off_t *size, size_t *blocksize)
+ssize_t nbd_receive_reply(int csock, struct nbd_reply *reply)

 {
 {

-       char buf[8 + 8 + 8 + 128];
-       uint64_t magic;
-
-       TRACE("Receiving negotation.");
+    uint8_t buf[NBD_REPLY_SIZE];
+    uint32_t magic;
+    ssize_t ret;

 
 

-       if (read_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
-               LOG("read failed");
-               errno = EINVAL;
-               return -1;
-       }
+    ret = read_sync(csock, buf, sizeof(buf));
+    if (ret < 0) {
+        return ret;
+    }

 
 

-       magic = be64_to_cpup((uint64_t*)(buf + 8));
-       *size = be64_to_cpup((uint64_t*)(buf + 16));
-       *blocksize = 1024;
+    if (ret != sizeof(buf)) {
+        LOG("read failed");
+        return -EINVAL;
+    }

 
 

-       TRACE("Magic is %c%c%c%c%c%c%c%c",
-             qemu_isprint(buf[0]) ? buf[0] : '.',
-             qemu_isprint(buf[1]) ? buf[1] : '.',
-             qemu_isprint(buf[2]) ? buf[2] : '.',
-             qemu_isprint(buf[3]) ? buf[3] : '.',
-             qemu_isprint(buf[4]) ? buf[4] : '.',
-             qemu_isprint(buf[5]) ? buf[5] : '.',
-             qemu_isprint(buf[6]) ? buf[6] : '.',
-             qemu_isprint(buf[7]) ? buf[7] : '.');
-       TRACE("Magic is 0x%" PRIx64, magic);
-       TRACE("Size is %" PRIu64, *size);
+    /* Reply
+       [ 0 ..  3]    magic   (NBD_REPLY_MAGIC)
+       [ 4 ..  7]    error   (0 == no error)
+       [ 7 .. 15]    handle
+     */

 
 

-       if (memcmp(buf, "NBDMAGIC", 8) != 0) {
-               LOG("Invalid magic received");
-               errno = EINVAL;
-               return -1;
-       }
+    magic = be32_to_cpup((uint32_t*)buf);
+    reply->error  = be32_to_cpup((uint32_t*)(buf + 4));
+    reply->handle = be64_to_cpup((uint64_t*)(buf + 8));

 
 

-       TRACE("Checking magic");
+    TRACE("Got reply: "
+          "{ magic = 0x%x, .error = %d, handle = %" PRIu64" }",
+          magic, reply->error, reply->handle);

 
 

-       if (magic != 0x00420281861253LL) {
-               LOG("Bad magic received");
-               errno = EINVAL;
-               return -1;
-       }
-        return 0;
+    if (magic != NBD_REPLY_MAGIC) {
+        LOG("invalid magic (got 0x%x)", magic);
+        return -EINVAL;
+    }
+    return 0;

 }
 
 }
 

-#ifndef _WIN32
-int nbd_init(int fd, int csock, off_t size, size_t blocksize)
+static ssize_t nbd_send_reply(int csock, struct nbd_reply *reply)

 {
 {

-       TRACE("Setting block size to %lu", (unsigned long)blocksize);
+    uint8_t buf[4 + 4 + 8];
+    ssize_t ret;
+
+    /* Reply
+       [ 0 ..  3]    magic   (NBD_REPLY_MAGIC)
+       [ 4 ..  7]    error   (0 == no error)
+       [ 7 .. 15]    handle
+     */
+    cpu_to_be32w((uint32_t*)buf, NBD_REPLY_MAGIC);
+    cpu_to_be32w((uint32_t*)(buf + 4), reply->error);
+    cpu_to_be64w((uint64_t*)(buf + 8), reply->handle);
+
+    TRACE("Sending response to client");
+
+    ret = write_sync(csock, buf, sizeof(buf));
+    if (ret < 0) {
+        return ret;
+    }

 
 

-       if (ioctl(fd, NBD_SET_BLKSIZE, blocksize) == -1) {
-               int serrno = errno;
-               LOG("Failed setting NBD block size");
-               errno = serrno;
-               return -1;
-       }
+    if (ret != sizeof(buf)) {
+        LOG("writing to socket failed");
+        return -EINVAL;
+    }
+    return 0;
+}

