Archipelago » qemu

/*

 * Copyright (C) 2009-2010 Nippon Telegraph and Telephone Corporation.

 *

 * This program is free software; you can redistribute it and/or

 * modify it under the terms of the GNU General Public License version

 * 2 as published by the Free Software Foundation.

 *

 * You should have received a copy of the GNU General Public License

 * along with this program. If not, see <http://www.gnu.org/licenses/>.

 */

#include "qemu-common.h"

#include "qemu-error.h"

#include "qemu_socket.h"

#include "block_int.h"

#include "bitops.h"

#define SD_PROTO_VER 0x01

#define SD_DEFAULT_ADDR "localhost"

#define SD_DEFAULT_PORT "7000"

#define SD_OP_CREATE_AND_WRITE_OBJ  0x01

#define SD_OP_READ_OBJ       0x02

#define SD_OP_WRITE_OBJ      0x03

#define SD_OP_NEW_VDI        0x11

#define SD_OP_LOCK_VDI       0x12

#define SD_OP_RELEASE_VDI    0x13

#define SD_OP_GET_VDI_INFO   0x14

#define SD_OP_READ_VDIS      0x15

#define SD_FLAG_CMD_WRITE    0x01

#define SD_FLAG_CMD_COW      0x02

#define SD_RES_SUCCESS       0x00 /* Success */

#define SD_RES_UNKNOWN       0x01 /* Unknown error */

#define SD_RES_NO_OBJ        0x02 /* No object found */

#define SD_RES_EIO           0x03 /* I/O error */

#define SD_RES_VDI_EXIST     0x04 /* Vdi exists already */

#define SD_RES_INVALID_PARMS 0x05 /* Invalid parameters */

#define SD_RES_SYSTEM_ERROR  0x06 /* System error */

#define SD_RES_VDI_LOCKED    0x07 /* Vdi is locked */

#define SD_RES_NO_VDI        0x08 /* No vdi found */

#define SD_RES_NO_BASE_VDI   0x09 /* No base vdi found */

#define SD_RES_VDI_READ      0x0A /* Cannot read requested vdi */

#define SD_RES_VDI_WRITE     0x0B /* Cannot write requested vdi */

#define SD_RES_BASE_VDI_READ 0x0C /* Cannot read base vdi */

#define SD_RES_BASE_VDI_WRITE   0x0D /* Cannot write base vdi */

#define SD_RES_NO_TAG        0x0E /* Requested tag is not found */

#define SD_RES_STARTUP       0x0F /* Sheepdog is on starting up */

#define SD_RES_VDI_NOT_LOCKED   0x10 /* Vdi is not locked */

#define SD_RES_SHUTDOWN      0x11 /* Sheepdog is shutting down */

#define SD_RES_NO_MEM        0x12 /* Cannot allocate memory */

#define SD_RES_FULL_VDI      0x13 /* we already have the maximum vdis */

#define SD_RES_VER_MISMATCH  0x14 /* Protocol version mismatch */

#define SD_RES_NO_SPACE      0x15 /* Server has no room for new objects */

#define SD_RES_WAIT_FOR_FORMAT  0x16 /* Waiting for a format operation */

#define SD_RES_WAIT_FOR_JOIN    0x17 /* Waiting for other nodes joining */

#define SD_RES_JOIN_FAILED   0x18 /* Target node had failed to join sheepdog */

/*

 * Object ID rules

 *

 *  0 - 19 (20 bits): data object space

 * 20 - 31 (12 bits): reserved data object space

 * 32 - 55 (24 bits): vdi object space

 * 56 - 59 ( 4 bits): reserved vdi object space

 * 60 - 63 ( 4 bits): object type indentifier space

 */

#define VDI_SPACE_SHIFT   32

#define VDI_BIT (UINT64_C(1) << 63)

#define VMSTATE_BIT (UINT64_C(1) << 62)

#define MAX_DATA_OBJS (UINT64_C(1) << 20)

#define MAX_CHILDREN 1024

#define SD_MAX_VDI_LEN 256

#define SD_MAX_VDI_TAG_LEN 256

#define SD_NR_VDIS   (1U << 24)

