]> git.proxmox.com Git - qemu.git/blobdiff - savevm.c
qemu-timer: Introduce clock reset notifier
[qemu.git] / savevm.c
index 95947a9495992c3a4d481d3eb832ed13a6db67b5..8139bc7e297db9604bfc7a607025a435b6aec75f 100644 (file)
--- a/savevm.c
+++ b/savevm.c
@@ -23,7 +23,6 @@
  */
 #include <unistd.h>
 #include <fcntl.h>
-#include <signal.h>
 #include <time.h>
 #include <errno.h>
 #include <sys/time.h>
 #include <sys/socket.h>
 #include <netinet/in.h>
 #include <net/if.h>
-#if defined(__NetBSD__)
-#include <net/if_tap.h>
-#endif
-#ifdef __linux__
-#include <linux/if_tun.h>
-#endif
 #include <arpa/inet.h>
 #include <dirent.h>
 #include <netdb.h>
 #include <sys/select.h>
 #ifdef CONFIG_BSD
 #include <sys/stat.h>
-#if defined(__FreeBSD__) || defined(__DragonFly__)
+#if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) || defined(__DragonFly__)
 #include <libutil.h>
 #else
 #include <util.h>
 #endif
-#elif defined (__GLIBC__) && defined (__FreeBSD_kernel__)
-#include <freebsd/stdlib.h>
-#else
 #ifdef __linux__
 #include <pty.h>
 #include <malloc.h>
 
 #include "qemu-common.h"
 #include "hw/hw.h"
+#include "hw/qdev.h"
 #include "net.h"
 #include "monitor.h"
 #include "sysemu.h"
 #include "qemu-timer.h"
 #include "qemu-char.h"
-#include "block.h"
 #include "audio/audio.h"
 #include "migration.h"
 #include "qemu_socket.h"
-
-/* point to the block driver where the snapshots are managed */
-static BlockDriverState *bs_snapshots;
+#include "qemu-queue.h"
+#include "cpus.h"
 
 #define SELF_ANNOUNCE_ROUNDS 5
-#define ETH_P_EXPERIMENTAL 0x01F1 /* just a number */
-//#define ETH_P_EXPERIMENTAL 0x0012 /* make it the size of the packet */
-#define EXPERIMENTAL_MAGIC 0xf1f23f4f
 
-static int announce_self_create(uint8_t *buf, 
+#ifndef ETH_P_RARP
+#define ETH_P_RARP 0x8035
+#endif
+#define ARP_HTYPE_ETH 0x0001
+#define ARP_PTYPE_IP 0x0800
+#define ARP_OP_REQUEST_REV 0x3
+
+static int announce_self_create(uint8_t *buf,
                                uint8_t *mac_addr)
 {
-    uint32_t magic = EXPERIMENTAL_MAGIC;
-    uint16_t proto = htons(ETH_P_EXPERIMENTAL);
+    /* Ethernet header. */
+    memset(buf, 0xff, 6);         /* destination MAC addr */
+    memcpy(buf + 6, mac_addr, 6); /* source MAC addr */
+    *(uint16_t *)(buf + 12) = htons(ETH_P_RARP); /* ethertype */
+
+    /* RARP header. */
+    *(uint16_t *)(buf + 14) = htons(ARP_HTYPE_ETH); /* hardware addr space */
+    *(uint16_t *)(buf + 16) = htons(ARP_PTYPE_IP); /* protocol addr space */
+    *(buf + 18) = 6; /* hardware addr length (ethernet) */
+    *(buf + 19) = 4; /* protocol addr length (IPv4) */
+    *(uint16_t *)(buf + 20) = htons(ARP_OP_REQUEST_REV); /* opcode */
+    memcpy(buf + 22, mac_addr, 6); /* source hw addr */
+    memset(buf + 28, 0x00, 4);     /* source protocol addr */
+    memcpy(buf + 32, mac_addr, 6); /* target hw addr */
+    memset(buf + 38, 0x00, 4);     /* target protocol addr */
+
+    /* Padding to get up to 60 bytes (ethernet min packet size, minus FCS). */
+    memset(buf + 42, 0x00, 18);
+
+    return 60; /* len (FCS will be added by hardware) */
+}
 
-    /* FIXME: should we send a different packet (arp/rarp/ping)? */
+static void qemu_announce_self_iter(NICState *nic, void *opaque)
+{
+    uint8_t buf[60];
+    int len;
 
-    memset(buf, 0, 64);
-    memset(buf, 0xff, 6);         /* h_dst */
-    memcpy(buf + 6, mac_addr, 6); /* h_src */
-    memcpy(buf + 12, &proto, 2);  /* h_proto */
-    memcpy(buf + 14, &magic, 4);  /* magic */
+    len = announce_self_create(buf, nic->conf->macaddr.a);
 
-    return 64; /* len */
+    qemu_send_packet_raw(&nic->nc, buf, len);
 }
 
+
 static void qemu_announce_self_once(void *opaque)
 {
-    int i, len;
-    VLANState *vlan;
-    VLANClientState *vc;
-    uint8_t buf[256];
     static int count = SELF_ANNOUNCE_ROUNDS;
     QEMUTimer *timer = *(QEMUTimer **)opaque;
 
-    for (i = 0; i < MAX_NICS; i++) {
-        if (!nd_table[i].used)
-            continue;
-        len = announce_self_create(buf, nd_table[i].macaddr);
-        vlan = nd_table[i].vlan;
-       for(vc = vlan->first_client; vc != NULL; vc = vc->next) {
-            vc->receive(vc, buf, len);
-        }
-    }
-    if (count--) {
-           qemu_mod_timer(timer, qemu_get_clock(rt_clock) + 100);
+    qemu_foreach_nic(qemu_announce_self_iter, NULL);
+
+    if (--count) {
+        /* delay 50ms, 150ms, 250ms, ... */
+        qemu_mod_timer(timer, qemu_get_clock_ms(rt_clock) +
+                       50 + (SELF_ANNOUNCE_ROUNDS - count - 1) * 100);
     } else {
            qemu_del_timer(timer);
            qemu_free_timer(timer);
@@ -145,7 +148,7 @@ static void qemu_announce_self_once(void *opaque)
 void qemu_announce_self(void)
 {
        static QEMUTimer *timer;
-       timer = qemu_new_timer(rt_clock, qemu_announce_self_once, &timer);
+       timer = qemu_new_timer_ms(rt_clock, qemu_announce_self_once, &timer);
        qemu_announce_self_once(&timer);
 }
 
@@ -160,6 +163,7 @@ struct QEMUFile {
     QEMUFileCloseFunc *close;
     QEMUFileRateLimit *rate_limit;
     QEMUFileSetRateLimit *set_rate_limit;
+    QEMUFileGetRateLimit *get_rate_limit;
     void *opaque;
     int is_write;
 
@@ -228,9 +232,10 @@ static int stdio_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 static int stdio_pclose(void *opaque)
 {
     QEMUFileStdio *s = opaque;
-    pclose(s->stdio_file);
+    int ret;
+    ret = pclose(s->stdio_file);
     qemu_free(s);
-    return 0;
+    return ret;
 }
 
 static int stdio_fclose(void *opaque)
@@ -255,9 +260,11 @@ QEMUFile *qemu_popen(FILE *stdio_file, const char *mode)
     s->stdio_file = stdio_file;
 
     if(mode[0] == 'r') {
-        s->file = qemu_fopen_ops(s, NULL, stdio_get_buffer, stdio_pclose, NULL, NULL);
+        s->file = qemu_fopen_ops(s, NULL, stdio_get_buffer, stdio_pclose, 
+                                NULL, NULL, NULL);
     } else {
-        s->file = qemu_fopen_ops(s, stdio_put_buffer, NULL, stdio_pclose, NULL, NULL);
+        s->file = qemu_fopen_ops(s, stdio_put_buffer, NULL, stdio_pclose, 
+                                NULL, NULL, NULL);
     }
     return s->file;
 }
@@ -302,9 +309,11 @@ QEMUFile *qemu_fdopen(int fd, const char *mode)
         goto fail;
 
     if(mode[0] == 'r') {
-        s->file = qemu_fopen_ops(s, NULL, stdio_get_buffer, stdio_fclose, NULL, NULL);
+        s->file = qemu_fopen_ops(s, NULL, stdio_get_buffer, stdio_fclose, 
+                                NULL, NULL, NULL);
     } else {
-        s->file = qemu_fopen_ops(s, stdio_put_buffer, NULL, stdio_fclose, NULL, NULL);
+        s->file = qemu_fopen_ops(s, stdio_put_buffer, NULL, stdio_fclose, 
+                                NULL, NULL, NULL);
     }
     return s->file;
 