 
 

-        TRACE("Setting size to %zd block(s)", (size_t)(size / blocksize));
+#define MAX_NBD_REQUESTS 16

 
 

-       if (ioctl(fd, NBD_SET_SIZE_BLOCKS, size / blocksize) == -1) {
-               int serrno = errno;
-               LOG("Failed setting size (in blocks)");
-               errno = serrno;
-               return -1;
-       }
+typedef struct NBDRequest NBDRequest;

 
 

-       TRACE("Clearing NBD socket");
+struct NBDRequest {
+    QSIMPLEQ_ENTRY(NBDRequest) entry;
+    NBDClient *client;
+    uint8_t *data;
+};

 
 

-       if (ioctl(fd, NBD_CLEAR_SOCK) == -1) {
-               int serrno = errno;
-               LOG("Failed clearing NBD socket");
-               errno = serrno;
-               return -1;
-       }
+struct NBDExport {
+    BlockDriverState *bs;
+    off_t dev_offset;
+    off_t size;
+    uint32_t nbdflags;
+    QSIMPLEQ_HEAD(, NBDRequest) requests;
+};

 
 

-       TRACE("Setting NBD socket");
+struct NBDClient {
+    int refcount;
+    void (*close)(NBDClient *client);

 
 

-       if (ioctl(fd, NBD_SET_SOCK, csock) == -1) {
-               int serrno = errno;
-               LOG("Failed to set NBD socket");
-               errno = serrno;
-               return -1;
-       }
+    NBDExport *exp;
+    int sock;

 
 

-       TRACE("Negotiation ended");
+    Coroutine *recv_coroutine;

 
 

-       return 0;
-}
+    CoMutex send_lock;
+    Coroutine *send_coroutine;

 
 

-int nbd_disconnect(int fd)
+    int nb_requests;
+};
+
+static void nbd_client_get(NBDClient *client)

 {
 {

-       ioctl(fd, NBD_CLEAR_QUE);
-       ioctl(fd, NBD_DISCONNECT);
-       ioctl(fd, NBD_CLEAR_SOCK);
-       return 0;
+    client->refcount++;

 }
 
 }
 

-int nbd_client(int fd, int csock)
+static void nbd_client_put(NBDClient *client)

 {
 {

-       int ret;
-       int serrno;
-
-       TRACE("Doing NBD loop");
-
-       ret = ioctl(fd, NBD_DO_IT);
-       serrno = errno;
-
-       TRACE("NBD loop returned %d: %s", ret, strerror(serrno));
-
-       TRACE("Clearing NBD queue");
-       ioctl(fd, NBD_CLEAR_QUE);
-
-       TRACE("Clearing NBD socket");
-       ioctl(fd, NBD_CLEAR_SOCK);
+    if (--client->refcount == 0) {
+        g_free(client);
+    }
+}

 
 

-       errno = serrno;
-       return ret;
+static void nbd_client_close(NBDClient *client)
+{
+    qemu_set_fd_handler2(client->sock, NULL, NULL, NULL, NULL);
+    close(client->sock);
+    client->sock = -1;
+    if (client->close) {
+        client->close(client);
+    }
+    nbd_client_put(client);

 }
 }

-#else
-int nbd_init(int fd, int csock, off_t size, size_t blocksize)
+
+static NBDRequest *nbd_request_get(NBDClient *client)

 {
 {

-    errno = ENOTSUP;
-    return -1;
+    NBDRequest *req;
+    NBDExport *exp = client->exp;
+
+    assert(client->nb_requests <= MAX_NBD_REQUESTS - 1);
+    client->nb_requests++;
+
+    if (QSIMPLEQ_EMPTY(&exp->requests)) {
+        req = g_malloc0(sizeof(NBDRequest));
+        req->data = qemu_blockalign(exp->bs, NBD_BUFFER_SIZE);
+    } else {
+        req = QSIMPLEQ_FIRST(&exp->requests);
+        QSIMPLEQ_REMOVE_HEAD(&exp->requests, entry);
+    }
+    nbd_client_get(client);
+    req->client = client;
+    return req;

 }
 
 }
 

-int nbd_disconnect(int fd)
+static void nbd_request_put(NBDRequest *req)

