[Qemu-devel] [patch 0/5] NBD support improvement

[Qemu-devel] [patch 0/5] NBD support improvement

[Laurent.Vivier]

This series of patches improves NBD support in QEMU.

[PATCH 1/5] qemu-nbd: cleanup
[PATCH 2/5] qemu-nbd: merge NBD client/server
[PATCH 3/5] qemu-nbd: add a parameter to disable host cache.
[PATCH 4/5] qemu-nbd: allow to share a disk image
[PATCH 5/5] Allow QEMU to connect directly to an NBD server

Laurent
--

[Qemu-devel] [patch 1/5] qemu-nbd: cleanup

[Laurent.Vivier]

[-- Attachment #1: nbd-cleanup.patch --]
[-- Type: text/plain, Size: 2254 bytes --]

This patch removes "#include <malloc.h>" and correct a printf format.
(from Carlo Marcelo Arenas Belon <carenas@sajinet.com.pe>).
Fill the last element of lopt (for getopt_long()) to mark the end of parameters
list. Add an entry in qemu-doc.texi.
---
 nbd.c         |    4 ++--
 qemu-doc.texi |    6 ++++++
 qemu-nbd.c    |    2 +-
 3 files changed, 9 insertions(+), 3 deletions(-)

Index: qemu/nbd.c
===================================================================
--- qemu.orig/nbd.c	2008-06-27 11:34:18.000000000 +0200
+++ qemu/nbd.c	2008-06-27 11:34:28.000000000 +0200
@@ -388,8 +388,8 @@ int nbd_trip(BlockDriverState *bs, int c
 	}
 
 	if (len > sizeof(data)) {
-		LOG("len (%u) is larger than max len (%u)",
-		    len, sizeof(data));
+		LOG("len (%u) is larger than max len (%lu)",
+		    len, (unsigned long)sizeof(data));
 		errno = EINVAL;
 		return -1;
 	}
Index: qemu/qemu-nbd.c
===================================================================
--- qemu.orig/qemu-nbd.c	2008-06-27 11:34:18.000000000 +0200
+++ qemu/qemu-nbd.c	2008-06-27 11:34:28.000000000 +0200
@@ -21,7 +21,6 @@
 #include "block_int.h"
 #include "nbd.h"
 
-#include <malloc.h>
 #include <stdarg.h>
 #include <stdio.h>
 #include <getopt.h>
@@ -168,6 +167,7 @@ int main(int argc, char **argv)
         { "partition", 1, 0, 'P' },
         { "snapshot", 0, 0, 's' },
         { "verbose", 0, 0, 'v' },
+        { NULL, 0, 0, 0 }
     };
     int ch;
     int opt_ind = 0;
Index: qemu/qemu-doc.texi
===================================================================
--- qemu.orig/qemu-doc.texi	2008-06-27 11:34:18.000000000 +0200
+++ qemu/qemu-doc.texi	2008-06-27 11:34:28.000000000 +0200
@@ -1307,6 +1307,7 @@ snapshots.
 * disk_images_snapshot_mode:: Snapshot mode
 * vm_snapshots::              VM snapshots
 * qemu_img_invocation::       qemu-img Invocation
+* qemu_nbd_invocation::       qemu-nbd Invocation
 * host_drives::               Using host drives
 * disk_images_fat_images::    Virtual FAT disk images
 @end menu
@@ -1389,6 +1390,11 @@ state is not saved or restored properly 
 
 @include qemu-img.texi
 
+@node qemu_nbd_invocation
+@subsection @code{qemu-nbd} Invocation
+
+@include qemu-nbd.texi
+
 @node host_drives
 @subsection Using host drives
 

--

[Qemu-devel] [patch 2/5] qemu-nbd: merge NBD client/server

[Laurent.Vivier]

[-- Attachment #1: qemu-nbd-local.patch --]
[-- Type: text/plain, Size: 11968 bytes --]

This patch allows to connect directly a disk image file to an NBD
device. It introduces the use of a unix socket (instead of inet).

- To connect a file to a device:

# qemu-nbd --connect=/dev/nbd0 my_disk.qcow2

Then you can see directly your disk (no need of nbd-client):

# fdisk -l /dev/nbd0

Disk /dev/nbd0: 4294 MB, 4294967296 bytes
255 heads, 63 sectors/track, 522 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes

     Device Boot      Start         End      Blocks   Id  System
/dev/nbd0p1   *           1         492     3951958+  83  Linux
/dev/nbd0p2             493         522      240975    5  Extended
/dev/nbd0p5             493         522      240943+  82  Linux swap /
Solaris

- To disconnect the file from the device:

# qemu-nbd --disconnect /dev/nbd0

Changelog:
- v2: call show_parts() from client and avoid the sleep(1). Thank you to
Avi and Anthony. Include my cleanup patch and comments from Carlo
Marcelo Arenas Belon.
- v3: allow to specify unix socket name with "--socket" and update 
documentation.
---
 nbd.c         |   74 ++++++++++++++++++++++++++++---
 nbd.h         |    5 +-
 qemu-nbd.c    |  137 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++--
 qemu-nbd.texi |    6 ++
 4 files changed, 212 insertions(+), 10 deletions(-)

Index: qemu/nbd.c
===================================================================
--- qemu.orig/nbd.c	2008-06-27 11:34:28.000000000 +0200
+++ qemu/nbd.c	2008-06-27 11:34:35.000000000 +0200
@@ -25,6 +25,7 @@
 #include <ctype.h>
 #include <inttypes.h>
 #include <sys/socket.h>
+#include <sys/un.h>
 #include <netinet/in.h>
 #include <netinet/tcp.h>
 #include <arpa/inet.h>
@@ -183,6 +184,65 @@ error:
     return -1;
 }
 
+int unix_socket_incoming(const char *path)
+{
+    int s;
+    struct sockaddr_un addr;
+    int serrno;
+
+    s = socket(PF_UNIX, SOCK_STREAM, 0);
+    if (s == -1) {
+        return -1;
+    }
+
+    memset(&addr, 0, sizeof(addr));
+    addr.sun_family = AF_UNIX;
+    pstrcpy(addr.sun_path, sizeof(addr.sun_path), path);
+
+    if (bind(s, (struct sockaddr *)&addr, sizeof(addr)) == -1) {
+        goto error;
+    }
+
+    if (listen(s, 128) == -1) {
+        goto error;
+    }
+
+    return s;
+error:
+    serrno = errno;
+    close(s);
+    errno = serrno;
+    return -1;
+}
+
+int unix_socket_outgoing(const char *path)
+{
+    int s;
+    struct sockaddr_un addr;
+    int serrno;
+
+    s = socket(PF_UNIX, SOCK_STREAM, 0);
+    if (s == -1) {
+        return -1;
+    }
+
+    memset(&addr, 0, sizeof(addr));
+    addr.sun_family = AF_UNIX;
+    pstrcpy(addr.sun_path, sizeof(addr.sun_path), path);
+
+    if (connect(s, (struct sockaddr *)&addr, sizeof(addr)) == -1) {
+        goto error;
+    }
+
+    return s;
+error:
+    serrno = errno;
+    close(s);
+    errno = serrno;
+    return -1;
+}
+
+
 /* Basic flow
 
    Server         Client
@@ -225,12 +285,10 @@ int nbd_negotiate(BlockDriverState *bs, 
 	return 0;
 }
 
-int nbd_receive_negotiate(int fd, int csock)
+int nbd_receive_negotiate(int csock, off_t *size, size_t *blocksize)
 {
 	char buf[8 + 8 + 8 + 128];
 	uint64_t magic;
-	off_t size;
-	size_t blocksize;
 
 	TRACE("Receiving negotation.");
 
@@ -241,8 +299,8 @@ int nbd_receive_negotiate(int fd, int cs
 	}
 
 	magic = be64_to_cpup((uint64_t*)(buf + 8));
-	size = be64_to_cpup((uint64_t*)(buf + 16));
-	blocksize = 1024;
+	*size = be64_to_cpup((uint64_t*)(buf + 16));
+	*blocksize = 1024;
 
 	TRACE("Magic is %c%c%c%c%c%c%c%c",
 	      isprint(buf[0]) ? buf[0] : '.',
@@ -254,7 +312,7 @@ int nbd_receive_negotiate(int fd, int cs
 	      isprint(buf[6]) ? buf[6] : '.',
 	      isprint(buf[7]) ? buf[7] : '.');
 	TRACE("Magic is 0x%" PRIx64, magic);
-	TRACE("Size is %" PRIu64, size);
+	TRACE("Size is %" PRIu64, *size);
 
 	if (memcmp(buf, "NBDMAGIC", 8) != 0) {
 		LOG("Invalid magic received");
@@ -269,7 +327,11 @@ int nbd_receive_negotiate(int fd, int cs
 		errno = EINVAL;
 		return -1;
 	}
+        return 0;
+}
 