@@ -318,7 +327,8 @@ QEMUFile *qemu_fopen_socket(int fd)
     QEMUFileSocket *s = qemu_mallocz(sizeof(QEMUFileSocket));
 
     s->fd = fd;
-    s->file = qemu_fopen_ops(s, NULL, socket_get_buffer, socket_close, NULL, NULL);
+    s->file = qemu_fopen_ops(s, NULL, socket_get_buffer, socket_close, 
+                            NULL, NULL, NULL);
     return s->file;
 }
 
@@ -327,8 +337,7 @@ static int file_put_buffer(void *opaque, const uint8_t *buf,
 {
     QEMUFileStdio *s = opaque;
     fseek(s->stdio_file, pos, SEEK_SET);
-    fwrite(buf, 1, size, s->stdio_file);
-    return size;
+    return fwrite(buf, 1, size, s->stdio_file);
 }
 
 static int file_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
@@ -345,7 +354,7 @@ QEMUFile *qemu_fopen(const char *filename, const char *mode)
     if (mode == NULL ||
        (mode[0] != 'r' && mode[0] != 'w') ||
        mode[1] != 'b' || mode[2] != 0) {
-        fprintf(stderr, "qemu_fdopen: Argument validity check failed\n");
+        fprintf(stderr, "qemu_fopen: Argument validity check failed\n");
         return NULL;
     }
 
@@ -354,11 +363,13 @@ QEMUFile *qemu_fopen(const char *filename, const char *mode)
     s->stdio_file = fopen(filename, mode);
     if (!s->stdio_file)
         goto fail;
-
+    
     if(mode[0] == 'w') {
-        s->file = qemu_fopen_ops(s, file_put_buffer, NULL, stdio_fclose, NULL, NULL);
+        s->file = qemu_fopen_ops(s, file_put_buffer, NULL, stdio_fclose, 
+                                NULL, NULL, NULL);
     } else {
-        s->file = qemu_fopen_ops(s, NULL, file_get_buffer, stdio_fclose, NULL, NULL);
+        s->file = qemu_fopen_ops(s, NULL, file_get_buffer, stdio_fclose, 
+                              NULL, NULL, NULL);
     }
     return s->file;
 fail:
@@ -386,15 +397,17 @@ static int bdrv_fclose(void *opaque)
 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable)
 {
     if (is_writable)
-        return qemu_fopen_ops(bs, block_put_buffer, NULL, bdrv_fclose, NULL, NULL);
-    return qemu_fopen_ops(bs, NULL, block_get_buffer, bdrv_fclose, NULL, NULL);
+        return qemu_fopen_ops(bs, block_put_buffer, NULL, bdrv_fclose, 
+                             NULL, NULL, NULL);
+    return qemu_fopen_ops(bs, NULL, block_get_buffer, bdrv_fclose, NULL, NULL, NULL);
 }
 
 QEMUFile *qemu_fopen_ops(void *opaque, QEMUFilePutBufferFunc *put_buffer,
                          QEMUFileGetBufferFunc *get_buffer,
                          QEMUFileCloseFunc *close,
                          QEMUFileRateLimit *rate_limit,
-                         QEMUFileSetRateLimit *set_rate_limit)
+                         QEMUFileSetRateLimit *set_rate_limit,
+                         QEMUFileGetRateLimit *get_rate_limit)
 {
     QEMUFile *f;
 
@@ -406,6 +419,7 @@ QEMUFile *qemu_fopen_ops(void *opaque, QEMUFilePutBufferFunc *put_buffer,
     f->close = close;
     f->rate_limit = rate_limit;
     f->set_rate_limit = set_rate_limit;
+    f->get_rate_limit = get_rate_limit;
     f->is_write = 0;
 
     return f;
@@ -536,6 +550,19 @@ int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size1)
     return size1 - size;
 }
 
+static int qemu_peek_byte(QEMUFile *f)
+{
+    if (f->is_write)
+        abort();
+
+    if (f->buf_index >= f->buf_size) {
+        qemu_fill_buffer(f);
+        if (f->buf_index >= f->buf_size)
+            return 0;
+    }
+    return f->buf[f->buf_index];
+}
+
 int qemu_get_byte(QEMUFile *f)
 {
     if (f->is_write)
@@ -583,7 +610,15 @@ int qemu_file_rate_limit(QEMUFile *f)
     return 0;
 }
 
-size_t qemu_file_set_rate_limit(QEMUFile *f, size_t new_rate)
+int64_t qemu_file_get_rate_limit(QEMUFile *f)
+{
+    if (f->get_rate_limit)
+        return f->get_rate_limit(f->opaque);
+
+    return 0;
+}
+
+int64_t qemu_file_set_rate_limit(QEMUFile *f, int64_t new_rate)
 {
     /* any failed or completed migration keeps its state to allow probing of
      * migration data, but has no associated file anymore */
@@ -639,311 +674,1084 @@ uint64_t qemu_get_be64(QEMUFile *f)
     return v;
 }
 
-typedef struct SaveStateEntry {
-    char idstr[256];
-    int instance_id;
-    int version_id;
-    int section_id;
-    SaveLiveStateHandler *save_live_state;
-    SaveStateHandler *save_state;
-    LoadStateHandler *load_state;
-    void *opaque;
-    struct SaveStateEntry *next;
-} SaveStateEntry;
+/* bool */
 
-static SaveStateEntry *first_se;
+static int get_bool(QEMUFile *f, void *pv, size_t size)
+{
+    bool *v = pv;
+    *v = qemu_get_byte(f);
+    return 0;
+}
 
-/* TODO: Individual devices generally have very little idea about the rest
-   of the system, so instance_id should be removed/replaced.
-   Meanwhile pass -1 as instance_id if you do not already have a clearly
-   distinguishing id for all instances of your device class. */
-int register_savevm_live(const char *idstr,
-                         int instance_id,
-                         int version_id,
-                         SaveLiveStateHandler *save_live_state,
-                         SaveStateHandler *save_state,
-                         LoadStateHandler *load_state,
-                         void *opaque)
+static void put_bool(QEMUFile *f, void *pv, size_t size)
 {
-    SaveStateEntry *se, **pse;
-    static int global_section_id;
+    bool *v = pv;
+    qemu_put_byte(f, *v);
+}
 
-    se = qemu_malloc(sizeof(SaveStateEntry));
-    pstrcpy(se->idstr, sizeof(se->idstr), idstr);
-    se->instance_id = (instance_id == -1) ? 0 : instance_id;
-    se->version_id = version_id;
-    se->section_id = global_section_id++;
-    se->save_live_state = save_live_state;
-    se->save_state = save_state;
-    se->load_state = load_state;
-    se->opaque = opaque;
-    se->next = NULL;
+const VMStateInfo vmstate_info_bool = {
+    .name = "bool",
+    .get  = get_bool,
+    .put  = put_bool,
+};
 
-    /* add at the end of list */
-    pse = &first_se;
-    while (*pse != NULL) {
-        if (instance_id == -1
-                && strcmp(se->idstr, (*pse)->idstr) == 0
-                && se->instance_id <= (*pse)->instance_id)
-            se->instance_id = (*pse)->instance_id + 1;
-        pse = &(*pse)->next;
-    }
-    *pse = se;
+/* 8 bit int */
+
+static int get_int8(QEMUFile *f, void *pv, size_t size)
+{
+    int8_t *v = pv;
+    qemu_get_s8s(f, v);
     return 0;
 }
 
-int register_savevm(const char *idstr,
-                    int instance_id,
-                    int version_id,
-                    SaveStateHandler *save_state,
-                    LoadStateHandler *load_state,
-                    void *opaque)
+static void put_int8(QEMUFile *f, void *pv, size_t size)
 {
-    return register_savevm_live(idstr, instance_id, version_id,
-                                NULL, save_state, load_state, opaque);
+    int8_t *v = pv;
+    qemu_put_s8s(f, v);
 }
 
-void unregister_savevm(const char *idstr, void *opaque)
+const VMStateInfo vmstate_info_int8 = {
+    .name = "int8",
+    .get  = get_int8,
+    .put  = put_int8,
+};
+
+/* 16 bit int */
+
+static int get_int16(QEMUFile *f, void *pv, size_t size)
 {
-    SaveStateEntry **pse;
+    int16_t *v = pv;
+    qemu_get_sbe16s(f, v);
+    return 0;
+}
 
-    pse = &first_se;
-    while (*pse != NULL) {
-        if (strcmp((*pse)->idstr, idstr) == 0 && (*pse)->opaque == opaque) {
-            SaveStateEntry *next = (*pse)->next;
-            qemu_free(*pse);
-            *pse = next;
-            continue;
-        }
-        pse = &(*pse)->next;
-    }
+static void put_int16(QEMUFile *f, void *pv, size_t size)
+{
+    int16_t *v = pv;
+    qemu_put_sbe16s(f, v);
 }
 