 {
 {

-    errno = ENOTSUP;
-    return -1;
+    NBDClient *client = req->client;
+    QSIMPLEQ_INSERT_HEAD(&client->exp->requests, req, entry);
+    if (client->nb_requests-- == MAX_NBD_REQUESTS) {
+        qemu_notify_event();
+    }
+    nbd_client_put(client);

 }
 
 }
 

-int nbd_client(int fd, int csock)
+NBDExport *nbd_export_new(BlockDriverState *bs, off_t dev_offset,
+                          off_t size, uint32_t nbdflags)

 {
 {

-    errno = ENOTSUP;
-    return -1;
+    NBDExport *exp = g_malloc0(sizeof(NBDExport));
+    QSIMPLEQ_INIT(&exp->requests);
+    exp->bs = bs;
+    exp->dev_offset = dev_offset;
+    exp->nbdflags = nbdflags;
+    exp->size = size == -1 ? bdrv_getlength(bs) : size;
+    return exp;

 }
 }

-#endif

 
 

-int nbd_send_request(int csock, struct nbd_request *request)
+void nbd_export_close(NBDExport *exp)

 {
 {

-       uint8_t buf[4 + 4 + 8 + 8 + 4];
+    while (!QSIMPLEQ_EMPTY(&exp->requests)) {
+        NBDRequest *first = QSIMPLEQ_FIRST(&exp->requests);
+        QSIMPLEQ_REMOVE_HEAD(&exp->requests, entry);
+        qemu_vfree(first->data);
+        g_free(first);
+    }
+
+    bdrv_close(exp->bs);
+    g_free(exp);
+}

 
 

-       cpu_to_be32w((uint32_t*)buf, NBD_REQUEST_MAGIC);
-       cpu_to_be32w((uint32_t*)(buf + 4), request->type);
-       cpu_to_be64w((uint64_t*)(buf + 8), request->handle);
-       cpu_to_be64w((uint64_t*)(buf + 16), request->from);
-       cpu_to_be32w((uint32_t*)(buf + 24), request->len);
+static int nbd_can_read(void *opaque);
+static void nbd_read(void *opaque);
+static void nbd_restart_write(void *opaque);

 
 

-       TRACE("Sending request to client");
+static ssize_t nbd_co_send_reply(NBDRequest *req, struct nbd_reply *reply,
+                                 int len)
+{
+    NBDClient *client = req->client;
+    int csock = client->sock;
+    ssize_t rc, ret;
+
+    qemu_co_mutex_lock(&client->send_lock);
+    qemu_set_fd_handler2(csock, nbd_can_read, nbd_read,
+                         nbd_restart_write, client);
+    client->send_coroutine = qemu_coroutine_self();
+
+    if (!len) {
+        rc = nbd_send_reply(csock, reply);
+    } else {
+        socket_set_cork(csock, 1);
+        rc = nbd_send_reply(csock, reply);
+        if (rc >= 0) {
+            ret = qemu_co_send(csock, req->data, len);
+            if (ret != len) {
+                rc = -EIO;
+            }
+        }
+        socket_set_cork(csock, 0);
+    }

 
 

-       if (write_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
-               LOG("writing to socket failed");
-               errno = EINVAL;
-               return -1;
-       }
-       return 0;
+    client->send_coroutine = NULL;
+    qemu_set_fd_handler2(csock, nbd_can_read, nbd_read, NULL, client);
+    qemu_co_mutex_unlock(&client->send_lock);
+    return rc;

 }
 
 }
 

-
-static int nbd_receive_request(int csock, struct nbd_request *request)
+static ssize_t nbd_co_receive_request(NBDRequest *req, struct nbd_request *request)

 {
 {

-       uint8_t buf[4 + 4 + 8 + 8 + 4];
-       uint32_t magic;
+    NBDClient *client = req->client;
+    int csock = client->sock;
+    ssize_t rc;
+
+    client->recv_coroutine = qemu_coroutine_self();
+    rc = nbd_receive_request(csock, request);
+    if (rc < 0) {
+        if (rc != -EAGAIN) {
+            rc = -EIO;
+        }
+        goto out;
+    }

 
 