#define SD_DATA_OBJ_SIZE (UINT64_C(1) << 22)

#define SD_MAX_VDI_SIZE (SD_DATA_OBJ_SIZE * MAX_DATA_OBJS)

#define SECTOR_SIZE 512

#define SD_INODE_SIZE (sizeof(SheepdogInode))

#define CURRENT_VDI_ID 0

typedef struct SheepdogReq {

    uint8_t proto_ver;

    uint8_t opcode;

    uint16_t flags;

    uint32_t epoch;

    uint32_t id;

    uint32_t data_length;

    uint32_t opcode_specific[8];

} SheepdogReq;

typedef struct SheepdogRsp {

    uint8_t proto_ver;

    uint8_t opcode;

    uint16_t flags;

    uint32_t epoch;

    uint32_t id;

    uint32_t data_length;

    uint32_t result;

    uint32_t opcode_specific[7];

} SheepdogRsp;

typedef struct SheepdogObjReq {

    uint8_t proto_ver;

    uint8_t opcode;

    uint16_t flags;

    uint32_t epoch;

    uint32_t id;

    uint32_t data_length;

    uint64_t oid;

    uint64_t cow_oid;

    uint32_t copies;

    uint32_t rsvd;

    uint64_t offset;

} SheepdogObjReq;

typedef struct SheepdogObjRsp {

    uint8_t proto_ver;

    uint8_t opcode;

    uint16_t flags;

    uint32_t epoch;

    uint32_t id;

    uint32_t data_length;

    uint32_t result;

    uint32_t copies;

    uint32_t pad[6];

} SheepdogObjRsp;

typedef struct SheepdogVdiReq {

    uint8_t proto_ver;

    uint8_t opcode;

    uint16_t flags;

    uint32_t epoch;

    uint32_t id;

    uint32_t data_length;

    uint64_t vdi_size;

    uint32_t base_vdi_id;

    uint32_t copies;

    uint32_t snapid;

    uint32_t pad[3];

} SheepdogVdiReq;

typedef struct SheepdogVdiRsp {

    uint8_t proto_ver;

    uint8_t opcode;

    uint16_t flags;

    uint32_t epoch;

    uint32_t id;

    uint32_t data_length;

    uint32_t result;

    uint32_t rsvd;

    uint32_t vdi_id;

    uint32_t pad[5];

} SheepdogVdiRsp;

typedef struct SheepdogInode {

    char name[SD_MAX_VDI_LEN];

    char tag[SD_MAX_VDI_TAG_LEN];

    uint64_t ctime;

    uint64_t snap_ctime;

    uint64_t vm_clock_nsec;

    uint64_t vdi_size;

    uint64_t vm_state_size;

    uint16_t copy_policy;

    uint8_t nr_copies;

    uint8_t block_size_shift;

    uint32_t snap_id;

    uint32_t vdi_id;

    uint32_t parent_vdi_id;

    uint32_t child_vdi_id[MAX_CHILDREN];

    uint32_t data_vdi_id[MAX_DATA_OBJS];

} SheepdogInode;

/*

 * 64 bit FNV-1a non-zero initial basis

 */

#define FNV1A_64_INIT ((uint64_t)0xcbf29ce484222325ULL)

/*

 * 64 bit Fowler/Noll/Vo FNV-1a hash code

 */

static inline uint64_t fnv_64a_buf(void *buf, size_t len, uint64_t hval)

{

    unsigned char *bp = buf;

    unsigned char *be = bp + len;

    while (bp < be) {

        hval ^= (uint64_t) *bp++;

        hval += (hval << 1) + (hval << 4) + (hval << 5) +

            (hval << 7) + (hval << 8) + (hval << 40);

    }

    return hval;

}

static inline int is_data_obj_writable(SheepdogInode *inode, unsigned int idx)

{

    return inode->vdi_id == inode->data_vdi_id[idx];

}

static inline int is_data_obj(uint64_t oid)

{

    return !(VDI_BIT & oid);

}

static inline uint64_t data_oid_to_idx(uint64_t oid)

{

    return oid & (MAX_DATA_OBJS - 1);