-#define QEMU_VM_FILE_MAGIC           0x5145564d
-#define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
-#define QEMU_VM_FILE_VERSION         0x00000003
+const VMStateInfo vmstate_info_int16 = {
+    .name = "int16",
+    .get  = get_int16,
+    .put  = put_int16,
+};
 
-#define QEMU_VM_EOF                  0x00
-#define QEMU_VM_SECTION_START        0x01
-#define QEMU_VM_SECTION_PART         0x02
-#define QEMU_VM_SECTION_END          0x03
-#define QEMU_VM_SECTION_FULL         0x04
+/* 32 bit int */
 
-int qemu_savevm_state_begin(QEMUFile *f)
+static int get_int32(QEMUFile *f, void *pv, size_t size)
 {
-    SaveStateEntry *se;
+    int32_t *v = pv;
+    qemu_get_sbe32s(f, v);
+    return 0;
+}
 
-    qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
-    qemu_put_be32(f, QEMU_VM_FILE_VERSION);
+static void put_int32(QEMUFile *f, void *pv, size_t size)
+{
+    int32_t *v = pv;
+    qemu_put_sbe32s(f, v);
+}
 
-    for (se = first_se; se != NULL; se = se->next) {
-        int len;
+const VMStateInfo vmstate_info_int32 = {
+    .name = "int32",
+    .get  = get_int32,
+    .put  = put_int32,
+};
 
-        if (se->save_live_state == NULL)
-            continue;
+/* 32 bit int. See that the received value is the same than the one
+   in the field */
 
-        /* Section type */
-        qemu_put_byte(f, QEMU_VM_SECTION_START);
-        qemu_put_be32(f, se->section_id);
+static int get_int32_equal(QEMUFile *f, void *pv, size_t size)
+{
+    int32_t *v = pv;
+    int32_t v2;
+    qemu_get_sbe32s(f, &v2);
 
-        /* ID string */
-        len = strlen(se->idstr);
-        qemu_put_byte(f, len);
-        qemu_put_buffer(f, (uint8_t *)se->idstr, len);
+    if (*v == v2)
+        return 0;
+    return -EINVAL;
+}
 
-        qemu_put_be32(f, se->instance_id);
-        qemu_put_be32(f, se->version_id);
+const VMStateInfo vmstate_info_int32_equal = {
+    .name = "int32 equal",
+    .get  = get_int32_equal,
+    .put  = put_int32,
+};
 
-        se->save_live_state(f, QEMU_VM_SECTION_START, se->opaque);
-    }
+/* 32 bit int. See that the received value is the less or the same
+   than the one in the field */
 
-    if (qemu_file_has_error(f))
-        return -EIO;
+static int get_int32_le(QEMUFile *f, void *pv, size_t size)
+{
+    int32_t *old = pv;
+    int32_t new;
+    qemu_get_sbe32s(f, &new);
+
+    if (*old <= new)
+        return 0;
+    return -EINVAL;
+}
 
+const VMStateInfo vmstate_info_int32_le = {
+    .name = "int32 equal",
+    .get  = get_int32_le,
+    .put  = put_int32,
+};
+
+/* 64 bit int */
+
+static int get_int64(QEMUFile *f, void *pv, size_t size)
+{
+    int64_t *v = pv;
+    qemu_get_sbe64s(f, v);
     return 0;
 }
 
-int qemu_savevm_state_iterate(QEMUFile *f)
+static void put_int64(QEMUFile *f, void *pv, size_t size)
 {
-    SaveStateEntry *se;
-    int ret = 1;
+    int64_t *v = pv;
+    qemu_put_sbe64s(f, v);
+}
 
-    for (se = first_se; se != NULL; se = se->next) {
-        if (se->save_live_state == NULL)
-            continue;
+const VMStateInfo vmstate_info_int64 = {
+    .name = "int64",
+    .get  = get_int64,
+    .put  = put_int64,
+};
 
-        /* Section type */
-        qemu_put_byte(f, QEMU_VM_SECTION_PART);
-        qemu_put_be32(f, se->section_id);
+/* 8 bit unsigned int */
 
-        ret &= !!se->save_live_state(f, QEMU_VM_SECTION_PART, se->opaque);
-    }
+static int get_uint8(QEMUFile *f, void *pv, size_t size)
+{
+    uint8_t *v = pv;
+    qemu_get_8s(f, v);
+    return 0;
+}
 
-    if (ret)
-        return 1;
+static void put_uint8(QEMUFile *f, void *pv, size_t size)
+{
+    uint8_t *v = pv;
+    qemu_put_8s(f, v);
+}
 
-    if (qemu_file_has_error(f))
-        return -EIO;
+const VMStateInfo vmstate_info_uint8 = {
+    .name = "uint8",
+    .get  = get_uint8,
+    .put  = put_uint8,
+};
 
+/* 16 bit unsigned int */
+
+static int get_uint16(QEMUFile *f, void *pv, size_t size)
+{
+    uint16_t *v = pv;
+    qemu_get_be16s(f, v);
     return 0;
 }
 
-int qemu_savevm_state_complete(QEMUFile *f)
+static void put_uint16(QEMUFile *f, void *pv, size_t size)
 {
-    SaveStateEntry *se;
-
-    for (se = first_se; se != NULL; se = se->next) {
-        if (se->save_live_state == NULL)
-            continue;
+    uint16_t *v = pv;
+    qemu_put_be16s(f, v);
+}
 
-        /* Section type */
-        qemu_put_byte(f, QEMU_VM_SECTION_END);
-        qemu_put_be32(f, se->section_id);
+const VMStateInfo vmstate_info_uint16 = {
+    .name = "uint16",
+    .get  = get_uint16,
+    .put  = put_uint16,
+};
 
-        se->save_live_state(f, QEMU_VM_SECTION_END, se->opaque);
-    }
+/* 32 bit unsigned int */
 
-    for(se = first_se; se != NULL; se = se->next) {
-        int len;
+static int get_uint32(QEMUFile *f, void *pv, size_t size)
+{
+    uint32_t *v = pv;
+    qemu_get_be32s(f, v);
+    return 0;
+}
 
-       if (se->save_state == NULL)
-           continue;
+static void put_uint32(QEMUFile *f, void *pv, size_t size)
+{
+    uint32_t *v = pv;
+    qemu_put_be32s(f, v);
+}
 
-        /* Section type */
-        qemu_put_byte(f, QEMU_VM_SECTION_FULL);
-        qemu_put_be32(f, se->section_id);
+const VMStateInfo vmstate_info_uint32 = {
+    .name = "uint32",
+    .get  = get_uint32,
+    .put  = put_uint32,
+};
 
-        /* ID string */
-        len = strlen(se->idstr);
-        qemu_put_byte(f, len);
-        qemu_put_buffer(f, (uint8_t *)se->idstr, len);
+/* 32 bit uint. See that the received value is the same than the one
+   in the field */
 
-        qemu_put_be32(f, se->instance_id);
-        qemu_put_be32(f, se->version_id);
+static int get_uint32_equal(QEMUFile *f, void *pv, size_t size)
+{
+    uint32_t *v = pv;
+    uint32_t v2;
+    qemu_get_be32s(f, &v2);
 
-        se->save_state(f, se->opaque);
+    if (*v == v2) {
+        return 0;
     }
+    return -EINVAL;
+}
 
-    qemu_put_byte(f, QEMU_VM_EOF);
+const VMStateInfo vmstate_info_uint32_equal = {
+    .name = "uint32 equal",
+    .get  = get_uint32_equal,
+    .put  = put_uint32,
+};
 
-    if (qemu_file_has_error(f))
-        return -EIO;
+/* 64 bit unsigned int */
 
+static int get_uint64(QEMUFile *f, void *pv, size_t size)
+{
+    uint64_t *v = pv;
+    qemu_get_be64s(f, v);
     return 0;
 }
 
-int qemu_savevm_state(QEMUFile *f)
+static void put_uint64(QEMUFile *f, void *pv, size_t size)
 {
-    int saved_vm_running;
-    int ret;
+    uint64_t *v = pv;
+    qemu_put_be64s(f, v);
+}
 
-    saved_vm_running = vm_running;
-    vm_stop(0);
+const VMStateInfo vmstate_info_uint64 = {
+    .name = "uint64",
+    .get  = get_uint64,
+    .put  = put_uint64,
+};
 
-    bdrv_flush_all();
+/* 8 bit int. See that the received value is the same than the one
+   in the field */
 
-    ret = qemu_savevm_state_begin(f);
-    if (ret < 0)
-        goto out;
+static int get_uint8_equal(QEMUFile *f, void *pv, size_t size)
+{
+    uint8_t *v = pv;
+    uint8_t v2;
+    qemu_get_8s(f, &v2);
 