+int nbd_init(int fd, int csock, off_t size, size_t blocksize)
+{
 	TRACE("Setting block size to %lu", (unsigned long)blocksize);
 
 	if (ioctl(fd, NBD_SET_BLKSIZE, blocksize) == -1) {
Index: qemu/nbd.h
===================================================================
--- qemu.orig/nbd.h	2008-06-27 11:34:18.000000000 +0200
+++ qemu/nbd.h	2008-06-27 11:34:35.000000000 +0200
@@ -27,9 +27,12 @@
 #include "block_int.h"
 
 int tcp_socket_incoming(const char *address, uint16_t port);
+int unix_socket_outgoing(const char *path);
+int unix_socket_incoming(const char *path);
 
 int nbd_negotiate(BlockDriverState *bs, int csock, off_t size);
-int nbd_receive_negotiate(int fd, int csock);
+int nbd_receive_negotiate(int csock, off_t *size, size_t *blocksize);
+int nbd_init(int fd, int csock, off_t size, size_t blocksize);
 int nbd_trip(BlockDriverState *bs, int csock, off_t size, uint64_t dev_offset, off_t *offset, bool readonly);
 int nbd_client(int fd, int csock);
 int nbd_disconnect(int fd);
Index: qemu/qemu-nbd.c
===================================================================
--- qemu.orig/qemu-nbd.c	2008-06-27 11:34:28.000000000 +0200
+++ qemu/qemu-nbd.c	2008-06-27 11:34:35.000000000 +0200
@@ -25,10 +25,14 @@
 #include <stdio.h>
 #include <getopt.h>
 #include <err.h>
+#include <sys/types.h>
 #include <sys/socket.h>
 #include <netinet/in.h>
 #include <netinet/tcp.h>
 #include <arpa/inet.h>
+#include <signal.h>
+
+#define SOCKET_PATH    "/var/lock/qemu-nbd-%s"
 
 int verbose;
 
@@ -41,14 +45,18 @@ static void usage(const char *name)
 "  -p, --port=PORT      port to listen on (default `1024')\n"
 "  -o, --offset=OFFSET  offset into the image\n"
 "  -b, --bind=IFACE     interface to bind to (default `0.0.0.0')\n"
+"  -k, --socket=PATH    path to the unix socket\n"
+"                       (default '"SOCKET_PATH"')\n"
 "  -r, --read-only      export read-only\n"
 "  -P, --partition=NUM  only expose partition NUM\n"
+"  -c, --connect=DEV    connect FILE to the local NBD device DEV\n"
+"  -d, --disconnect     disconnect the specified device\n"
 "  -v, --verbose        display extra debugging information\n"
 "  -h, --help           display this help and exit\n"
 "  -V, --version        output version information and exit\n"
 "\n"
 "Report bugs to <anthony@codemonkey.ws>\n"
-    , name);
+    , name, "DEVICE");
 }
 
 static void version(const char *name)
@@ -144,27 +152,51 @@ static int find_partition(BlockDriverSta
     return -1;
 }
 
+static void show_parts(const char *device)
+{
+    if (fork() == 0) {
+        int nbd;
+
+        /* linux just needs an open() to trigger
+         * the partition table update
+         * but remember to load the module with max_part != 0 :
+         *     modprobe nbd max_part=63
+         */
+        nbd = open(device, O_RDWR);
+        if (nbd != -1)
+              close(nbd);
+        exit(0);
+    }
+}
+
 int main(int argc, char **argv)
 {
     BlockDriverState *bs;
     off_t dev_offset = 0;
     off_t offset = 0;
     bool readonly = false;
+    bool disconnect = false;
     const char *bindto = "0.0.0.0";
     int port = 1024;
     int sock, csock;
     struct sockaddr_in addr;
     socklen_t addr_len = sizeof(addr);
     off_t fd_size;
-    const char *sopt = "hVbo:p:rsP:v";
+    char *device = NULL;
+    char *socket = NULL;
+    char sockpath[128];
+    const char *sopt = "hVbo:p:rsP:c:dvk:";
     struct option lopt[] = {
         { "help", 0, 0, 'h' },
         { "version", 0, 0, 'V' },
         { "bind", 1, 0, 'b' },
         { "port", 1, 0, 'p' },
+        { "socket", 1, 0, 'k' },
         { "offset", 1, 0, 'o' },
         { "read-only", 0, 0, 'r' },
         { "partition", 1, 0, 'P' },
+        { "connect", 1, 0, 'c' },
+        { "disconnect", 0, 0, 'd' },
         { "snapshot", 0, 0, 's' },
         { "verbose", 0, 0, 'v' },
         { NULL, 0, 0, 0 }
@@ -175,6 +207,8 @@ int main(int argc, char **argv)
     char *end;
     bool snapshot = false;
     int partition = -1;
+    int fd;
+    int ret;
 
     while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
         switch (ch) {
@@ -213,6 +247,17 @@ int main(int argc, char **argv)
             if (partition < 1 || partition > 8)
                 errx(EINVAL, "Invalid partition %d", partition);
             break;
+        case 'k':
+            socket = optarg;
+            if (socket[0] != '/')
+                errx(EINVAL, "socket path must be absolute\n");
+            break;
+        case 'd':
+            disconnect = true;
+            break;
+        case 'c':
+            device = optarg;
+            break;
         case 'v':
             verbose = 1;
             break;
@@ -236,6 +281,20 @@ int main(int argc, char **argv)
              argv[0]);
     }
 
+    if (disconnect) {
+        fd = open(argv[optind], O_RDWR);
+        if (fd == -1)
+            errx(errno, "Cannot open %s", argv[optind]);
+
+        nbd_disconnect(fd);
+
+        close(fd);
+
+        printf("%s disconnected\n", argv[optind]);
+
+	return 0;
+    }
+
     bdrv_init();
 
     bs = bdrv_new("hda");
@@ -251,7 +310,77 @@ int main(int argc, char **argv)
         find_partition(bs, partition, &dev_offset, &fd_size))
         errx(errno, "Could not find partition %d", partition);
 
-    sock = tcp_socket_incoming(bindto, port);
+    if (device) {
+	pid_t pid;
+	if (!verbose)
+            daemon(0, 0);	/* detach client and server */
+
+        if (socket == NULL) {
+            sprintf(sockpath, SOCKET_PATH, basename(device));
+            socket = sockpath;
+        }
+
+        pid = fork();
+        if (pid < 0)
+            return 1;
+	if (pid != 0) {
+            off_t size;
+            size_t blocksize;
+
+            ret = 0;
+            bdrv_close(bs);
+
+            do {
+                sock = unix_socket_outgoing(socket);
+                if (sock == -1) {
+                    if (errno != ENOENT && errno != ECONNREFUSED)
+                        goto exit;
+                    sleep(1);	/* wait children */
+                }
+            } while (sock == -1);
+
+            fd = open(device, O_RDWR);
+            if (fd == -1) {
+                ret = 1;
+                goto exit;
+            }
+
+            ret = nbd_receive_negotiate(sock, &size, &blocksize);
+            if (ret == -1) {
+                ret = 1;
+                goto exit;
+            }
+
+            ret = nbd_init(fd, sock, size, blocksize);
+            if (ret == -1) {
+                ret = 1;
+                goto exit;
+            }
+
+            printf("NBD device %s is now connected to file %s\n",
+                    device, argv[optind]);
+
+	    /* update partition table */
+
+            show_parts(device);
+
+            nbd_client(fd, sock);
+            close(fd);
+exit:
+            kill(pid, SIGTERM);
+            unlink(socket);
+
+            return ret;
+        }
+        /* children */
+    }
+
+    if (socket) {
+        sock = unix_socket_incoming(socket);
+    } else {
+        sock = tcp_socket_incoming(bindto, port);
+    }
+
     if (sock == -1)
         return 1;
 
@@ -270,6 +399,8 @@ int main(int argc, char **argv)
     close(csock);
     close(sock);
     bdrv_close(bs);
+    if (socket)
+        unlink(socket);
 
     return 0;
 }
Index: qemu/qemu-nbd.texi
===================================================================
--- qemu.orig/qemu-nbd.texi	2008-06-27 11:34:18.000000000 +0200
+++ qemu/qemu-nbd.texi	2008-06-27 11:34:35.000000000 +0200
@@ -20,10 +20,16 @@ Export Qemu disk image using NBD protoco
   offset into the image
 @item -b, --bind=IFACE
   interface to bind to (default `0.0.0.0')
+@item -k, --socket=PATH
+  Use a unix socket with path PATH
 @item -r, --read-only
   export read-only
 @item -P, --partition=NUM
   only expose partition NUM
+@item -c, --connect
+  connect FILE to NBD device DEV
+@item -d, --disconnect
+  disconnect the specified device
 @item -v, --verbose
   display extra debugging information
 @item -h, --help

--

[Qemu-devel] [patch 3/5] qemu-nbd: add a parameter to disable host cache.