}

static inline uint64_t vid_to_vdi_oid(uint32_t vid)

{

    return VDI_BIT | ((uint64_t)vid << VDI_SPACE_SHIFT);

}

static inline uint64_t vid_to_vmstate_oid(uint32_t vid, uint32_t idx)

{

    return VMSTATE_BIT | ((uint64_t)vid << VDI_SPACE_SHIFT) | idx;

}

static inline uint64_t vid_to_data_oid(uint32_t vid, uint32_t idx)

{

    return ((uint64_t)vid << VDI_SPACE_SHIFT) | idx;

}

static inline int is_snapshot(struct SheepdogInode *inode)

{

    return !!inode->snap_ctime;

}

#undef dprintf

#ifdef DEBUG_SDOG

#define dprintf(fmt, args...)                                       \

    do {                                                            \

        fprintf(stdout, "%s %d: " fmt, __func__, __LINE__, ##args); \

    } while (0)

#else

#define dprintf(fmt, args...)

#endif

typedef struct SheepdogAIOCB SheepdogAIOCB;

typedef struct AIOReq {

    SheepdogAIOCB *aiocb;

    unsigned int iov_offset;

    uint64_t oid;

    uint64_t base_oid;

    uint64_t offset;

    unsigned int data_len;

    uint8_t flags;

    uint32_t id;

    QLIST_ENTRY(AIOReq) outstanding_aio_siblings;

    QLIST_ENTRY(AIOReq) aioreq_siblings;

} AIOReq;

enum AIOCBState {

    AIOCB_WRITE_UDATA,

    AIOCB_READ_UDATA,

};

struct SheepdogAIOCB {

    BlockDriverAIOCB common;

    QEMUIOVector *qiov;

    int64_t sector_num;

    int nb_sectors;

    int ret;

    enum AIOCBState aiocb_type;

    QEMUBH *bh;

    void (*aio_done_func)(SheepdogAIOCB *);

    int canceled;

    QLIST_HEAD(aioreq_head, AIOReq) aioreq_head;

};

typedef struct BDRVSheepdogState {

    SheepdogInode inode;

    uint32_t min_dirty_data_idx;

    uint32_t max_dirty_data_idx;

    char name[SD_MAX_VDI_LEN];

    int is_snapshot;

    char *addr;

    char *port;

    int fd;

    uint32_t aioreq_seq_num;

    QLIST_HEAD(outstanding_aio_head, AIOReq) outstanding_aio_head;

} BDRVSheepdogState;

static const char * sd_strerror(int err)

{

    int i;

    static const struct {

        int err;

        const char *desc;

    } errors[] = {

        {SD_RES_SUCCESS, "Success"},

        {SD_RES_UNKNOWN, "Unknown error"},

        {SD_RES_NO_OBJ, "No object found"},

        {SD_RES_EIO, "I/O error"},

        {SD_RES_VDI_EXIST, "VDI exists already"},

        {SD_RES_INVALID_PARMS, "Invalid parameters"},

        {SD_RES_SYSTEM_ERROR, "System error"},

        {SD_RES_VDI_LOCKED, "VDI is already locked"},

        {SD_RES_NO_VDI, "No vdi found"},

        {SD_RES_NO_BASE_VDI, "No base VDI found"},

        {SD_RES_VDI_READ, "Failed read the requested VDI"},

        {SD_RES_VDI_WRITE, "Failed to write the requested VDI"},

        {SD_RES_BASE_VDI_READ, "Failed to read the base VDI"},

        {SD_RES_BASE_VDI_WRITE, "Failed to write the base VDI"},

        {SD_RES_NO_TAG, "Failed to find the requested tag"},

        {SD_RES_STARTUP, "The system is still booting"},

        {SD_RES_VDI_NOT_LOCKED, "VDI isn't locked"},

        {SD_RES_SHUTDOWN, "The system is shutting down"},

        {SD_RES_NO_MEM, "Out of memory on the server"},

        {SD_RES_FULL_VDI, "We already have the maximum vdis"},

        {SD_RES_VER_MISMATCH, "Protocol version mismatch"},

        {SD_RES_NO_SPACE, "Server has no space for new objects"},

        {SD_RES_WAIT_FOR_FORMAT, "Sheepdog is waiting for a format operation"},

        {SD_RES_WAIT_FOR_JOIN, "Sheepdog is waiting for other nodes joining"},

        {SD_RES_JOIN_FAILED, "Target node had failed to join sheepdog"},

    };

    for (i = 0; i < ARRAY_SIZE(errors); ++i) {

        if (errors[i].err == err) {

            return errors[i].desc;