-    do {
-        ret = qemu_savevm_state_iterate(f);
-        if (ret < 0)
-            goto out;
-    } while (ret == 0);
+    if (*v == v2)
+        return 0;
+    return -EINVAL;
+}
 
-    ret = qemu_savevm_state_complete(f);
+const VMStateInfo vmstate_info_uint8_equal = {
+    .name = "uint8 equal",
+    .get  = get_uint8_equal,
+    .put  = put_uint8,
+};
 
-out:
-    if (qemu_file_has_error(f))
-        ret = -EIO;
+/* 16 bit unsigned int int. See that the received value is the same than the one
+   in the field */
 
-    if (!ret && saved_vm_running)
-        vm_start();
+static int get_uint16_equal(QEMUFile *f, void *pv, size_t size)
+{
+    uint16_t *v = pv;
+    uint16_t v2;
+    qemu_get_be16s(f, &v2);
 
-    return ret;
+    if (*v == v2)
+        return 0;
+    return -EINVAL;
 }
 
-static SaveStateEntry *find_se(const char *idstr, int instance_id)
-{
-    SaveStateEntry *se;
+const VMStateInfo vmstate_info_uint16_equal = {
+    .name = "uint16 equal",
+    .get  = get_uint16_equal,
+    .put  = put_uint16,
+};
 
-    for(se = first_se; se != NULL; se = se->next) {
-        if (!strcmp(se->idstr, idstr) &&
-            instance_id == se->instance_id)
-            return se;
-    }
-    return NULL;
-}
+/* timers  */
 
-typedef struct LoadStateEntry {
-    SaveStateEntry *se;
-    int section_id;
-    int version_id;
-    struct LoadStateEntry *next;
-} LoadStateEntry;
+static int get_timer(QEMUFile *f, void *pv, size_t size)
+{
+    QEMUTimer *v = pv;
+    qemu_get_timer(f, v);
+    return 0;
+}
 
-static int qemu_loadvm_state_v2(QEMUFile *f)
+static void put_timer(QEMUFile *f, void *pv, size_t size)
 {
-    SaveStateEntry *se;
-    int len, ret, instance_id, record_len, version_id;
-    int64_t total_len, end_pos, cur_pos;
-    char idstr[256];
+    QEMUTimer *v = pv;
+    qemu_put_timer(f, v);
+}
 
-    total_len = qemu_get_be64(f);
-    end_pos = total_len + qemu_ftell(f);
-    for(;;) {
-        if (qemu_ftell(f) >= end_pos)
-            break;
-        len = qemu_get_byte(f);
-        qemu_get_buffer(f, (uint8_t *)idstr, len);
-        idstr[len] = '\0';
-        instance_id = qemu_get_be32(f);
-        version_id = qemu_get_be32(f);
-        record_len = qemu_get_be32(f);
-        cur_pos = qemu_ftell(f);
-        se = find_se(idstr, instance_id);
-        if (!se) {
-            fprintf(stderr, "qemu: warning: instance 0x%x of device '%s' not present in current VM\n",
-                    instance_id, idstr);
-        } else {
-            ret = se->load_state(f, se->opaque, version_id);
-            if (ret < 0) {
-                fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
-                        instance_id, idstr);
-                return ret;
-            }
-        }
-        /* always seek to exact end of record */
-        qemu_fseek(f, cur_pos + record_len, SEEK_SET);
-    }
+const VMStateInfo vmstate_info_timer = {
+    .name = "timer",
+    .get  = get_timer,
+    .put  = put_timer,
+};
 
-    if (qemu_file_has_error(f))
-        return -EIO;
+/* uint8_t buffers */
 
+static int get_buffer(QEMUFile *f, void *pv, size_t size)
+{
+    uint8_t *v = pv;
+    qemu_get_buffer(f, v, size);
     return 0;
 }
 
-int qemu_loadvm_state(QEMUFile *f)
+static void put_buffer(QEMUFile *f, void *pv, size_t size)
 {
-    LoadStateEntry *first_le = NULL;
-    uint8_t section_type;
-    unsigned int v;
-    int ret;
+    uint8_t *v = pv;
+    qemu_put_buffer(f, v, size);
+}
 
-    v = qemu_get_be32(f);
-    if (v != QEMU_VM_FILE_MAGIC)
-        return -EINVAL;
+const VMStateInfo vmstate_info_buffer = {
+    .name = "buffer",
+    .get  = get_buffer,
+    .put  = put_buffer,
+};
 
-    v = qemu_get_be32(f);
-    if (v == QEMU_VM_FILE_VERSION_COMPAT)
-        return qemu_loadvm_state_v2(f);
-    if (v != QEMU_VM_FILE_VERSION)
-        return -ENOTSUP;
+/* unused buffers: space that was used for some fields that are
+   not useful anymore */
 