[Laurent.Vivier]

[-- Attachment #1: qemu-nbd-direct.patch --]
[-- Type: text/plain, Size: 8602 bytes --]

This patch add a parameter "--nocache" to disable host cache for disk image.
It avoid to duplicate page for NBD device and disk image file.
---
 Makefile          |    5 ++++-
 block-raw-posix.c |   16 ++++++++--------
 nbd.c             |   13 ++++---------
 nbd.h             |    3 ++-
 qemu-nbd.c        |   22 +++++++++++++++++-----
 qemu-nbd.texi     |    4 ++++
 6 files changed, 39 insertions(+), 24 deletions(-)

Index: qemu/Makefile
===================================================================
--- qemu.orig/Makefile	2008-06-27 11:34:17.000000000 +0200
+++ qemu/Makefile	2008-06-27 11:34:38.000000000 +0200
@@ -172,8 +172,11 @@ qemu-img-%.o: %.c
 %.o: %.c
 	$(CC) $(CFLAGS) $(CPPFLAGS) -c -o $@ $<
 
+qemu-nbd-%.o: %.c
+	$(CC) $(CFLAGS) $(CPPFLAGS) -DQEMU_NBD -c -o $@ $<
+
 qemu-nbd$(EXESUF):  qemu-nbd.o nbd.o qemu-img-block.o \
-		    $(QEMU_IMG_BLOCK_OBJS)
+		    osdep.o qemu-nbd-block-raw-posix.o $(BLOCK_OBJS)
 	$(CC) $(LDFLAGS) -o $@ $^ -lz $(LIBS)
 
 # dyngen host tool
Index: qemu/block-raw-posix.c
===================================================================
--- qemu.orig/block-raw-posix.c	2008-06-27 11:34:18.000000000 +0200
+++ qemu/block-raw-posix.c	2008-06-27 11:34:38.000000000 +0200
@@ -22,7 +22,7 @@
  * THE SOFTWARE.
  */
 #include "qemu-common.h"
-#ifndef QEMU_IMG
+#if !defined(QEMU_IMG) && !defined(QEMU_NBD)
 #include "qemu-timer.h"
 #include "exec-all.h"
 #endif
@@ -59,7 +59,7 @@
 //#define DEBUG_FLOPPY
 
 //#define DEBUG_BLOCK
-#if defined(DEBUG_BLOCK) && !defined(QEMU_IMG)
+#if defined(DEBUG_BLOCK) && !defined(QEMU_IMG) && !defined(QEMU_NBD)
 #define DEBUG_BLOCK_PRINT(formatCstr, args...) do { if (loglevel != 0)	\
     { fprintf(logfile, formatCstr, ##args); fflush(logfile); } } while (0)
 #else
@@ -434,7 +434,7 @@ static int aio_initialized = 0;
 
 static void aio_signal_handler(int signum)
 {
-#ifndef QEMU_IMG
+#if !defined(QEMU_IMG) && !defined(QEMU_NBD)
     CPUState *env = cpu_single_env;
     if (env) {
         /* stop the currently executing cpu because a timer occured */
@@ -544,7 +544,7 @@ void qemu_aio_wait(void)
     sigset_t set;
     int nb_sigs;
 
-#ifndef QEMU_IMG
+#if !defined(QEMU_IMG) && !defined(QEMU_NBD)
     if (qemu_bh_poll())
         return;
 #endif
@@ -586,7 +586,7 @@ static RawAIOCB *raw_aio_setup(BlockDriv
     return acb;
 }
 
-#ifndef QEMU_IMG
+#if !defined(QEMU_IMG) && !defined(QEMU_NBD)
 static void raw_aio_em_cb(void* opaque)
 {
     RawAIOCB *acb = opaque;
@@ -605,7 +605,7 @@ static BlockDriverAIOCB *raw_aio_read(Bl
      * If O_DIRECT is used and the buffer is not aligned fall back
      * to synchronous IO.
      */
-#if defined(O_DIRECT) && !defined(QEMU_IMG)
+#if defined(O_DIRECT) && !defined(QEMU_IMG) && !defined(QEMU_NBD)
     BDRVRawState *s = bs->opaque;
 
     if (unlikely(s->aligned_buf != NULL && ((uintptr_t) buf % 512))) {
@@ -638,7 +638,7 @@ static BlockDriverAIOCB *raw_aio_write(B
      * If O_DIRECT is used and the buffer is not aligned fall back
      * to synchronous IO.
      */
-#if defined(O_DIRECT) && !defined(QEMU_IMG)
+#if defined(O_DIRECT) && !defined(QEMU_IMG) && !defined(QEMU_NBD)
     BDRVRawState *s = bs->opaque;
 
     if (unlikely(s->aligned_buf != NULL && ((uintptr_t) buf % 512))) {
@@ -941,7 +941,7 @@ static int hdev_open(BlockDriverState *b
     return 0;
 }
 
-#if defined(__linux__) && !defined(QEMU_IMG)
+#if defined(__linux__) && !defined(QEMU_IMG) && !defined(QEMU_NBD)
 
 /* Note: we do not have a reliable method to detect if the floppy is
    present. The current method is to try to open the floppy at every
Index: qemu/nbd.c
===================================================================
--- qemu.orig/nbd.c	2008-06-27 11:34:35.000000000 +0200
+++ qemu/nbd.c	2008-06-27 11:34:38.000000000 +0200
@@ -404,13 +404,8 @@ int nbd_client(int fd, int csock)
 	return ret;
 }
 
-int nbd_trip(BlockDriverState *bs, int csock, off_t size, uint64_t dev_offset, off_t *offset, bool readonly)
+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)
 {
-#ifndef _REENTRANT
-	static uint8_t data[1024 * 1024]; // keep this off of the stack
-#else
-	uint8_t data[1024 * 1024];
-#endif
 	uint8_t buf[4 + 4 + 8 + 8 + 4];
 	uint32_t magic;
 	uint32_t type;
@@ -449,9 +444,9 @@ int nbd_trip(BlockDriverState *bs, int c
 		return -1;
 	}
 
-	if (len > sizeof(data)) {
-		LOG("len (%u) is larger than max len (%lu)",
-		    len, (unsigned long)sizeof(data));
+	if (len > data_size) {
+		LOG("len (%u) is larger than max len (%u)",
+		    len, data_size);
 		errno = EINVAL;
 		return -1;
 	}
Index: qemu/qemu-nbd.c
===================================================================
--- qemu.orig/qemu-nbd.c	2008-06-27 11:34:35.000000000 +0200
+++ qemu/qemu-nbd.c	2008-06-27 11:34:38.000000000 +0200
@@ -34,6 +34,8 @@
 
 #define SOCKET_PATH    "/var/lock/qemu-nbd-%s"
 
+#define NBD_BUFFER_SIZE (1024*1024)
+
 int verbose;
 
 static void usage(const char *name)
@@ -49,6 +51,8 @@ static void usage(const char *name)
 "                       (default '"SOCKET_PATH"')\n"
 "  -r, --read-only      export read-only\n"
 "  -P, --partition=NUM  only expose partition NUM\n"
+"  -s, --snapshot       use snapshot file\n"
+"  -n, --nocache        disable host cache\n"
 "  -c, --connect=DEV    connect FILE to the local NBD device DEV\n"
 "  -d, --disconnect     disconnect the specified device\n"
 "  -v, --verbose        display extra debugging information\n"
@@ -185,7 +189,7 @@ int main(int argc, char **argv)
     char *device = NULL;
     char *socket = NULL;
     char sockpath[128];
-    const char *sopt = "hVbo:p:rsP:c:dvk:";
+    const char *sopt = "hVbo:p:rsnP:c:dvk:";
     struct option lopt[] = {
         { "help", 0, 0, 'h' },
         { "version", 0, 0, 'V' },
@@ -198,6 +202,7 @@ int main(int argc, char **argv)
         { "connect", 1, 0, 'c' },
         { "disconnect", 0, 0, 'd' },
         { "snapshot", 0, 0, 's' },
+        { "nocache", 0, 0, 'n' },
         { "verbose", 0, 0, 'v' },
         { NULL, 0, 0, 0 }
     };
@@ -205,15 +210,19 @@ int main(int argc, char **argv)
     int opt_ind = 0;
     int li;
     char *end;
-    bool snapshot = false;
+    int flags = 0;
     int partition = -1;
     int fd;
     int ret;
+    uint8_t *data;
 
     while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
         switch (ch) {
         case 's':
-            snapshot = true;
+            flags |= BDRV_O_SNAPSHOT;
+            break;
+        case 'n':
+            flags |= BDRV_O_DIRECT;
             break;
         case 'b':
             bindto = optarg;
@@ -301,7 +310,7 @@ int main(int argc, char **argv)
     if (bs == NULL)
         return 1;
 