-       if (read_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
-               LOG("read failed");
-               errno = EINVAL;
-               return -1;
-       }
+    if (request->len > NBD_BUFFER_SIZE) {
+        LOG("len (%u) is larger than max len (%u)",
+            request->len, NBD_BUFFER_SIZE);
+        rc = -EINVAL;
+        goto out;
+    }
+
+    if ((request->from + request->len) < request->from) {
+        LOG("integer overflow detected! "
+            "you're probably being attacked");
+        rc = -EINVAL;
+        goto out;
+    }

 
 

-       /* Request
-          [ 0 ..  3]   magic   (NBD_REQUEST_MAGIC)
-          [ 4 ..  7]   type    (0 == READ, 1 == WRITE)
-          [ 8 .. 15]   handle
-          [16 .. 23]   from
-          [24 .. 27]   len
-        */
+    TRACE("Decoding type");

 
 

-       magic = be32_to_cpup((uint32_t*)buf);
-       request->type  = be32_to_cpup((uint32_t*)(buf + 4));
-       request->handle = be64_to_cpup((uint64_t*)(buf + 8));
-       request->from  = be64_to_cpup((uint64_t*)(buf + 16));
-       request->len   = be32_to_cpup((uint32_t*)(buf + 24));
+    if ((request->type & NBD_CMD_MASK_COMMAND) == NBD_CMD_WRITE) {
+        TRACE("Reading %u byte(s)", request->len);

 
 

-       TRACE("Got request: "
-             "{ magic = 0x%x, .type = %d, from = %" PRIu64" , len = %u }",
-             magic, request->type, request->from, request->len);
+        if (qemu_co_recv(csock, req->data, request->len) != request->len) {
+            LOG("reading from socket failed");
+            rc = -EIO;
+            goto out;
+        }
+    }
+    rc = 0;

 
 

-       if (magic != NBD_REQUEST_MAGIC) {
-               LOG("invalid magic (got 0x%x)", magic);
-               errno = EINVAL;
-               return -1;
-       }
-       return 0;
+out:
+    client->recv_coroutine = NULL;
+    return rc;

 }
 
 }
 

-int nbd_receive_reply(int csock, struct nbd_reply *reply)
+static void nbd_trip(void *opaque)

 {
 {

-       uint8_t buf[4 + 4 + 8];
-       uint32_t magic;
+    NBDClient *client = opaque;
+    NBDRequest *req = nbd_request_get(client);
+    NBDExport *exp = client->exp;
+    struct nbd_request request;
+    struct nbd_reply reply;
+    ssize_t ret;
+
+    TRACE("Reading request.");
+
+    ret = nbd_co_receive_request(req, &request);
+    if (ret == -EAGAIN) {
+        goto done;
+    }
+    if (ret == -EIO) {
+        goto out;
+    }
+
+    reply.handle = request.handle;
+    reply.error = 0;
+
+    if (ret < 0) {
+        reply.error = -ret;
+        goto error_reply;
+    }
+
+    if ((request.from + request.len) > exp->size) {
+            LOG("From: %" PRIu64 ", Len: %u, Size: %" PRIu64
+            ", Offset: %" PRIu64 "\n",
+                    request.from, request.len,
+                    (uint64_t)exp->size, (uint64_t)exp->dev_offset);
+        LOG("requested operation past EOF--bad client?");
+        goto invalid_request;
+    }

 
 

-       memset(buf, 0xAA, sizeof(buf));
+    switch (request.type & NBD_CMD_MASK_COMMAND) {
+    case NBD_CMD_READ:
+        TRACE("Request type is READ");
+
+        if (request.type & NBD_CMD_FLAG_FUA) {
+            ret = bdrv_co_flush(exp->bs);
+            if (ret < 0) {
+                LOG("flush failed");
+                reply.error = -ret;
+                goto error_reply;
+            }
+        }
+
+        ret = bdrv_read(exp->bs, (request.from + exp->dev_offset) / 512,
+                        req->data, request.len / 512);
+        if (ret < 0) {
+            LOG("reading from file failed");
+            reply.error = -ret;
+            goto error_reply;
+        }

 
 