-    while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
-        uint32_t instance_id, version_id, section_id;
-        LoadStateEntry *le;
+static int get_unused_buffer(QEMUFile *f, void *pv, size_t size)
+{
+    uint8_t buf[1024];
+    int block_len;
+
+    while (size > 0) {
+        block_len = MIN(sizeof(buf), size);
+        size -= block_len;
+        qemu_get_buffer(f, buf, block_len);
+    }
+   return 0;
+}
+
+static void put_unused_buffer(QEMUFile *f, void *pv, size_t size)
+{
+    static const uint8_t buf[1024];
+    int block_len;
+
+    while (size > 0) {
+        block_len = MIN(sizeof(buf), size);
+        size -= block_len;
+        qemu_put_buffer(f, buf, block_len);
+    }
+}
+
+const VMStateInfo vmstate_info_unused_buffer = {
+    .name = "unused_buffer",
+    .get  = get_unused_buffer,
+    .put  = put_unused_buffer,
+};
+
+typedef struct CompatEntry {
+    char idstr[256];
+    int instance_id;
+} CompatEntry;
+
+typedef struct SaveStateEntry {
+    QTAILQ_ENTRY(SaveStateEntry) entry;
+    char idstr[256];
+    int instance_id;
+    int alias_id;
+    int version_id;
+    int section_id;
+    SaveSetParamsHandler *set_params;
+    SaveLiveStateHandler *save_live_state;
+    SaveStateHandler *save_state;
+    LoadStateHandler *load_state;
+    const VMStateDescription *vmsd;
+    void *opaque;
+    CompatEntry *compat;
+    int no_migrate;
+} SaveStateEntry;
+
+
+static QTAILQ_HEAD(savevm_handlers, SaveStateEntry) savevm_handlers =
+    QTAILQ_HEAD_INITIALIZER(savevm_handlers);
+static int global_section_id;
+
+static int calculate_new_instance_id(const char *idstr)
+{
+    SaveStateEntry *se;
+    int instance_id = 0;
+
+    QTAILQ_FOREACH(se, &savevm_handlers, entry) {
+        if (strcmp(idstr, se->idstr) == 0
+            && instance_id <= se->instance_id) {
+            instance_id = se->instance_id + 1;
+        }
+    }
+    return instance_id;
+}
+
+static int calculate_compat_instance_id(const char *idstr)
+{
+    SaveStateEntry *se;
+    int instance_id = 0;
+
+    QTAILQ_FOREACH(se, &savevm_handlers, entry) {
+        if (!se->compat)
+            continue;
+
+        if (strcmp(idstr, se->compat->idstr) == 0
+            && instance_id <= se->compat->instance_id) {
+            instance_id = se->compat->instance_id + 1;
+        }
+    }
+    return instance_id;
+}
+
+/* TODO: Individual devices generally have very little idea about the rest
+   of the system, so instance_id should be removed/replaced.
+   Meanwhile pass -1 as instance_id if you do not already have a clearly
+   distinguishing id for all instances of your device class. */
+int register_savevm_live(DeviceState *dev,
+                         const char *idstr,
+                         int instance_id,
+                         int version_id,
+                         SaveSetParamsHandler *set_params,
+                         SaveLiveStateHandler *save_live_state,
+                         SaveStateHandler *save_state,
+                         LoadStateHandler *load_state,
+                         void *opaque)
+{
+    SaveStateEntry *se;
+
+    se = qemu_mallocz(sizeof(SaveStateEntry));
+    se->version_id = version_id;
+    se->section_id = global_section_id++;
+    se->set_params = set_params;
+    se->save_live_state = save_live_state;
+    se->save_state = save_state;
+    se->load_state = load_state;
+    se->opaque = opaque;
+    se->vmsd = NULL;
+    se->no_migrate = 0;
+
+    if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
+        char *id = dev->parent_bus->info->get_dev_path(dev);
+        if (id) {
+            pstrcpy(se->idstr, sizeof(se->idstr), id);
+            pstrcat(se->idstr, sizeof(se->idstr), "/");
+            qemu_free(id);
+
+            se->compat = qemu_mallocz(sizeof(CompatEntry));
+            pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr);
+            se->compat->instance_id = instance_id == -1 ?
+                         calculate_compat_instance_id(idstr) : instance_id;
+            instance_id = -1;
+        }
+    }
+    pstrcat(se->idstr, sizeof(se->idstr), idstr);
+
+    if (instance_id == -1) {
+        se->instance_id = calculate_new_instance_id(se->idstr);
+    } else {
+        se->instance_id = instance_id;
+    }
+    assert(!se->compat || se->instance_id == 0);
+    /* add at the end of list */
+    QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
+    return 0;
+}
+
+int register_savevm(DeviceState *dev,
+                    const char *idstr,
+                    int instance_id,
+                    int version_id,
+                    SaveStateHandler *save_state,
+                    LoadStateHandler *load_state,
+                    void *opaque)
+{
+    return register_savevm_live(dev, idstr, instance_id, version_id,
+                                NULL, NULL, save_state, load_state, opaque);
+}
+
+void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque)
+{
+    SaveStateEntry *se, *new_se;
+    char id[256] = "";
+
+    if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
+        char *path = dev->parent_bus->info->get_dev_path(dev);
+        if (path) {
+            pstrcpy(id, sizeof(id), path);
+            pstrcat(id, sizeof(id), "/");
+            qemu_free(path);
+        }
+    }
+    pstrcat(id, sizeof(id), idstr);
+
+    QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
+        if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
+            QTAILQ_REMOVE(&savevm_handlers, se, entry);
+            if (se->compat) {
+                qemu_free(se->compat);
+            }
+            qemu_free(se);
+        }
+    }
+}
+
+/* mark a device as not to be migrated, that is the device should be
+   unplugged before migration */
+void register_device_unmigratable(DeviceState *dev, const char *idstr,
+                                                            void *opaque)
+{
+    SaveStateEntry *se;
+    char id[256] = "";
+
+    if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
+        char *path = dev->parent_bus->info->get_dev_path(dev);
+        if (path) {
+            pstrcpy(id, sizeof(id), path);
+            pstrcat(id, sizeof(id), "/");
+            qemu_free(path);
+        }
+    }
+    pstrcat(id, sizeof(id), idstr);
+
+    QTAILQ_FOREACH(se, &savevm_handlers, entry) {
+        if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
+            se->no_migrate = 1;
+        }
+    }
+}
+
+int vmstate_register_with_alias_id(DeviceState *dev, int instance_id,
+                                   const VMStateDescription *vmsd,
+                                   void *opaque, int alias_id,
+                                   int required_for_version)
+{
+    SaveStateEntry *se;
+
+    /* If this triggers, alias support can be dropped for the vmsd. */
+    assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
+
+    se = qemu_mallocz(sizeof(SaveStateEntry));
+    se->version_id = vmsd->version_id;
+    se->section_id = global_section_id++;
+    se->save_live_state = NULL;
+    se->save_state = NULL;
+    se->load_state = NULL;
+    se->opaque = opaque;
+    se->vmsd = vmsd;
+    se->alias_id = alias_id;
+
+    if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
+        char *id = dev->parent_bus->info->get_dev_path(dev);
+        if (id) {
+            pstrcpy(se->idstr, sizeof(se->idstr), id);
+            pstrcat(se->idstr, sizeof(se->idstr), "/");
+            qemu_free(id);
+
+            se->compat = qemu_mallocz(sizeof(CompatEntry));
+            pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name);
+            se->compat->instance_id = instance_id == -1 ?
+                         calculate_compat_instance_id(vmsd->name) : instance_id;
+            instance_id = -1;
+        }
+    }
+    pstrcat(se->idstr, sizeof(se->idstr), vmsd->name);
+
+    if (instance_id == -1) {
+        se->instance_id = calculate_new_instance_id(se->idstr);
+    } else {
+        se->instance_id = instance_id;
+    }
+    assert(!se->compat || se->instance_id == 0);
+    /* add at the end of list */
+    QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
+    return 0;
+}
+
+int vmstate_register(DeviceState *dev, int instance_id,
+                     const VMStateDescription *vmsd, void *opaque)
+{
+    return vmstate_register_with_alias_id(dev, instance_id, vmsd,
+                                          opaque, -1, 0);
+}
+
+void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd,
+                        void *opaque)
+{
+    SaveStateEntry *se, *new_se;
+
+    QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
+        if (se->vmsd == vmsd && se->opaque == opaque) {
+            QTAILQ_REMOVE(&savevm_handlers, se, entry);
+            if (se->compat) {
+                qemu_free(se->compat);
+            }
+            qemu_free(se);
+        }
+    }
+}
+
+static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
+                                    void *opaque);
+static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
+                                   void *opaque);
+
+int vmstate_load_state(QEMUFile *f, const VMStateDescription *vmsd,
+                       void *opaque, int version_id)
+{
+    VMStateField *field = vmsd->fields;
+    int ret;
+
+    if (version_id > vmsd->version_id) {
+        return -EINVAL;
+    }
+    if (version_id < vmsd->minimum_version_id_old) {
+        return -EINVAL;
+    }
+    if  (version_id < vmsd->minimum_version_id) {
+        return vmsd->load_state_old(f, opaque, version_id);
+    }
+    if (vmsd->pre_load) {
+        int ret = vmsd->pre_load(opaque);
+        if (ret)
+            return ret;
+    }
+    while(field->name) {
+        if ((field->field_exists &&
+             field->field_exists(opaque, version_id)) ||
+            (!field->field_exists &&
+             field->version_id <= version_id)) {
+            void *base_addr = opaque + field->offset;
+            int i, n_elems = 1;
+            int size = field->size;
+
+            if (field->flags & VMS_VBUFFER) {
+                size = *(int32_t *)(opaque+field->size_offset);
+                if (field->flags & VMS_MULTIPLY) {
+                    size *= field->size;
+                }
+            }
+            if (field->flags & VMS_ARRAY) {
+                n_elems = field->num;
+            } else if (field->flags & VMS_VARRAY_INT32) {
+                n_elems = *(int32_t *)(opaque+field->num_offset);
+            } else if (field->flags & VMS_VARRAY_UINT32) {
+                n_elems = *(uint32_t *)(opaque+field->num_offset);
+            } else if (field->flags & VMS_VARRAY_UINT16) {
+                n_elems = *(uint16_t *)(opaque+field->num_offset);
+            } else if (field->flags & VMS_VARRAY_UINT8) {
+                n_elems = *(uint8_t *)(opaque+field->num_offset);
+            }
+            if (field->flags & VMS_POINTER) {
+                base_addr = *(void **)base_addr + field->start;
+            }
+            for (i = 0; i < n_elems; i++) {
+                void *addr = base_addr + size * i;
+
+                if (field->flags & VMS_ARRAY_OF_POINTER) {
+                    addr = *(void **)addr;
+                }
+                if (field->flags & VMS_STRUCT) {
+                    ret = vmstate_load_state(f, field->vmsd, addr, field->vmsd->version_id);
+                } else {
+                    ret = field->info->get(f, addr, size);