-    if (bdrv_open(bs, argv[optind], snapshot) == -1)
+    if (bdrv_open(bs, argv[optind], flags) == -1)
         return 1;
 
     fd_size = bs->total_sectors * 512;
@@ -394,7 +403,10 @@ exit:
     if (nbd_negotiate(bs, csock, fd_size) == -1)
         return 1;
 
-    while (nbd_trip(bs, csock, fd_size, dev_offset, &offset, readonly) == 0);
+    data = qemu_memalign(512, NBD_BUFFER_SIZE);
+    while (nbd_trip(bs, csock, fd_size, dev_offset, &offset, readonly,
+                    data, NBD_BUFFER_SIZE) == 0);
+    qemu_free(data);
 
     close(csock);
     close(sock);
Index: qemu/nbd.h
===================================================================
--- qemu.orig/nbd.h	2008-06-27 11:34:35.000000000 +0200
+++ qemu/nbd.h	2008-06-27 11:34:38.000000000 +0200
@@ -33,7 +33,8 @@ int unix_socket_incoming(const char *pat
 int nbd_negotiate(BlockDriverState *bs, int csock, off_t size);
 int nbd_receive_negotiate(int csock, off_t *size, size_t *blocksize);
 int nbd_init(int fd, int csock, off_t size, size_t blocksize);
-int nbd_trip(BlockDriverState *bs, int csock, off_t size, uint64_t dev_offset, off_t *offset, bool readonly);
+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);
 int nbd_client(int fd, int csock);
 int nbd_disconnect(int fd);
 
Index: qemu/qemu-nbd.texi
===================================================================
--- qemu.orig/qemu-nbd.texi	2008-06-27 11:34:35.000000000 +0200
+++ qemu/qemu-nbd.texi	2008-06-27 11:34:38.000000000 +0200
@@ -26,6 +26,10 @@ Export Qemu disk image using NBD protoco
   export read-only
 @item -P, --partition=NUM
   only expose partition NUM
+@item -s, --snapshot
+  use snapshot file
+@item -n, --nocache
+  disable host cache
 @item -c, --connect
   connect FILE to NBD device DEV
 @item -d, --disconnect

--

[Qemu-devel] [patch 4/5] qemu-nbd: allow to share a disk image

[Laurent.Vivier]

[-- Attachment #1: qemu-nbd-shared.patch --]
[-- Type: text/plain, Size: 6352 bytes --]

This patch allows to share a disk image using an NBD server.

You must specify by how many clients it could be shared:

server$ qemu-nbd --share=2 my_disk.qcow2

client1# nbd-client server 1024 /dev/nbd0
client2# nbd-client server 1024 /dev/nbd0

if you start a third client, it will hang until one of the previous connected
clients disconnects.
---
 qemu-nbd.c    |   92 ++++++++++++++++++++++++++++++++++++++++++++++------------
 qemu-nbd.texi |    2 +
 2 files changed, 75 insertions(+), 19 deletions(-)

Index: qemu/qemu-nbd.c
===================================================================
--- qemu.orig/qemu-nbd.c	2008-06-27 11:34:38.000000000 +0200
+++ qemu/qemu-nbd.c	2008-06-27 11:34:41.000000000 +0200
@@ -55,6 +55,7 @@ static void usage(const char *name)
 "  -n, --nocache        disable host cache\n"
 "  -c, --connect=DEV    connect FILE to the local NBD device DEV\n"
 "  -d, --disconnect     disconnect the specified device\n"
+"  -e, --shared=NUM     device can be shared by NUM clients (default '1')\n"
 "  -v, --verbose        display extra debugging information\n"
 "  -h, --help           display this help and exit\n"
 "  -V, --version        output version information and exit\n"
@@ -182,14 +183,13 @@ int main(int argc, char **argv)
     bool disconnect = false;
     const char *bindto = "0.0.0.0";
     int port = 1024;
-    int sock, csock;
     struct sockaddr_in addr;
     socklen_t addr_len = sizeof(addr);
     off_t fd_size;
     char *device = NULL;
     char *socket = NULL;
     char sockpath[128];
-    const char *sopt = "hVbo:p:rsnP:c:dvk:";
+    const char *sopt = "hVbo:p:rsnP:c:dvk:e:";
     struct option lopt[] = {
         { "help", 0, 0, 'h' },
         { "version", 0, 0, 'V' },
@@ -203,6 +203,7 @@ int main(int argc, char **argv)
         { "disconnect", 0, 0, 'd' },
         { "snapshot", 0, 0, 's' },
         { "nocache", 0, 0, 'n' },
+        { "shared", 1, 0, 'e' },
         { "verbose", 0, 0, 'v' },
         { NULL, 0, 0, 0 }
     };
@@ -212,9 +213,15 @@ int main(int argc, char **argv)
     char *end;
     int flags = 0;
     int partition = -1;
-    int fd;
     int ret;
+    int shared = 1;
     uint8_t *data;
+    fd_set fds;
+    int *sharing_fds;
+    int fd;
+    int i;
+    int nb_fds = 0;
+    int max_fd;
 
     while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
         switch (ch) {
@@ -267,6 +274,15 @@ int main(int argc, char **argv)
         case 'c':
             device = optarg;
             break;
+        case 'e':
+            shared = strtol(optarg, &end, 0);
+            if (*end) {
+                errx(EINVAL, "Invalid shared device number '%s'", optarg);
+            }
+            if (shared < 1) {
+                errx(EINVAL, "Shared device number must be greater than 0\n");
+            }
+            break;
         case 'v':
             verbose = 1;
             break;
@@ -321,6 +337,8 @@ int main(int argc, char **argv)
 
     if (device) {
 	pid_t pid;
+        int sock;
+
 	if (!verbose)
             daemon(0, 0);	/* detach client and server */
 
@@ -384,33 +402,69 @@ exit:
         /* children */
     }
 
+    sharing_fds = qemu_malloc((shared + 1) * sizeof(int));
+    if (sharing_fds == NULL)
+        errx(ENOMEM, "Cannot allocate sharing fds");
+
     if (socket) {
-        sock = unix_socket_incoming(socket);
+        sharing_fds[0] = unix_socket_incoming(socket);
     } else {
-        sock = tcp_socket_incoming(bindto, port);
+        sharing_fds[0] = tcp_socket_incoming(bindto, port);
     }
 
-    if (sock == -1)
+    if (sharing_fds[0] == -1)
         return 1;
+    max_fd = sharing_fds[0];
+    nb_fds++;
 
-    csock = accept(sock,
-               (struct sockaddr *)&addr,
-               &addr_len);
-    if (csock == -1)
-        return 1;
+    data = qemu_memalign(512, NBD_BUFFER_SIZE);
+    if (data == NULL)
+        errx(ENOMEM, "Cannot allocate data buffer");
 
-    /* new fd_size is calculated by find_partition */
-    if (nbd_negotiate(bs, csock, fd_size) == -1)
-        return 1;
+    do {
 