        }

    }

    return "Invalid error code";

}

/*

 * Sheepdog I/O handling:

 *

 * 1. In the sd_aio_readv/writev, read/write requests are added to the

 *    QEMU Bottom Halves.

 *

 * 2. In sd_readv_writev_bh_cb, the callbacks of BHs, we send the I/O

 *    requests to the server and link the requests to the

 *    outstanding_list in the BDRVSheepdogState.  we exits the

 *    function without waiting for receiving the response.

 *

 * 3. We receive the response in aio_read_response, the fd handler to

 *    the sheepdog connection.  If metadata update is needed, we send

 *    the write request to the vdi object in sd_write_done, the write

 *    completion function.  The AIOCB callback is not called until all

 *    the requests belonging to the AIOCB are finished.

 */

static inline AIOReq *alloc_aio_req(BDRVSheepdogState *s, SheepdogAIOCB *acb,

                                    uint64_t oid, unsigned int data_len,

                                    uint64_t offset, uint8_t flags,

                                    uint64_t base_oid, unsigned int iov_offset)

{

    AIOReq *aio_req;

    aio_req = qemu_malloc(sizeof(*aio_req));

    aio_req->aiocb = acb;

    aio_req->iov_offset = iov_offset;

    aio_req->oid = oid;

    aio_req->base_oid = base_oid;

    aio_req->offset = offset;

    aio_req->data_len = data_len;

    aio_req->flags = flags;

    aio_req->id = s->aioreq_seq_num++;

    QLIST_INSERT_HEAD(&s->outstanding_aio_head, aio_req,

                      outstanding_aio_siblings);

    QLIST_INSERT_HEAD(&acb->aioreq_head, aio_req, aioreq_siblings);

    return aio_req;

}

static inline int free_aio_req(BDRVSheepdogState *s, AIOReq *aio_req)

{

    SheepdogAIOCB *acb = aio_req->aiocb;

    QLIST_REMOVE(aio_req, outstanding_aio_siblings);

    QLIST_REMOVE(aio_req, aioreq_siblings);

    qemu_free(aio_req);

    return !QLIST_EMPTY(&acb->aioreq_head);

}

static void sd_finish_aiocb(SheepdogAIOCB *acb)

{

    if (!acb->canceled) {

        acb->common.cb(acb->common.opaque, acb->ret);

    }

    qemu_aio_release(acb);

}

static void sd_aio_cancel(BlockDriverAIOCB *blockacb)

{

    SheepdogAIOCB *acb = (SheepdogAIOCB *)blockacb;

    /*

     * Sheepdog cannot cancel the requests which are already sent to

     * the servers, so we just complete the request with -EIO here.

     */

    acb->common.cb(acb->common.opaque, -EIO);

    acb->canceled = 1;

}

static AIOPool sd_aio_pool = {

    .aiocb_size = sizeof(SheepdogAIOCB),

    .cancel = sd_aio_cancel,

};

static SheepdogAIOCB *sd_aio_setup(BlockDriverState *bs, QEMUIOVector *qiov,

                                   int64_t sector_num, int nb_sectors,

                                   BlockDriverCompletionFunc *cb, void *opaque)

{

    SheepdogAIOCB *acb;

    acb = qemu_aio_get(&sd_aio_pool, bs, cb, opaque);

    acb->qiov = qiov;

    acb->sector_num = sector_num;

    acb->nb_sectors = nb_sectors;

    acb->aio_done_func = NULL;

    acb->canceled = 0;

    acb->bh = NULL;

    acb->ret = 0;

    QLIST_INIT(&acb->aioreq_head);

    return acb;