+
+                }
+                if (ret < 0) {
+                    return ret;
+                }
+            }
+        }
+        field++;
+    }
+    ret = vmstate_subsection_load(f, vmsd, opaque);
+    if (ret != 0) {
+        return ret;
+    }
+    if (vmsd->post_load) {
+        return vmsd->post_load(opaque, version_id);
+    }
+    return 0;
+}
+
+void vmstate_save_state(QEMUFile *f, const VMStateDescription *vmsd,
+                        void *opaque)
+{
+    VMStateField *field = vmsd->fields;
+
+    if (vmsd->pre_save) {
+        vmsd->pre_save(opaque);
+    }
+    while(field->name) {
+        if (!field->field_exists ||
+            field->field_exists(opaque, vmsd->version_id)) {
+            void *base_addr = opaque + field->offset;
+            int i, n_elems = 1;
+            int size = field->size;
+
+            if (field->flags & VMS_VBUFFER) {
+                size = *(int32_t *)(opaque+field->size_offset);
+                if (field->flags & VMS_MULTIPLY) {
+                    size *= field->size;
+                }
+            }
+            if (field->flags & VMS_ARRAY) {
+                n_elems = field->num;
+            } else if (field->flags & VMS_VARRAY_INT32) {
+                n_elems = *(int32_t *)(opaque+field->num_offset);
+            } else if (field->flags & VMS_VARRAY_UINT16) {
+                n_elems = *(uint16_t *)(opaque+field->num_offset);
+            } else if (field->flags & VMS_VARRAY_UINT8) {
+                n_elems = *(uint8_t *)(opaque+field->num_offset);
+            }
+            if (field->flags & VMS_POINTER) {
+                base_addr = *(void **)base_addr + field->start;
+            }
+            for (i = 0; i < n_elems; i++) {
+                void *addr = base_addr + size * i;
+
+                if (field->flags & VMS_ARRAY_OF_POINTER) {
+                    addr = *(void **)addr;
+                }
+                if (field->flags & VMS_STRUCT) {
+                    vmstate_save_state(f, field->vmsd, addr);
+                } else {
+                    field->info->put(f, addr, size);
+                }
+            }
+        }
+        field++;
+    }
+    vmstate_subsection_save(f, vmsd, opaque);
+}
+
+static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
+{
+    if (!se->vmsd) {         /* Old style */
+        return se->load_state(f, se->opaque, version_id);
+    }
+    return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
+}
+
+static void vmstate_save(QEMUFile *f, SaveStateEntry *se)
+{
+    if (!se->vmsd) {         /* Old style */
+        se->save_state(f, se->opaque);
+        return;
+    }
+    vmstate_save_state(f,se->vmsd, se->opaque);
+}
+
+#define QEMU_VM_FILE_MAGIC           0x5145564d
+#define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
+#define QEMU_VM_FILE_VERSION         0x00000003
+
+#define QEMU_VM_EOF                  0x00
+#define QEMU_VM_SECTION_START        0x01
+#define QEMU_VM_SECTION_PART         0x02
+#define QEMU_VM_SECTION_END          0x03
+#define QEMU_VM_SECTION_FULL         0x04
+#define QEMU_VM_SUBSECTION           0x05
+
+bool qemu_savevm_state_blocked(Monitor *mon)
+{
+    SaveStateEntry *se;
+
+    QTAILQ_FOREACH(se, &savevm_handlers, entry) {
+        if (se->no_migrate) {
+            monitor_printf(mon, "state blocked by non-migratable device '%s'\n",
+                           se->idstr);
+            return true;
+        }
+    }
+    return false;
+}
+
+int qemu_savevm_state_begin(Monitor *mon, QEMUFile *f, int blk_enable,
+                            int shared)
+{
+    SaveStateEntry *se;
+
+    QTAILQ_FOREACH(se, &savevm_handlers, entry) {
+        if(se->set_params == NULL) {
+            continue;
+       }
+       se->set_params(blk_enable, shared, se->opaque);
+    }
+    
+    qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
+    qemu_put_be32(f, QEMU_VM_FILE_VERSION);
+
+    QTAILQ_FOREACH(se, &savevm_handlers, entry) {
+        int len;
+
+        if (se->save_live_state == NULL)
+            continue;
+
+        /* Section type */
+        qemu_put_byte(f, QEMU_VM_SECTION_START);
+        qemu_put_be32(f, se->section_id);
+
+        /* ID string */
+        len = strlen(se->idstr);
+        qemu_put_byte(f, len);
+        qemu_put_buffer(f, (uint8_t *)se->idstr, len);
+
+        qemu_put_be32(f, se->instance_id);
+        qemu_put_be32(f, se->version_id);
+
+        se->save_live_state(mon, f, QEMU_VM_SECTION_START, se->opaque);
+    }
+
+    if (qemu_file_has_error(f)) {
+        qemu_savevm_state_cancel(mon, f);
+        return -EIO;
+    }
+
+    return 0;
+}
+
+int qemu_savevm_state_iterate(Monitor *mon, QEMUFile *f)
+{
+    SaveStateEntry *se;
+    int ret = 1;
+
+    QTAILQ_FOREACH(se, &savevm_handlers, entry) {
+        if (se->save_live_state == NULL)
+            continue;
+
+        /* Section type */
+        qemu_put_byte(f, QEMU_VM_SECTION_PART);
+        qemu_put_be32(f, se->section_id);
+
+        ret = se->save_live_state(mon, f, QEMU_VM_SECTION_PART, se->opaque);
+        if (!ret) {
+            /* Do not proceed to the next vmstate before this one reported
+               completion of the current stage. This serializes the migration
+               and reduces the probability that a faster changing state is
+               synchronized over and over again. */
+            break;
+        }
+    }
+
+    if (ret)
+        return 1;
+
+    if (qemu_file_has_error(f)) {
+        qemu_savevm_state_cancel(mon, f);
+        return -EIO;
+    }
+
+    return 0;
+}
+
+int qemu_savevm_state_complete(Monitor *mon, QEMUFile *f)
+{
+    SaveStateEntry *se;
+
+    cpu_synchronize_all_states();
+
+    QTAILQ_FOREACH(se, &savevm_handlers, entry) {
+        if (se->save_live_state == NULL)
+            continue;
+
+        /* Section type */
+        qemu_put_byte(f, QEMU_VM_SECTION_END);
+        qemu_put_be32(f, se->section_id);
+
+        se->save_live_state(mon, f, QEMU_VM_SECTION_END, se->opaque);
+    }
+
+    QTAILQ_FOREACH(se, &savevm_handlers, entry) {
+        int len;
+
+       if (se->save_state == NULL && se->vmsd == NULL)
+           continue;
+
+        /* Section type */
+        qemu_put_byte(f, QEMU_VM_SECTION_FULL);
+        qemu_put_be32(f, se->section_id);
+
+        /* ID string */
+        len = strlen(se->idstr);
+        qemu_put_byte(f, len);
+        qemu_put_buffer(f, (uint8_t *)se->idstr, len);
+
+        qemu_put_be32(f, se->instance_id);
+        qemu_put_be32(f, se->version_id);
+
+        vmstate_save(f, se);
+    }
+
+    qemu_put_byte(f, QEMU_VM_EOF);
+
+    if (qemu_file_has_error(f))
+        return -EIO;
+
+    return 0;
+}
+
+void qemu_savevm_state_cancel(Monitor *mon, QEMUFile *f)
+{
+    SaveStateEntry *se;
+
+    QTAILQ_FOREACH(se, &savevm_handlers, entry) {
+        if (se->save_live_state) {
+            se->save_live_state(mon, f, -1, se->opaque);
+        }
+    }
+}
+
+static int qemu_savevm_state(Monitor *mon, QEMUFile *f)
+{
+    int saved_vm_running;
+    int ret;
+
+    saved_vm_running = vm_running;
+    vm_stop(VMSTOP_SAVEVM);
+
+    if (qemu_savevm_state_blocked(mon)) {
+        ret = -EINVAL;
+        goto out;
+    }
+
+    ret = qemu_savevm_state_begin(mon, f, 0, 0);
+    if (ret < 0)
+        goto out;
+
+    do {
+        ret = qemu_savevm_state_iterate(mon, f);
+        if (ret < 0)
+            goto out;
+    } while (ret == 0);
+
+    ret = qemu_savevm_state_complete(mon, f);
+
+out:
+    if (qemu_file_has_error(f))
+        ret = -EIO;
+
+    if (!ret && saved_vm_running)
+        vm_start();
+
+    return ret;
+}
+
+static SaveStateEntry *find_se(const char *idstr, int instance_id)
+{
+    SaveStateEntry *se;
+
+    QTAILQ_FOREACH(se, &savevm_handlers, entry) {
+        if (!strcmp(se->idstr, idstr) &&
+            (instance_id == se->instance_id ||
+             instance_id == se->alias_id))
+            return se;
+        /* Migrating from an older version? */
+        if (strstr(se->idstr, idstr) && se->compat) {
+            if (!strcmp(se->compat->idstr, idstr) &&
+                (instance_id == se->compat->instance_id ||
+                 instance_id == se->alias_id))
+                return se;
+        }
+    }
+    return NULL;
+}
+
+static const VMStateDescription *vmstate_get_subsection(const VMStateSubsection *sub, char *idstr)
+{
+    while(sub && sub->needed) {
+        if (strcmp(idstr, sub->vmsd->name) == 0) {
+            return sub->vmsd;
+        }
+        sub++;
+    }
+    return NULL;
+}
+
+static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
+                                   void *opaque)
+{
+    const VMStateSubsection *sub = vmsd->subsections;
+
+    if (!sub || !sub->needed) {
+        return 0;
+    }
+
+    while (qemu_peek_byte(f) == QEMU_VM_SUBSECTION) {
+        char idstr[256];
+        int ret;
+        uint8_t version_id, len;
+        const VMStateDescription *sub_vmsd;
+
+        qemu_get_byte(f); /* subsection */
+        len = qemu_get_byte(f);
+        qemu_get_buffer(f, (uint8_t *)idstr, len);
+        idstr[len] = 0;
+        version_id = qemu_get_be32(f);
+
+        sub_vmsd = vmstate_get_subsection(sub, idstr);
+        if (sub_vmsd == NULL) {
+            return -ENOENT;
+        }
+        assert(!sub_vmsd->subsections);
+        ret = vmstate_load_state(f, sub_vmsd, opaque, version_id);
+        if (ret) {
+            return ret;
+        }
+    }
+    return 0;
+}
+
+static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
+                                    void *opaque)
+{
+    const VMStateSubsection *sub = vmsd->subsections;
+
+    while (sub && sub->needed) {
+        if (sub->needed(opaque)) {
+            const VMStateDescription *vmsd = sub->vmsd;
+            uint8_t len;
+
+            qemu_put_byte(f, QEMU_VM_SUBSECTION);
+            len = strlen(vmsd->name);
+            qemu_put_byte(f, len);
+            qemu_put_buffer(f, (uint8_t *)vmsd->name, len);
+            qemu_put_be32(f, vmsd->version_id);
+            assert(!vmsd->subsections);
+            vmstate_save_state(f, vmsd, opaque);
+        }
+        sub++;
+    }
+}
+
+typedef struct LoadStateEntry {
+    QLIST_ENTRY(LoadStateEntry) entry;
+    SaveStateEntry *se;
+    int section_id;
+    int version_id;
+} LoadStateEntry;
+
+int qemu_loadvm_state(QEMUFile *f)
+{
+    QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
+        QLIST_HEAD_INITIALIZER(loadvm_handlers);
+    LoadStateEntry *le, *new_le;
+    uint8_t section_type;
+    unsigned int v;
+    int ret;
+
+    if (qemu_savevm_state_blocked(default_mon)) {
+        return -EINVAL;
+    }
+
+    v = qemu_get_be32(f);
+    if (v != QEMU_VM_FILE_MAGIC)
+        return -EINVAL;
+
+    v = qemu_get_be32(f);
+    if (v == QEMU_VM_FILE_VERSION_COMPAT) {
+        fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
+        return -ENOTSUP;
+    }
+    if (v != QEMU_VM_FILE_VERSION)
+        return -ENOTSUP;
+
+    while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
+        uint32_t instance_id, version_id, section_id;
         SaveStateEntry *se;
         char idstr[257];
         int len;
@@ -981,23 +1789,36 @@ int qemu_loadvm_state(QEMUFile *f)
             le->se = se;
             le->section_id = section_id;
             le->version_id = version_id;
-            le->next = first_le;
-            first_le = le;
+            QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
 