-    data = qemu_memalign(512, NBD_BUFFER_SIZE);
-    while (nbd_trip(bs, csock, fd_size, dev_offset, &offset, readonly,
-                    data, NBD_BUFFER_SIZE) == 0);
+        FD_ZERO(&fds);
+        for (i = 0; i < nb_fds; i++)
+            FD_SET(sharing_fds[i], &fds);
+
+        ret = select(max_fd + 1, &fds, NULL, NULL, NULL);
+        if (ret == -1)
+            break;
+
+        if (FD_ISSET(sharing_fds[0], &fds))
+            ret--;
+        for (i = 1; i < nb_fds && ret; i++) {
+            if (FD_ISSET(sharing_fds[i], &fds)) {
+                if (nbd_trip(bs, sharing_fds[i], fd_size, dev_offset,
+                    &offset, readonly, data, NBD_BUFFER_SIZE) != 0) {
+                    close(sharing_fds[i]);
+                    nb_fds--;
+                    sharing_fds[i] = sharing_fds[nb_fds];
+                    i--;
+                }
+                ret--;
+            }
+        }
+        /* new connection ? */
+        if (FD_ISSET(sharing_fds[0], &fds)) {
+            if (nb_fds < shared + 1) {
+                sharing_fds[nb_fds] = accept(sharing_fds[0],
+                                             (struct sockaddr *)&addr,
+                                             &addr_len);
+                if (sharing_fds[nb_fds] != -1 &&
+                    nbd_negotiate(bs, sharing_fds[nb_fds], fd_size) != -1) {
+                        if (sharing_fds[nb_fds] > max_fd)
+                            max_fd = sharing_fds[nb_fds];
+                        nb_fds++;
+                }
+            }
+        }
+    } while (nb_fds > 1);
     qemu_free(data);
 
-    close(csock);
-    close(sock);
+    close(sharing_fds[0]);
     bdrv_close(bs);
+    qemu_free(sharing_fds);
     if (socket)
         unlink(socket);
 
Index: qemu/qemu-nbd.texi
===================================================================
--- qemu.orig/qemu-nbd.texi	2008-06-27 11:34:38.000000000 +0200
+++ qemu/qemu-nbd.texi	2008-06-27 11:34:41.000000000 +0200
@@ -34,6 +34,8 @@ Export Qemu disk image using NBD protoco
   connect FILE to NBD device DEV
 @item -d, --disconnect
   disconnect the specified device
+@item -e, --shared=NUM
+  device can be shared by NUM clients (default '1')
 @item -v, --verbose
   display extra debugging information
 @item -h, --help

--

[Qemu-devel] [patch 5/5] Allow QEMU to connect directly to an NBD server

[Laurent.Vivier]

[-- Attachment #1: block-nbd.patch --]
[-- Type: text/plain, Size: 22960 bytes --]

This patch add a new protocol to QEMU block interface: "nbd".

It allows to QEMU to connect to an NBD server without using kernel support
(/dev/nbd*).

example:
qemu linux.img -hdb nbd:my_nbd_server.mydomain.org:1024

If the NBD server is located on the same host, you can use an unix socket
instead of an inet socket:
qemu linux.img -hdb nbd:unix:/tmp/my_socket

In this case, the block device must be exported using qemu-nbd:
qemu-nbd --socket=/tmp/my_socket my_disk.qcow2

The use of qemu-nbd allows to share a disk between several guests:

qemu-nbd --socket=/tmp/my_socket --share=2 my_disk.qcow2

and then you can use it with two guests:

qemu linux1.img -hdb nbd:unix:/tmp/my_socket
qemu linux2.img -hdb nbd:unix:/tmp/my_socket
---
 Makefile      |    8 +-
 block-nbd.c   |  194 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 block.c       |    1 
 block.h       |    1 
 nbd.c         |  196 ++++++++++++++++++++++++++++++++++++++++------------------
 nbd.h         |   22 ++++++
 qemu-doc.texi |   35 ++++++++++
 qemu-nbd.c    |   10 ++
 qemu-nbd.texi |    2 
 9 files changed, 403 insertions(+), 66 deletions(-)

Index: qemu/block-nbd.c
===================================================================
--- /dev/null	1970-01-01 00:00:00.000000000 +0000
+++ qemu/block-nbd.c	2008-06-27 11:35:14.000000000 +0200
@@ -0,0 +1,194 @@
+/*
+ * QEMU Block driver for  NBD
+ *
+ * Copyright (C) 2008 Bull S.A.S.
+ *     Author: Laurent Vivier <Laurent.Vivier@bull;net>
+ *
+ * Some parts:
+ *    Copyright (C) 2007 Anthony Liguori <anthony@codemonkey.ws>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+
+#include "qemu-common.h"
+#include "nbd.h"
+
+#include <sys/types.h>
+#include <unistd.h>
+#include <sys/socket.h>
+#include <sys/un.h>
+#include <netinet/in.h>
+#include <arpa/inet.h>
+#include <pthread.h>
+
+typedef struct BDRVNBDState {
+    int sock;
+    off_t size;
+    size_t blocksize;
+} BDRVNBDState;
+
+static int nbd_open(BlockDriverState *bs, const char* filename, int flags)
+{
+    BDRVNBDState *s = bs->opaque;
+    const char *host;
+    const char *unixpath;
+    int sock;
+    off_t size;
+    size_t blocksize;
+    int ret;
+
+    if ((flags & BDRV_O_CREAT))
+        return -EINVAL;
+
+    if (!strstart(filename, "nbd:", &host))
+        return -EINVAL;
+
+    if (strstart(host, "unix:", &unixpath)) {
+
+        if (unixpath[0] != '/')
+            return -EINVAL;
+
+        sock = unix_socket_outgoing(unixpath);
+
+    } else {
+        uint16_t port;
+        char *p, *r;
+        char hostname[128];
+
+        pstrcpy(hostname, 128, host);
+
+        p = strchr(hostname, ':');
+        if (p == NULL)
+            return -EINVAL;
+
+        *p = '\0';
+        p++;
+
+        port = strtol(p, &r, 0);
+        if (r == p)
+            return -EINVAL;
+        sock = tcp_socket_outgoing(hostname, port);
+    }
+
+    if (sock == -1)
+        return -errno;
+
+    ret = nbd_receive_negotiate(sock, &size, &blocksize);
+    if (ret == -1)
+        return -errno;
+
+    s->sock = sock;
+    s->size = size;
+    s->blocksize = blocksize;
+
+    return 0;
+}
+
+static int nbd_read(BlockDriverState *bs, int64_t sector_num,
+                    uint8_t *buf, int nb_sectors)
+{
+    BDRVNBDState *s = bs->opaque;
+    struct nbd_request request;
+    struct nbd_reply reply;
+
+    request.type = NBD_CMD_READ;
+    request.handle = (uint64_t)bs;
+    request.from = sector_num * 512;;
+    request.len = nb_sectors * 512;
+
+    if (nbd_send_request(s->sock, &request) == -1)
+        return -errno;
+
+    if (nbd_receive_reply(s->sock, &reply) == -1)
+        return -errno;
+
+    if (reply.error !=0)
+        return -reply.error;
+
+    if (reply.handle != request.handle)
+        return -EIO;
+
+    if (nbd_wr_sync(s->sock, buf, request.len, 1) != request.len)
+        return -EIO;
+
+    return 0;
+}
+
+static int nbd_write(BlockDriverState *bs, int64_t sector_num,
+                     const uint8_t *buf, int nb_sectors)
+{
+    BDRVNBDState *s = bs->opaque;
+    struct nbd_request request;
+    struct nbd_reply reply;
+
+    request.type = NBD_CMD_WRITE;
+    request.handle = (uint64_t)bs;
+    request.from = sector_num * 512;;
+    request.len = nb_sectors * 512;
+
+    if (nbd_send_request(s->sock, &request) == -1)
+        return -errno;
+
+    if (nbd_wr_sync(s->sock, (uint8_t*)buf, request.len, 0) != request.len)
+        return -EIO;
+
+    if (nbd_receive_reply(s->sock, &reply) == -1)
+        return -errno;
+
+    if (reply.error !=0)
+        return -reply.error;
+
+    if (reply.handle != request.handle)
+        return -EIO;
+
+    return 0;
+}
+
+static void nbd_close(BlockDriverState *bs)
+{
+    BDRVNBDState *s = bs->opaque;
+    struct nbd_request request;
+
+    request.type = NBD_CMD_DISC;
+    request.handle = (uint64_t)bs;
+    request.from = 0;
+    request.len = 0;
+    nbd_send_request(s->sock, &request);
+
+    close(s->sock);
+}
+
+static int64_t  nbd_getlength(BlockDriverState *bs)
+{
+    BDRVNBDState *s = bs->opaque;
+
+    return s->size;
+}
+
+BlockDriver bdrv_nbd = {
+    "nbd",
+    sizeof(BDRVNBDState),
+    NULL, /* no probe for protocols */
+    nbd_open,
+    nbd_read,
+    nbd_write,
+    nbd_close,
+    .bdrv_getlength = nbd_getlength,
+    .protocol_name = "nbd",
+};
Index: qemu/Makefile
===================================================================
--- qemu.orig/Makefile	2008-06-27 11:35:11.000000000 +0200
+++ qemu/Makefile	2008-06-27 11:35:14.000000000 +0200
@@ -42,7 +42,7 @@ recurse-all: $(SUBDIR_RULES)
 BLOCK_OBJS=cutils.o qemu-malloc.o
 BLOCK_OBJS+=block-cow.o block-qcow.o aes.o block-vmdk.o block-cloop.o
 BLOCK_OBJS+=block-dmg.o block-bochs.o block-vpc.o block-vvfat.o
-BLOCK_OBJS+=block-qcow2.o block-parallels.o
+BLOCK_OBJS+=block-qcow2.o block-parallels.o block-nbd.o
 
 ######################################################################
 # libqemu_common.a: Target independent part of system emulation. The
@@ -50,7 +50,7 @@ BLOCK_OBJS+=block-qcow2.o block-parallel
 # system emulation, i.e. a single QEMU executable should support all
 # CPUs and machines.
 