-       if (read_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
-               LOG("read failed");
-               errno = EINVAL;
-               return -1;
-       }
+        TRACE("Read %u byte(s)", request.len);
+        if (nbd_co_send_reply(req, &reply, request.len) < 0)
+            goto out;
+        break;
+    case NBD_CMD_WRITE:
+        TRACE("Request type is WRITE");
+
+        if (exp->nbdflags & NBD_FLAG_READ_ONLY) {
+            TRACE("Server is read-only, return error");
+            reply.error = EROFS;
+            goto error_reply;
+        }

 
 

-       /* Reply
-          [ 0 ..  3]    magic   (NBD_REPLY_MAGIC)
-          [ 4 ..  7]    error   (0 == no error)
-          [ 7 .. 15]    handle
-        */
+        TRACE("Writing to device");

 
 

-       magic = be32_to_cpup((uint32_t*)buf);
-       reply->error  = be32_to_cpup((uint32_t*)(buf + 4));
-       reply->handle = be64_to_cpup((uint64_t*)(buf + 8));
+        ret = bdrv_write(exp->bs, (request.from + exp->dev_offset) / 512,
+                         req->data, request.len / 512);
+        if (ret < 0) {
+            LOG("writing to file failed");
+            reply.error = -ret;
+            goto error_reply;
+        }

 
 

-       TRACE("Got reply: "
-             "{ magic = 0x%x, .error = %d, handle = %" PRIu64" }",
-             magic, reply->error, reply->handle);
+        if (request.type & NBD_CMD_FLAG_FUA) {
+            ret = bdrv_co_flush(exp->bs);
+            if (ret < 0) {
+                LOG("flush failed");
+                reply.error = -ret;
+                goto error_reply;
+            }
+        }

 
 

-       if (magic != NBD_REPLY_MAGIC) {
-               LOG("invalid magic (got 0x%x)", magic);
-               errno = EINVAL;
-               return -1;
-       }
-       return 0;
+        if (nbd_co_send_reply(req, &reply, 0) < 0) {
+            goto out;
+        }
+        break;
+    case NBD_CMD_DISC:
+        TRACE("Request type is DISCONNECT");
+        errno = 0;
+        goto out;
+    case NBD_CMD_FLUSH:
+        TRACE("Request type is FLUSH");
+
+        ret = bdrv_co_flush(exp->bs);
+        if (ret < 0) {
+            LOG("flush failed");
+            reply.error = -ret;
+        }
+        if (nbd_co_send_reply(req, &reply, 0) < 0) {
+            goto out;
+        }
+        break;
+    case NBD_CMD_TRIM:
+        TRACE("Request type is TRIM");
+        ret = bdrv_co_discard(exp->bs, (request.from + exp->dev_offset) / 512,
+                              request.len / 512);
+        if (ret < 0) {
+            LOG("discard failed");
+            reply.error = -ret;
+        }
+        if (nbd_co_send_reply(req, &reply, 0) < 0) {
+            goto out;
+        }
+        break;
+    default:
+        LOG("invalid request type (%u) received", request.type);
+    invalid_request:
+        reply.error = -EINVAL;
+    error_reply:
+        if (nbd_co_send_reply(req, &reply, 0) < 0) {
+            goto out;
+        }
+        break;
+    }
+
+    TRACE("Request/Reply complete");
+
+done:
+    nbd_request_put(req);
+    return;
+
+out:
+    nbd_request_put(req);
+    nbd_client_close(client);
+}
+
+static int nbd_can_read(void *opaque)
+{
+    NBDClient *client = opaque;
+
+    return client->recv_coroutine || client->nb_requests < MAX_NBD_REQUESTS;
+}
+
+static void nbd_read(void *opaque)
+{
+    NBDClient *client = opaque;
+
+    if (client->recv_coroutine) {
+        qemu_coroutine_enter(client->recv_coroutine, NULL);
+    } else {
+        qemu_coroutine_enter(qemu_coroutine_create(nbd_trip), client);
+    }

 }
 
 }
 

-static int nbd_send_reply(int csock, struct nbd_reply *reply)
-{
-       uint8_t buf[4 + 4 + 8];
+static void nbd_restart_write(void *opaque)
+{
+    NBDClient *client = opaque;
+
+    qemu_coroutine_enter(client->send_coroutine, NULL);
+}

 
 