}

static int sd_schedule_bh(QEMUBHFunc *cb, SheepdogAIOCB *acb)

{

    if (acb->bh) {

        error_report("bug: %d %d", acb->aiocb_type, acb->aiocb_type);

        return -EIO;

    }

    acb->bh = qemu_bh_new(cb, acb);

    qemu_bh_schedule(acb->bh);

    return 0;

}

#ifdef _WIN32

struct msghdr {

    struct iovec *msg_iov;

    size_t        msg_iovlen;

};

static ssize_t sendmsg(int s, const struct msghdr *msg, int flags)

{

    size_t size = 0;

    char *buf, *p;

    int i, ret;

    /* count the msg size */

    for (i = 0; i < msg->msg_iovlen; i++) {

        size += msg->msg_iov[i].iov_len;

    }

    buf = qemu_malloc(size);

    p = buf;

    for (i = 0; i < msg->msg_iovlen; i++) {

        memcpy(p, msg->msg_iov[i].iov_base, msg->msg_iov[i].iov_len);

        p += msg->msg_iov[i].iov_len;

    }

    ret = send(s, buf, size, flags);

    qemu_free(buf);

    return ret;

}

static ssize_t recvmsg(int s, struct msghdr *msg, int flags)

{

    size_t size = 0;

    char *buf, *p;

    int i, ret;

    /* count the msg size */

    for (i = 0; i < msg->msg_iovlen; i++) {

        size += msg->msg_iov[i].iov_len;

    }

    buf = qemu_malloc(size);

    ret = qemu_recv(s, buf, size, flags);

    if (ret < 0) {

        goto out;

    }

    p = buf;

    for (i = 0; i < msg->msg_iovlen; i++) {

        memcpy(msg->msg_iov[i].iov_base, p, msg->msg_iov[i].iov_len);

        p += msg->msg_iov[i].iov_len;

    }

out:

    qemu_free(buf);

    return ret;

}

#endif

/*

 * Send/recv data with iovec buffers

 *

 * This function send/recv data from/to the iovec buffer directly.

 * The first `offset' bytes in the iovec buffer are skipped and next

 * `len' bytes are used.

 *

 * For example,

 *

 *   do_send_recv(sockfd, iov, len, offset, 1);

 *

 * is equals to

 *

 *   char *buf = malloc(size);

 *   iov_to_buf(iov, iovcnt, buf, offset, size);

 *   send(sockfd, buf, size, 0);

 *   free(buf);

 */

static int do_send_recv(int sockfd, struct iovec *iov, int len, int offset,

                        int write)

{

    struct msghdr msg;

    int ret, diff;

    memset(&msg, 0, sizeof(msg));

    msg.msg_iov = iov;

    msg.msg_iovlen = 1;

    len += offset;

    while (iov->iov_len < len) {

        len -= iov->iov_len;

        iov++;

        msg.msg_iovlen++;

    }

    diff = iov->iov_len - len;

    iov->iov_len -= diff;

    while (msg.msg_iov->iov_len <= offset) {

        offset -= msg.msg_iov->iov_len;

        msg.msg_iov++;

        msg.msg_iovlen--;

    }

    msg.msg_iov->iov_base = (char *) msg.msg_iov->iov_base + offset;

    msg.msg_iov->iov_len -= offset;

    if (write) {

        ret = sendmsg(sockfd, &msg, 0);

    } else {

        ret = recvmsg(sockfd, &msg, 0);

    }

    msg.msg_iov->iov_base = (char *) msg.msg_iov->iov_base - offset;

    msg.msg_iov->iov_len += offset;

    iov->iov_len += diff;

    return ret;

}

static int connect_to_sdog(const char *addr, const char *port)

{

    char hbuf[NI_MAXHOST], sbuf[NI_MAXSERV];

    int fd, ret;

    struct addrinfo hints, *res, *res0;

    if (!addr) {

        addr = SD_DEFAULT_ADDR;

        port = SD_DEFAULT_PORT;