-OBJS=$(BLOCK_OBJS)
+OBJS=nbd.o $(BLOCK_OBJS)
 OBJS+=readline.o console.o
 OBJS+=block.o
 
@@ -154,7 +154,7 @@ libqemu_user.a: $(USER_OBJS)
 	rm -f $@ 
 	$(AR) rcs $@ $(USER_OBJS)
 
-QEMU_IMG_BLOCK_OBJS = $(BLOCK_OBJS)
+QEMU_IMG_BLOCK_OBJS = nbd.o $(BLOCK_OBJS)
 ifdef CONFIG_WIN32
 QEMU_IMG_BLOCK_OBJS += qemu-img-block-raw-win32.o
 else
@@ -175,7 +175,7 @@ qemu-img-%.o: %.c
 qemu-nbd-%.o: %.c
 	$(CC) $(CFLAGS) $(CPPFLAGS) -DQEMU_NBD -c -o $@ $<
 
-qemu-nbd$(EXESUF):  qemu-nbd.o nbd.o qemu-img-block.o \
+qemu-nbd$(EXESUF):  qemu-nbd.o qemu-nbd-nbd.o qemu-img-block.o \
 		    osdep.o qemu-nbd-block-raw-posix.o $(BLOCK_OBJS)
 	$(CC) $(LDFLAGS) -o $@ $^ -lz $(LIBS)
 
Index: qemu/nbd.c
===================================================================
--- qemu.orig/nbd.c	2008-06-27 11:35:11.000000000 +0200
+++ qemu/nbd.c	2008-06-27 11:35:14.000000000 +0200
@@ -31,17 +31,21 @@
 #include <arpa/inet.h>
 #include <netdb.h>
 
+#if defined(QEMU_NBD)
 extern int verbose;
+#else
+static int verbose = 0;
+#endif
+
+#define TRACE(msg, ...) do { \
+    if (verbose) LOG(msg, ## __VA_ARGS__); \
+} while(0)
 
 #define LOG(msg, ...) do { \
     fprintf(stderr, "%s:%s():L%d: " msg "\n", \
             __FILE__, __FUNCTION__, __LINE__, ## __VA_ARGS__); \
 } while(0)
 
-#define TRACE(msg, ...) do { \
-    if (verbose) LOG(msg, ## __VA_ARGS__); \
-} while(0)
-
 /* This is all part of the "official" NBD API */
 
 #define NBD_REQUEST_MAGIC       0x25609513
@@ -59,10 +63,10 @@ extern int verbose;
 
 /* That's all folks */
 
-#define read_sync(fd, buffer, size) wr_sync(fd, buffer, size, true)
-#define write_sync(fd, buffer, size) wr_sync(fd, buffer, size, false)
+#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)
 
-static size_t wr_sync(int fd, void *buffer, size_t size, bool do_read)
+size_t nbd_wr_sync(int fd, void *buffer, size_t size, bool do_read)
 {
     size_t offset = 0;
 
@@ -76,7 +80,7 @@ static size_t wr_sync(int fd, void *buff
         }
 
         /* recoverable error */
-        if (len == -1 && errno == EAGAIN) {
+        if (len == -1 && (errno == EAGAIN || errno == EINTR)) {
             continue;
         }
 
@@ -96,7 +100,7 @@ static size_t wr_sync(int fd, void *buff
     return offset;
 }
 
-static int tcp_socket_outgoing(const char *address, uint16_t port)
+int tcp_socket_outgoing(const char *address, uint16_t port)
 {
     int s;
     struct in_addr in;
@@ -404,15 +408,31 @@ int nbd_client(int fd, int csock)
 	return ret;
 }
 
-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)
+int nbd_send_request(int csock, struct nbd_request *request)
 {
 	uint8_t buf[4 + 4 + 8 + 8 + 4];
-	uint32_t magic;
-	uint32_t type;
-	uint64_t from;
-	uint32_t len;
 
-	TRACE("Reading request.");
+	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");
+
+	if (write_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
+		LOG("writing to socket failed");
+		errno = EINVAL;
+		return -1;
+	}
+	return 0;
+}
+
+
+static int nbd_receive_request(int csock, struct nbd_request *request)
+{
+	uint8_t buf[4 + 4 + 8 + 8 + 4];
+	uint32_t magic;
 
 	if (read_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
 		LOG("read failed");
@@ -429,89 +449,150 @@ int nbd_trip(BlockDriverState *bs, int c
 	 */
 
 	magic = be32_to_cpup((uint32_t*)buf);
-	type  = be32_to_cpup((uint32_t*)(buf + 4));
-	from  = be64_to_cpup((uint64_t*)(buf + 16));
-	len   = be32_to_cpup((uint32_t*)(buf + 24));
+	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, type, from, len);
-
+	      magic, request->type, request->from, request->len);
 
 	if (magic != NBD_REQUEST_MAGIC) {
 		LOG("invalid magic (got 0x%x)", magic);
 		errno = EINVAL;
 		return -1;
 	}
+}
+
+int nbd_receive_reply(int csock, struct nbd_reply *reply)
+{
+	uint8_t buf[4 + 4 + 8];
+	uint32_t magic;
+int i;
+
+for (i = 0; i < sizeof(buf); i++) buf[i] = 0xAA;
+
+	if (read_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
+		LOG("read failed");
+		errno = EINVAL;
+		return -1;
+	}
 
-	if (len > data_size) {
+	/* Reply
+	 [ 0 ..  3]    magic   (NBD_REPLY_MAGIC)
+	 [ 4 ..  7]    error   (0 == no error)
+         [ 7 .. 15]    handle
+	 */
+
+	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("Got reply: "
+	      "{ magic = 0x%x, .error = %d, handle = %" PRIu64" }",
+	      magic, reply->error, reply->handle);
+
+	if (magic != NBD_REPLY_MAGIC) {
+		LOG("invalid magic (got 0x%x)", magic);
+		errno = EINVAL;
+		return -1;
+	}
+	return 0;
+}
+
+static int nbd_send_reply(int csock, struct nbd_reply *reply)
+{
+	uint8_t buf[4 + 4 + 8];
+
+	/* 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)",
-		    len, data_size);
+		    request.len, data_size);
 		errno = EINVAL;
 		return -1;
 	}
 
-	if ((from + len) < from) {
+	if ((request.from + request.len) < request.from) {
 		LOG("integer overflow detected! "
 		    "you're probably being attacked");
 		errno = EINVAL;
 		return -1;
 	}
 
-	if ((from + len) > size) {
+	if ((request.from + request.len) > size) {
 	        LOG("From: %" PRIu64 ", Len: %u, Size: %" PRIu64
 		    ", Offset: %" PRIu64 "\n",
-		     from, len, size, dev_offset);
+		     request.from, request.len, size, dev_offset);
 		LOG("requested operation past EOF--bad client?");
 		errno = EINVAL;
 		return -1;
 	}
 
-	/* 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), 0);
-
 	TRACE("Decoding type");
 
-	switch (type) {
-	case 0:
+	reply.handle = request.handle;
+	reply.error = 0;
+
+	switch (request.type) {
+	case NBD_CMD_READ:
 		TRACE("Request type is READ");
 
-		if (bdrv_read(bs, (from + dev_offset) / 512, data, len / 512) == -1) {
+		if (bdrv_read(bs, (request.from + dev_offset) / 512, data,
+			      request.len / 512) == -1) {
 			LOG("reading from file failed");
 			errno = EINVAL;
 			return -1;
 		}
-		*offset += len;
-
-		TRACE("Read %u byte(s)", len);
+		*offset += request.len;
 