-       /* Reply
-          [ 0 ..  3]    magic   (NBD_REPLY_MAGIC)
-          [ 4 ..  7]    error   (0 == no error)
-          [ 7 .. 15]    handle
-        */
-       cpu_to_be32w((uint32_t*)buf, NBD_REPLY_MAGIC);
-       cpu_to_be32w((uint32_t*)(buf + 4), reply->error);
-       cpu_to_be64w((uint64_t*)(buf + 8), reply->handle);
-
-       TRACE("Sending response to client");
-
-       if (write_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
-               LOG("writing to socket failed");
-               errno = EINVAL;
-               return -1;
-       }
-       return 0;
-}
-
-int nbd_trip(BlockDriverState *bs, int csock, off_t size, uint64_t dev_offset,
-             off_t *offset, bool readonly, uint8_t *data, int data_size)
-{
-       struct nbd_request request;
-       struct nbd_reply reply;
-
-       TRACE("Reading request.");
-
-       if (nbd_receive_request(csock, &request) == -1)
-               return -1;
-
-       if (request.len > data_size) {
-               LOG("len (%u) is larger than max len (%u)",
-                   request.len, data_size);
-               errno = EINVAL;
-               return -1;
-       }
-
-       if ((request.from + request.len) < request.from) {
-               LOG("integer overflow detected! "
-                   "you're probably being attacked");
-               errno = EINVAL;
-               return -1;
-       }
-
-       if ((request.from + request.len) > size) {
-               LOG("From: %" PRIu64 ", Len: %u, Size: %" PRIu64
-                   ", Offset: %" PRIu64 "\n",
-                    request.from, request.len, (uint64_t)size, dev_offset);
-               LOG("requested operation past EOF--bad client?");
-               errno = EINVAL;
-               return -1;
-       }
-
-       TRACE("Decoding type");
-
-       reply.handle = request.handle;
-       reply.error = 0;
-
-       switch (request.type) {
-       case NBD_CMD_READ:
-               TRACE("Request type is READ");
-
-               if (bdrv_read(bs, (request.from + dev_offset) / 512, data,
-                             request.len / 512) == -1) {
-                       LOG("reading from file failed");
-                       errno = EINVAL;
-                       return -1;
-               }
-               *offset += request.len;
-
-               TRACE("Read %u byte(s)", request.len);
-
-               if (nbd_send_reply(csock, &reply) == -1)
-                       return -1;
-
-               TRACE("Sending data to client");
-
-               if (write_sync(csock, data, request.len) != request.len) {
-                       LOG("writing to socket failed");
-                       errno = EINVAL;
-                       return -1;
-               }
-               break;
-       case NBD_CMD_WRITE:
-               TRACE("Request type is WRITE");
-
-               TRACE("Reading %u byte(s)", request.len);
-
-               if (read_sync(csock, data, request.len) != request.len) {
-                       LOG("reading from socket failed");
-                       errno = EINVAL;
-                       return -1;
-               }
-
-               if (readonly) {
-                       TRACE("Server is read-only, return error");
-                       reply.error = 1;
-               } else {
-                       TRACE("Writing to device");
-
-                       if (bdrv_write(bs, (request.from + dev_offset) / 512,
-                                      data, request.len / 512) == -1) {
-                               LOG("writing to file failed");
-                               errno = EINVAL;
-                               return -1;
-                       }
-
-                       *offset += request.len;
-               }
-
-               if (nbd_send_reply(csock, &reply) == -1)
-                       return -1;
-               break;
-       case NBD_CMD_DISC:
-               TRACE("Request type is DISCONNECT");
-               errno = 0;
-               return 1;
-       default:
-               LOG("invalid request type (%u) received", request.type);
-               errno = EINVAL;
-               return -1;
-       }
-
-       TRACE("Request/Reply complete");
-
-       return 0;
+NBDClient *nbd_client_new(NBDExport *exp, int csock,
+                          void (*close)(NBDClient *))
+{
+    NBDClient *client;
+    if (nbd_send_negotiate(csock, exp->size, exp->nbdflags) < 0) {
+        return NULL;
+    }
+    client = g_malloc0(sizeof(NBDClient));
+    client->refcount = 1;
+    client->exp = exp;
+    client->sock = csock;
+    client->close = close;
+    qemu_co_mutex_init(&client->send_lock);
+    qemu_set_fd_handler2(csock, nbd_can_read, nbd_read, NULL, client);
+    return client;

 }
 }