    }

    memset(&hints, 0, sizeof(hints));

    hints.ai_socktype = SOCK_STREAM;

    ret = getaddrinfo(addr, port, &hints, &res0);

    if (ret) {

        error_report("unable to get address info %s, %s",

                     addr, strerror(errno));

        return -1;

    }

    for (res = res0; res; res = res->ai_next) {

        ret = getnameinfo(res->ai_addr, res->ai_addrlen, hbuf, sizeof(hbuf),

                          sbuf, sizeof(sbuf), NI_NUMERICHOST | NI_NUMERICSERV);

        if (ret) {

            continue;

        }

        fd = socket(res->ai_family, res->ai_socktype, res->ai_protocol);

        if (fd < 0) {

            continue;

        }

    reconnect:

        ret = connect(fd, res->ai_addr, res->ai_addrlen);

        if (ret < 0) {

            if (errno == EINTR) {

                goto reconnect;

            }

            break;

        }

        dprintf("connected to %s:%s\n", addr, port);

        goto success;

    }

    fd = -1;

    error_report("failed connect to %s:%s", addr, port);

success:

    freeaddrinfo(res0);

    return fd;

}

static int do_readv_writev(int sockfd, struct iovec *iov, int len,

                           int iov_offset, int write)

{

    int ret;

again:

    ret = do_send_recv(sockfd, iov, len, iov_offset, write);

    if (ret < 0) {

        if (errno == EINTR || errno == EAGAIN) {

            goto again;

        }

        error_report("failed to recv a rsp, %s", strerror(errno));

        return 1;

    }

    iov_offset += ret;

    len -= ret;

    if (len) {

        goto again;

    }

    return 0;

}

static int do_readv(int sockfd, struct iovec *iov, int len, int iov_offset)

{

    return do_readv_writev(sockfd, iov, len, iov_offset, 0);

}

static int do_writev(int sockfd, struct iovec *iov, int len, int iov_offset)

{

    return do_readv_writev(sockfd, iov, len, iov_offset, 1);

}

static int do_read_write(int sockfd, void *buf, int len, int write)

{

    struct iovec iov;

    iov.iov_base = buf;

    iov.iov_len = len;

    return do_readv_writev(sockfd, &iov, len, 0, write);

}

static int do_read(int sockfd, void *buf, int len)

{

    return do_read_write(sockfd, buf, len, 0);

}

static int do_write(int sockfd, void *buf, int len)

{

    return do_read_write(sockfd, buf, len, 1);

}

static int send_req(int sockfd, SheepdogReq *hdr, void *data,

                    unsigned int *wlen)

{

    int ret;

    struct iovec iov[2];

    iov[0].iov_base = hdr;

    iov[0].iov_len = sizeof(*hdr);

    if (*wlen) {

        iov[1].iov_base = data;

        iov[1].iov_len = *wlen;

    }

    ret = do_writev(sockfd, iov, sizeof(*hdr) + *wlen, 0);

    if (ret) {

        error_report("failed to send a req, %s", strerror(errno));

        ret = -1;

    }

    return ret;

}

static int do_req(int sockfd, SheepdogReq *hdr, void *data,

                  unsigned int *wlen, unsigned int *rlen)

{

    int ret;

    ret = send_req(sockfd, hdr, data, wlen);

    if (ret) {

        ret = -1;

        goto out;

    }

    ret = do_read(sockfd, hdr, sizeof(*hdr));

    if (ret) {

        error_report("failed to get a rsp, %s", strerror(errno));

        ret = -1;

        goto out;

    }

    if (*rlen > hdr->data_length) {

        *rlen = hdr->data_length;

    }

    if (*rlen) {

        ret = do_read(sockfd, data, *rlen);

        if (ret) {

            error_report("failed to get the data, %s", strerror(errno));

            ret = -1;

            goto out;

        }

    }

    ret = 0;

out:

    return ret;

}

static int add_aio_request(BDRVSheepdogState *s, AIOReq *aio_req,

                           struct iovec *iov, int niov, int create,

                           enum AIOCBState aiocb_type);

/*

 * This function searchs pending requests to the object `oid', and

 * sends them.