-		TRACE("Sending OK response");
+		TRACE("Read %u byte(s)", request.len);
 
-		if (write_sync(csock, buf, 16) != 16) {
-			LOG("writing to socket failed");
-			errno = EINVAL;
+		if (nbd_send_reply(csock, &reply) == -1)
 			return -1;
-		}
 
 		TRACE("Sending data to client");
 
-		if (write_sync(csock, data, len) != len) {
+		if (write_sync(csock, data, request.len) != request.len) {
 			LOG("writing to socket failed");
 			errno = EINVAL;
 			return -1;
 		}
 		break;
-	case 1:
+	case NBD_CMD_WRITE:
 		TRACE("Request type is WRITE");
 
-		TRACE("Reading %u byte(s)", len);
+		TRACE("Reading %u byte(s)", request.len);
 
-		if (read_sync(csock, data, len) != len) {
+		if (read_sync(csock, data, request.len) != request.len) {
 			LOG("reading from socket failed");
 			errno = EINVAL;
 			return -1;
@@ -519,34 +600,29 @@ int nbd_trip(BlockDriverState *bs, int c
 
 		if (readonly) {
 			TRACE("Server is read-only, return error");
-
-			cpu_to_be32w((uint32_t*)(buf + 4), 1);
+			reply.error = 1;
 		} else {
 			TRACE("Writing to device");
 
-			if (bdrv_write(bs, (from + dev_offset) / 512, data, len / 512) == -1) {
+			if (bdrv_write(bs, (request.from + dev_offset) / 512,
+				       data, request.len / 512) == -1) {
 				LOG("writing to file failed");
 				errno = EINVAL;
 				return -1;
 			}
 
-			*offset += len;
+			*offset += request.len;
 		}
 
-		TRACE("Sending response to client");
-
-		if (write_sync(csock, buf, 16) != 16) {
-			LOG("writing to socket failed");
-			errno = EINVAL;
+		if (nbd_send_reply(csock, &reply) == -1)
 			return -1;
-		}
 		break;
-	case 2:
+	case NBD_CMD_DISC:
 		TRACE("Request type is DISCONNECT");
 		errno = 0;
 		return 1;
 	default:
-		LOG("invalid request type (%u) received", type);
+		LOG("invalid request type (%u) received", request.type);
 		errno = EINVAL;
 		return -1;
 	}
Index: qemu/nbd.h
===================================================================
--- qemu.orig/nbd.h	2008-06-27 11:35:11.000000000 +0200
+++ qemu/nbd.h	2008-06-27 11:35:14.000000000 +0200
@@ -26,6 +26,26 @@
 #include <qemu-common.h>
 #include "block_int.h"
 
+struct nbd_request {
+    uint32_t type;
+    uint64_t handle;
+    uint64_t from;
+    uint32_t len;
+};
+
+struct nbd_reply {
+    uint32_t error;
+    uint64_t handle;
+};
+
+enum {
+    NBD_CMD_READ = 0,
+    NBD_CMD_WRITE = 1,
+    NBD_CMD_DISC = 2
+};
+
+size_t nbd_wr_sync(int fd, void *buffer, size_t size, bool do_read);
+int tcp_socket_outgoing(const char *address, uint16_t port);
 int tcp_socket_incoming(const char *address, uint16_t port);
 int unix_socket_outgoing(const char *path);
 int unix_socket_incoming(const char *path);
@@ -33,6 +53,8 @@ int unix_socket_incoming(const char *pat
 int nbd_negotiate(BlockDriverState *bs, int csock, off_t size);
 int nbd_receive_negotiate(int csock, off_t *size, size_t *blocksize);
 int nbd_init(int fd, int csock, off_t size, size_t blocksize);
+int nbd_send_request(int csock, struct nbd_request *request);
+int nbd_receive_reply(int csock, struct nbd_reply *reply);
 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);
 int nbd_client(int fd, int csock);
Index: qemu/block.c
===================================================================
--- qemu.orig/block.c	2008-06-27 11:35:11.000000000 +0200
+++ qemu/block.c	2008-06-27 11:35:14.000000000 +0200
@@ -1332,6 +1332,7 @@ void bdrv_init(void)
     bdrv_register(&bdrv_vvfat);
     bdrv_register(&bdrv_qcow2);
     bdrv_register(&bdrv_parallels);
+    bdrv_register(&bdrv_nbd);
 }
 
 void *qemu_aio_get(BlockDriverState *bs, BlockDriverCompletionFunc *cb,
Index: qemu/block.h
===================================================================
--- qemu.orig/block.h	2008-06-27 11:35:11.000000000 +0200
+++ qemu/block.h	2008-06-27 11:35:14.000000000 +0200
@@ -16,6 +16,7 @@ extern BlockDriver bdrv_vpc;
 extern BlockDriver bdrv_vvfat;
 extern BlockDriver bdrv_qcow2;
 extern BlockDriver bdrv_parallels;
+extern BlockDriver bdrv_nbd;
 
 typedef struct BlockDriverInfo {
     /* in bytes, 0 if irrelevant */
Index: qemu/qemu-nbd.c
===================================================================
--- qemu.orig/qemu-nbd.c	2008-06-27 11:35:11.000000000 +0200
+++ qemu/qemu-nbd.c	2008-06-27 11:35:14.000000000 +0200
@@ -56,6 +56,7 @@ static void usage(const char *name)
 "  -c, --connect=DEV    connect FILE to the local NBD device DEV\n"
 "  -d, --disconnect     disconnect the specified device\n"
 "  -e, --shared=NUM     device can be shared by NUM clients (default '1')\n"
+"  -t, --persistent     don't exit on the last connection\n"
 "  -v, --verbose        display extra debugging information\n"
 "  -h, --help           display this help and exit\n"
 "  -V, --version        output version information and exit\n"
@@ -189,7 +190,7 @@ int main(int argc, char **argv)
     char *device = NULL;
     char *socket = NULL;
     char sockpath[128];
-    const char *sopt = "hVbo:p:rsnP:c:dvk:e:";
+    const char *sopt = "hVbo:p:rsnP:c:dvk:e:t";
     struct option lopt[] = {
         { "help", 0, 0, 'h' },
         { "version", 0, 0, 'V' },
@@ -204,6 +205,7 @@ int main(int argc, char **argv)
         { "snapshot", 0, 0, 's' },
         { "nocache", 0, 0, 'n' },
         { "shared", 1, 0, 'e' },
+        { "persistent", 0, 0, 't' },
         { "verbose", 0, 0, 'v' },
         { NULL, 0, 0, 0 }
     };
@@ -222,6 +224,7 @@ int main(int argc, char **argv)
     int i;
     int nb_fds = 0;
     int max_fd;
+    int persistent = 0;
 
     while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
         switch (ch) {
@@ -283,6 +286,9 @@ int main(int argc, char **argv)
                 errx(EINVAL, "Shared device number must be greater than 0\n");
             }
             break;
+	case 't':
+	    persistent = 1;
+	    break;
         case 'v':
             verbose = 1;
             break;
@@ -459,7 +465,7 @@ exit:
                 }
             }
         }
-    } while (nb_fds > 1);
+    } while (persistent || nb_fds > 1);
     qemu_free(data);
 
     close(sharing_fds[0]);
Index: qemu/qemu-nbd.texi
===================================================================
--- qemu.orig/qemu-nbd.texi	2008-06-27 11:35:11.000000000 +0200
+++ qemu/qemu-nbd.texi	2008-06-27 11:35:14.000000000 +0200
@@ -36,6 +36,8 @@ Export Qemu disk image using NBD protoco
   disconnect the specified device
 @item -e, --shared=NUM
   device can be shared by NUM clients (default '1')
+@item -t, --persistent
+  don't exit on the last connection
 @item -v, --verbose
   display extra debugging information
 @item -h, --help
Index: qemu/qemu-doc.texi
===================================================================
--- qemu.orig/qemu-doc.texi	2008-06-27 11:35:11.000000000 +0200
+++ qemu/qemu-doc.texi	2008-06-27 11:35:14.000000000 +0200
@@ -1310,6 +1310,7 @@ snapshots.
 * qemu_nbd_invocation::       qemu-nbd Invocation
 * host_drives::               Using host drives
 * disk_images_fat_images::    Virtual FAT disk images
+* disk_images_nbd::           NBD access
 @end menu
 
 @node disk_images_quickstart
@@ -1492,6 +1493,40 @@ What you should @emph{never} do:
 @item write to the FAT directory on the host system while accessing it with the guest system.
 @end itemize
 