-            le->se->load_state(f, le->se->opaque, le->version_id);
+            ret = vmstate_load(f, le->se, le->version_id);
+            if (ret < 0) {
+                fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
+                        instance_id, idstr);
+                goto out;
+            }
             break;
         case QEMU_VM_SECTION_PART:
         case QEMU_VM_SECTION_END:
             section_id = qemu_get_be32(f);
 
-            for (le = first_le; le && le->section_id != section_id; le = le->next);
+            QLIST_FOREACH(le, &loadvm_handlers, entry) {
+                if (le->section_id == section_id) {
+                    break;
+                }
+            }
             if (le == NULL) {
                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
                 ret = -EINVAL;
                 goto out;
             }
 
-            le->se->load_state(f, le->se->opaque, le->version_id);
+            ret = vmstate_load(f, le->se, le->version_id);
+            if (ret < 0) {
+                fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
+                        section_id);
+                goto out;
+            }
             break;
         default:
             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
@@ -1006,12 +1827,13 @@ int qemu_loadvm_state(QEMUFile *f)
         }
     }
 
+    cpu_synchronize_all_post_init();
+
     ret = 0;
 
 out:
-    while (first_le) {
-        LoadStateEntry *le = first_le;
-        first_le = first_le->next;
+    QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
+        QLIST_REMOVE(le, entry);
         qemu_free(le);
     }
 
@@ -1021,40 +1843,6 @@ out:
     return ret;
 }
 
-/* device can contain snapshots */
-static int bdrv_can_snapshot(BlockDriverState *bs)
-{
-    return (bs &&
-            !bdrv_is_removable(bs) &&
-            !bdrv_is_read_only(bs));
-}
-
-/* device must be snapshots in order to have a reliable snapshot */
-static int bdrv_has_snapshot(BlockDriverState *bs)
-{
-    return (bs &&
-            !bdrv_is_removable(bs) &&
-            !bdrv_is_read_only(bs));
-}
-
-static BlockDriverState *get_bs_snapshots(void)
-{
-    BlockDriverState *bs;
-    DriveInfo *dinfo;
-
-    if (bs_snapshots)
-        return bs_snapshots;
-    TAILQ_FOREACH(dinfo, &drives, next) {
-        bs = dinfo->bdrv;
-        if (bdrv_can_snapshot(bs))
-            goto ok;
-    }
-    return NULL;
- ok:
-    bs_snapshots = bs;
-    return bs;
-}
-
 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
                               const char *name)
 {
@@ -1077,48 +1865,75 @@ static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
     return ret;
 }
 
-void do_savevm(Monitor *mon, const char *name)
+/*
+ * Deletes snapshots of a given name in all opened images.
+ */
+static int del_existing_snapshots(Monitor *mon, const char *name)
+{
+    BlockDriverState *bs;
+    QEMUSnapshotInfo sn1, *snapshot = &sn1;
+    int ret;
+
+    bs = NULL;
+    while ((bs = bdrv_next(bs))) {
+        if (bdrv_can_snapshot(bs) &&
+            bdrv_snapshot_find(bs, snapshot, name) >= 0)
+        {
+            ret = bdrv_snapshot_delete(bs, name);
+            if (ret < 0) {
+                monitor_printf(mon,
+                               "Error while deleting snapshot on '%s'\n",
+                               bdrv_get_device_name(bs));
+                return -1;
+            }
+        }
+    }
+
+    return 0;
+}
+
+void do_savevm(Monitor *mon, const QDict *qdict)
 {
-    DriveInfo *dinfo;
     BlockDriverState *bs, *bs1;
     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
-    int must_delete, ret;
+    int ret;
     QEMUFile *f;
     int saved_vm_running;
     uint32_t vm_state_size;
 #ifdef _WIN32
     struct _timeb tb;
+    struct tm *ptm;
 #else
     struct timeval tv;
+    struct tm tm;
 #endif
+    const char *name = qdict_get_try_str(qdict, "name");
+
+    /* Verify if there is a device that doesn't support snapshots and is writable */
+    bs = NULL;
+    while ((bs = bdrv_next(bs))) {
+
+        if (bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
+            continue;
+        }
+
+        if (!bdrv_can_snapshot(bs)) {
+            monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
+                               bdrv_get_device_name(bs));
+            return;
+        }
+    }
 
-    bs = get_bs_snapshots();
+    bs = bdrv_snapshots();
     if (!bs) {
         monitor_printf(mon, "No block device can accept snapshots\n");
         return;
     }
 
-    /* ??? Should this occur after vm_stop?  */
-    qemu_aio_flush();
-
     saved_vm_running = vm_running;
-    vm_stop(0);
+    vm_stop(VMSTOP_SAVEVM);
 
-    must_delete = 0;
-    if (name) {
-        ret = bdrv_snapshot_find(bs, old_sn, name);
-        if (ret >= 0) {
-            must_delete = 1;
-        }
-    }
     memset(sn, 0, sizeof(*sn));
-    if (must_delete) {
-        pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
-        pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
-    } else {
-        if (name)
-            pstrcpy(sn->name, sizeof(sn->name), name);
-    }
 
     /* fill auxiliary fields */
 #ifdef _WIN32
@@ -1130,7 +1945,31 @@ void do_savevm(Monitor *mon, const char *name)
     sn->date_sec = tv.tv_sec;
     sn->date_nsec = tv.tv_usec * 1000;
 #endif
-    sn->vm_clock_nsec = qemu_get_clock(vm_clock);
+    sn->vm_clock_nsec = qemu_get_clock_ns(vm_clock);
+
+    if (name) {
+        ret = bdrv_snapshot_find(bs, old_sn, name);
+        if (ret >= 0) {
+            pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
+            pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
+        } else {
+            pstrcpy(sn->name, sizeof(sn->name), name);
+        }
+    } else {
+#ifdef _WIN32
+        ptm = localtime(&tb.time);
+        strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", ptm);
+#else
+        /* cast below needed for OpenBSD where tv_sec is still 'long' */
+        localtime_r((const time_t *)&tv.tv_sec, &tm);
+        strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
+#endif
+    }
+
+    /* Delete old snapshots of the same name */
+    if (name && del_existing_snapshots(mon, name) < 0) {
+        goto the_end;
+    }
 
     /* save the VM state */
     f = qemu_fopen_bdrv(bs, 1);
@@ -1138,7 +1977,7 @@ void do_savevm(Monitor *mon, const char *name)
         monitor_printf(mon, "Could not open VM state file\n");
         goto the_end;
     }
-    ret = qemu_savevm_state(f);
+    ret = qemu_savevm_state(mon, f);
     vm_state_size = qemu_ftell(f);
     qemu_fclose(f);
     if (ret < 0) {
@@ -1148,17 +1987,9 @@ void do_savevm(Monitor *mon, const char *name)
 
     /* create the snapshots */
 
-    TAILQ_FOREACH(dinfo, &drives, next) {
-        bs1 = dinfo->bdrv;
-        if (bdrv_has_snapshot(bs1)) {
-            if (must_delete) {
-                ret = bdrv_snapshot_delete(bs1, old_sn->id_str);
-                if (ret < 0) {
-                    monitor_printf(mon,
-                                   "Error while deleting snapshot on '%s'\n",
-                                   bdrv_get_device_name(bs1));
-                }
-            }
+    bs1 = NULL;
+    while ((bs1 = bdrv_next(bs1))) {
+        if (bdrv_can_snapshot(bs1)) {
             /* Write VM state size only to the image that contains the state */
             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
             ret = bdrv_snapshot_create(bs1, sn);
@@ -1174,97 +2005,101 @@ void do_savevm(Monitor *mon, const char *name)
         vm_start();
 }
 
-void load_vmstate(Monitor *mon, const char *name)
+int load_vmstate(const char *name)
 {
-    DriveInfo *dinfo;
-    BlockDriverState *bs, *bs1;
+    BlockDriverState *bs, *bs_vm_state;
     QEMUSnapshotInfo sn;
     QEMUFile *f;
     int ret;
 
-    bs = get_bs_snapshots();
-    if (!bs) {
-        monitor_printf(mon, "No block device supports snapshots\n");
-        return;
+    bs_vm_state = bdrv_snapshots();
+    if (!bs_vm_state) {
+        error_report("No block device supports snapshots");
+        return -ENOTSUP;
+    }
+
+    /* Don't even try to load empty VM states */
+    ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
+    if (ret < 0) {
+        return ret;
+    } else if (sn.vm_state_size == 0) {
+        error_report("This is a disk-only snapshot. Revert to it offline "
+            "using qemu-img.");
+        return -EINVAL;
+    }
+
+    /* Verify if there is any device that doesn't support snapshots and is
+    writable and check if the requested snapshot is available too. */
+    bs = NULL;
+    while ((bs = bdrv_next(bs))) {
+
+        if (bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
+            continue;
+        }
+
+        if (!bdrv_can_snapshot(bs)) {
+            error_report("Device '%s' is writable but does not support snapshots.",
+                               bdrv_get_device_name(bs));
+            return -ENOTSUP;
+        }
+
+        ret = bdrv_snapshot_find(bs, &sn, name);
+        if (ret < 0) {
+            error_report("Device '%s' does not have the requested snapshot '%s'",
+                           bdrv_get_device_name(bs), name);
+            return ret;
+        }
     }
 
     /* Flush all IO requests so they don't interfere with the new state.  */
     qemu_aio_flush();
 
-    TAILQ_FOREACH(dinfo, &drives, next) {
-        bs1 = dinfo->bdrv;
-        if (bdrv_has_snapshot(bs1)) {
-            ret = bdrv_snapshot_goto(bs1, name);
+    bs = NULL;
+    while ((bs = bdrv_next(bs))) {
+        if (bdrv_can_snapshot(bs)) {
+            ret = bdrv_snapshot_goto(bs, name);
             if (ret < 0) {
-                if (bs != bs1)
-                    monitor_printf(mon, "Warning: ");
-                switch(ret) {
-                case -ENOTSUP:
-                    monitor_printf(mon,
-                                   "Snapshots not supported on device '%s'\n",
-                                   bdrv_get_device_name(bs1));
-                    break;
-                case -ENOENT:
-                    monitor_printf(mon, "Could not find snapshot '%s' on "
-                                   "device '%s'\n",
-                                   name, bdrv_get_device_name(bs1));
-                    break;
-                default:
-                    monitor_printf(mon, "Error %d while activating snapshot on"
-                                   " '%s'\n", ret, bdrv_get_device_name(bs1));
-                    break;
-                }
-                /* fatal on snapshot block device */
-                if (bs == bs1)
-                    return;
+                error_report("Error %d while activating snapshot '%s' on '%s'",
+                             ret, name, bdrv_get_device_name(bs));
+                return ret;
             }
         }
     }
 
-    /* Don't even try to load empty VM states */
-    ret = bdrv_snapshot_find(bs, &sn, name);
-    if ((ret >= 0) && (sn.vm_state_size == 0))
-        return;
-
     /* restore the VM state */
-    f = qemu_fopen_bdrv(bs, 0);
+    f = qemu_fopen_bdrv(bs_vm_state, 0);
     if (!f) {
-        monitor_printf(mon, "Could not open VM state file\n");
-        return;
+        error_report("Could not open VM state file");
+        return -EINVAL;
     }
+
+    qemu_system_reset(VMRESET_SILENT);
     ret = qemu_loadvm_state(f);
+
     qemu_fclose(f);
     if (ret < 0) {
-        monitor_printf(mon, "Error %d while loading VM state\n", ret);
+        error_report("Error %d while loading VM state", ret);
+        return ret;
     }
-}
 
-void do_loadvm(Monitor *mon, const char *name)
-{
-    int saved_vm_running  = vm_running;
-
-    vm_stop(0);
-
-    load_vmstate(mon, name);
-    if (saved_vm_running)
-        vm_start();
+    return 0;
 }
 
-void do_delvm(Monitor *mon, const char *name)
+void do_delvm(Monitor *mon, const QDict *qdict)
 {
-    DriveInfo *dinfo;
     BlockDriverState *bs, *bs1;
     int ret;
+    const char *name = qdict_get_str(qdict, "name");
 
-    bs = get_bs_snapshots();
+    bs = bdrv_snapshots();
     if (!bs) {
         monitor_printf(mon, "No block device supports snapshots\n");
         return;
     }
 
-    TAILQ_FOREACH(dinfo, &drives, next) {
-        bs1 = dinfo->bdrv;
-        if (bdrv_has_snapshot(bs1)) {
+    bs1 = NULL;
+    while ((bs1 = bdrv_next(bs1))) {
+        if (bdrv_can_snapshot(bs1)) {
             ret = bdrv_snapshot_delete(bs1, name);
             if (ret < 0) {
                 if (ret == -ENOTSUP)
@@ -1281,38 +2116,64 @@ void do_delvm(Monitor *mon, const char *name)
 
 void do_info_snapshots(Monitor *mon)
 {
-    DriveInfo *dinfo;
     BlockDriverState *bs, *bs1;
-    QEMUSnapshotInfo *sn_tab, *sn;
-    int nb_sns, i;
+    QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
+    int nb_sns, i, ret, available;
+    int total;
+    int *available_snapshots;
     char buf[256];
 
-    bs = get_bs_snapshots();
+    bs = bdrv_snapshots();
     if (!bs) {
         monitor_printf(mon, "No available block device supports snapshots\n");
         return;
     }
-    monitor_printf(mon, "Snapshot devices:");
-    TAILQ_FOREACH(dinfo, &drives, next) {
-        bs1 = dinfo->bdrv;
-        if (bdrv_has_snapshot(bs1)) {
-            if (bs == bs1)
-                monitor_printf(mon, " %s", bdrv_get_device_name(bs1));
-        }
-    }
-    monitor_printf(mon, "\n");
 
     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
     if (nb_sns < 0) {
         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
         return;
     }
-    monitor_printf(mon, "Snapshot list (from %s):\n",
-                   bdrv_get_device_name(bs));
-    monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
-    for(i = 0; i < nb_sns; i++) {
+
+    if (nb_sns == 0) {
+        monitor_printf(mon, "There is no snapshot available.\n");
+        return;
+    }
+
+    available_snapshots = qemu_mallocz(sizeof(int) * nb_sns);
+    total = 0;
+    for (i = 0; i < nb_sns; i++) {
         sn = &sn_tab[i];
-        monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
+        available = 1;
+        bs1 = NULL;
+
+        while ((bs1 = bdrv_next(bs1))) {
+            if (bdrv_can_snapshot(bs1) && bs1 != bs) {
+                ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
+                if (ret < 0) {
+                    available = 0;
+                    break;
+                }
+            }
+        }
+
+        if (available) {
+            available_snapshots[total] = i;
+            total++;
+        }
+    }
+
+    if (total > 0) {
+        monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
+        for (i = 0; i < total; i++) {
+            sn = &sn_tab[available_snapshots[i]];
+            monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
+        }
+    } else {
+        monitor_printf(mon, "There is no suitable snapshot available\n");
     }
+
     qemu_free(sn_tab);
+    qemu_free(available_snapshots);
+
 }