+@node disk_images_nbd
+@subsection NBD access
+
+QEMU can access directly to block device exported using the Network Block Device
+protocol.
+
+@example
+qemu linux.img -hdb nbd:my_nbd_server.mydomain.org:1024
+@end example
+
+If the NBD server is located on the same host, you can use an unix socket instead
+of an inet socket:
+
+@example
+qemu linux.img -hdb nbd:unix:/tmp/my_socket
+@end example
+
+In this case, the block device must be exported using qemu-nbd:
+
+@example
+qemu-nbd --socket=/tmp/my_socket my_disk.qcow2
+@end example
+
+The use of qemu-nbd allows to share a disk between several guests:
+@example
+qemu-nbd --socket=/tmp/my_socket --share=2 my_disk.qcow2
+@end example
+
+and then you can use it with two guests:
+@example
+qemu linux1.img -hdb nbd:unix:/tmp/my_socket
+qemu linux2.img -hdb nbd:unix:/tmp/my_socket
+@end example
+
 @node pcsys_network
 @section Network emulation
 

--

Re: [Qemu-devel] [patch 2/5] qemu-nbd: merge NBD client/server

[Thiemo Seufer]

Laurent.Vivier@bull.net wrote:
> This patch allows to connect directly a disk image file to an NBD
> device. It introduces the use of a unix socket (instead of inet).
> 
> - To connect a file to a device:
> 
> # qemu-nbd --connect=/dev/nbd0 my_disk.qcow2
> 
> Then you can see directly your disk (no need of nbd-client):
> 
> # fdisk -l /dev/nbd0
> 
> Disk /dev/nbd0: 4294 MB, 4294967296 bytes
> 255 heads, 63 sectors/track, 522 cylinders
> Units = cylinders of 16065 * 512 = 8225280 bytes
> 
>      Device Boot      Start         End      Blocks   Id  System
> /dev/nbd0p1   *           1         492     3951958+  83  Linux
> /dev/nbd0p2             493         522      240975    5  Extended
> /dev/nbd0p5             493         522      240943+  82  Linux swap /
> Solaris
> 
> - To disconnect the file from the device:
> 
> # qemu-nbd --disconnect /dev/nbd0
> 
> Changelog:
> - v2: call show_parts() from client and avoid the sleep(1). Thank you to
> Avi and Anthony. Include my cleanup patch and comments from Carlo
> Marcelo Arenas Belon.
> - v3: allow to specify unix socket name with "--socket" and update 
> documentation.

There's still a call to sleep() in this patch.


Thiemo

Re: [Qemu-devel] [patch 2/5] qemu-nbd: merge NBD client/server

[Kevin Wolf]

Thiemo Seufer schrieb:
> Laurent.Vivier@bull.net wrote:
>> This patch allows to connect directly a disk image file to an NBD
>> device. It introduces the use of a unix socket (instead of inet).
>>
>> - To connect a file to a device:
>>
>> # qemu-nbd --connect=/dev/nbd0 my_disk.qcow2
>>
>> Then you can see directly your disk (no need of nbd-client):
>>
>> # fdisk -l /dev/nbd0
>>
>> Disk /dev/nbd0: 4294 MB, 4294967296 bytes
>> 255 heads, 63 sectors/track, 522 cylinders
>> Units = cylinders of 16065 * 512 = 8225280 bytes
>>
>>      Device Boot      Start         End      Blocks   Id  System
>> /dev/nbd0p1   *           1         492     3951958+  83  Linux
>> /dev/nbd0p2             493         522      240975    5  Extended
>> /dev/nbd0p5             493         522      240943+  82  Linux swap /
>> Solaris
>>
>> - To disconnect the file from the device:
>>
>> # qemu-nbd --disconnect /dev/nbd0
>>
>> Changelog:
>> - v2: call show_parts() from client and avoid the sleep(1). Thank you to
>> Avi and Anthony. Include my cleanup patch and comments from Carlo
>> Marcelo Arenas Belon.
>> - v3: allow to specify unix socket name with "--socket" and update 
>> documentation.
> 
> There's still a call to sleep() in this patch.

And that's fine, IMHO. sleep() is not bad per se. The remaining call is
not a "guess it'll take at most a second" style sleep but waiting in a
loop until the child is ready. You don't want the loop to eat up 100%
CPU, do you?

Kevin

Re: [Qemu-devel] [patch 2/5] qemu-nbd: merge NBD client/server

[Laurent Vivier]

Le jeudi 03 juillet 2008 à 09:21 +0200, Kevin Wolf a écrit :
> Thiemo Seufer schrieb:
> > Laurent.Vivier@bull.net wrote:
> >> This patch allows to connect directly a disk image file to an NBD
> >> device. It introduces the use of a unix socket (instead of inet).
> >>
> >> - To connect a file to a device:
> >>
> >> # qemu-nbd --connect=/dev/nbd0 my_disk.qcow2
> >>
> >> Then you can see directly your disk (no need of nbd-client):
> >>
> >> # fdisk -l /dev/nbd0
> >>
> >> Disk /dev/nbd0: 4294 MB, 4294967296 bytes
> >> 255 heads, 63 sectors/track, 522 cylinders
> >> Units = cylinders of 16065 * 512 = 8225280 bytes
> >>
> >>      Device Boot      Start         End      Blocks   Id  System
> >> /dev/nbd0p1   *           1         492     3951958+  83  Linux
> >> /dev/nbd0p2             493         522      240975    5  Extended
> >> /dev/nbd0p5             493         522      240943+  82  Linux swap /
> >> Solaris
> >>
> >> - To disconnect the file from the device:
> >>
> >> # qemu-nbd --disconnect /dev/nbd0
> >>
> >> Changelog:
> >> - v2: call show_parts() from client and avoid the sleep(1). Thank you to
> >> Avi and Anthony. Include my cleanup patch and comments from Carlo
> >> Marcelo Arenas Belon.
> >> - v3: allow to specify unix socket name with "--socket" and update 
> >> documentation.
> > 
> > There's still a call to sleep() in this patch.
> 
> And that's fine, IMHO. sleep() is not bad per se. The remaining call is
> not a "guess it'll take at most a second" style sleep but waiting in a
> loop until the child is ready. You don't want the loop to eat up 100%
> CPU, do you?

I agree.

Laurent
-- 
------------- Laurent.Vivier@bull.net ---------------
"The best way to predict the future is to invent it."
- Alan Kay

Re: [Qemu-devel] [patch 2/5] qemu-nbd: merge NBD client/server

[Thiemo Seufer]

Laurent Vivier wrote:
> Le jeudi 03 juillet 2008 à 09:21 +0200, Kevin Wolf a écrit :
> > Thiemo Seufer schrieb:
> > > Laurent.Vivier@bull.net wrote:
> > >> This patch allows to connect directly a disk image file to an NBD
> > >> device. It introduces the use of a unix socket (instead of inet).
> > >>
> > >> - To connect a file to a device:
> > >>
> > >> # qemu-nbd --connect=/dev/nbd0 my_disk.qcow2
> > >>
> > >> Then you can see directly your disk (no need of nbd-client):
> > >>
> > >> # fdisk -l /dev/nbd0
> > >>
> > >> Disk /dev/nbd0: 4294 MB, 4294967296 bytes
> > >> 255 heads, 63 sectors/track, 522 cylinders
> > >> Units = cylinders of 16065 * 512 = 8225280 bytes
> > >>
> > >>      Device Boot      Start         End      Blocks   Id  System
> > >> /dev/nbd0p1   *           1         492     3951958+  83  Linux
> > >> /dev/nbd0p2             493         522      240975    5  Extended
> > >> /dev/nbd0p5             493         522      240943+  82  Linux swap /
> > >> Solaris
> > >>
> > >> - To disconnect the file from the device:
> > >>
> > >> # qemu-nbd --disconnect /dev/nbd0
> > >>
> > >> Changelog:
> > >> - v2: call show_parts() from client and avoid the sleep(1). Thank you to
> > >> Avi and Anthony. Include my cleanup patch and comments from Carlo
> > >> Marcelo Arenas Belon.
> > >> - v3: allow to specify unix socket name with "--socket" and update 
> > >> documentation.
> > > 
> > > There's still a call to sleep() in this patch.
> > 
> > And that's fine, IMHO. sleep() is not bad per se. The remaining call is
> > not a "guess it'll take at most a second" style sleep but waiting in a
> > loop until the child is ready. You don't want the loop to eat up 100%
> > CPU, do you?
> 
> I agree.

I committed it with a small change (exit is not the best name for a label).


Thiemo