 */

static void send_pending_req(BDRVSheepdogState *s, uint64_t oid, uint32_t id)

{

    AIOReq *aio_req, *next;

    SheepdogAIOCB *acb;

    int ret;

    QLIST_FOREACH_SAFE(aio_req, &s->outstanding_aio_head,

                       outstanding_aio_siblings, next) {

        if (id == aio_req->id) {

            continue;

        }

        if (aio_req->oid != oid) {

            continue;

        }

        acb = aio_req->aiocb;

        ret = add_aio_request(s, aio_req, acb->qiov->iov,

                              acb->qiov->niov, 0, acb->aiocb_type);

        if (ret < 0) {

            error_report("add_aio_request is failed");

            free_aio_req(s, aio_req);

            if (QLIST_EMPTY(&acb->aioreq_head)) {

                sd_finish_aiocb(acb);

            }

        }

    }

}

/*

 * Receive responses of the I/O requests.

 *

 * This function is registered as a fd handler, and called from the

 * main loop when s->fd is ready for reading responses.

 */

static void aio_read_response(void *opaque)

{

    SheepdogObjRsp rsp;

    BDRVSheepdogState *s = opaque;

    int fd = s->fd;

    int ret;

    AIOReq *aio_req = NULL;

    SheepdogAIOCB *acb;

    int rest;

    unsigned long idx;

    if (QLIST_EMPTY(&s->outstanding_aio_head)) {

        return;

    }

    /* read a header */

    ret = do_read(fd, &rsp, sizeof(rsp));

    if (ret) {

        error_report("failed to get the header, %s", strerror(errno));

        return;

    }

    /* find the right aio_req from the outstanding_aio list */

    QLIST_FOREACH(aio_req, &s->outstanding_aio_head, outstanding_aio_siblings) {

        if (aio_req->id == rsp.id) {

            break;

        }

    }

    if (!aio_req) {

        error_report("cannot find aio_req %x", rsp.id);

        return;

    }

    acb = aio_req->aiocb;

    switch (acb->aiocb_type) {

    case AIOCB_WRITE_UDATA:

        if (!is_data_obj(aio_req->oid)) {

            break;

        }

        idx = data_oid_to_idx(aio_req->oid);

        if (s->inode.data_vdi_id[idx] != s->inode.vdi_id) {

            /*

             * If the object is newly created one, we need to update

             * the vdi object (metadata object).  min_dirty_data_idx

             * and max_dirty_data_idx are changed to include updated

             * index between them.

             */

            s->inode.data_vdi_id[idx] = s->inode.vdi_id;

            s->max_dirty_data_idx = MAX(idx, s->max_dirty_data_idx);

            s->min_dirty_data_idx = MIN(idx, s->min_dirty_data_idx);

            /*

             * Some requests may be blocked because simultaneous

             * create requests are not allowed, so we search the

             * pending requests here.

             */

            send_pending_req(s, vid_to_data_oid(s->inode.vdi_id, idx), rsp.id);

        }

        break;

    case AIOCB_READ_UDATA:

        ret = do_readv(fd, acb->qiov->iov, rsp.data_length,

                       aio_req->iov_offset);

        if (ret) {

            error_report("failed to get the data, %s", strerror(errno));

            return;

        }

        break;

    }

    if (rsp.result != SD_RES_SUCCESS) {

        acb->ret = -EIO;

        error_report("%s", sd_strerror(rsp.result));

    }

    rest = free_aio_req(s, aio_req);

    if (!rest) {

        /*

         * We've finished all requests which belong to the AIOCB, so

         * we can call the callback now.

         */

        acb->aio_done_func(acb);

    }

}

static int aio_flush_request(void *opaque)

{

    BDRVSheepdogState *s = opaque;

    return !QLIST_EMPTY(&s->outstanding_aio_head);

}

#if !defined(SOL_TCP) || !defined(TCP_CORK)

static int set_cork(int fd, int v)

{

    return 0;

}

#else

static int set_cork(int fd, int v)

{

    return setsockopt(fd, SOL_TCP, TCP_CORK, &v, sizeof(v));

}

#endif

static int set_nodelay(int fd)

{

    int ret, opt;

    opt = 1;

    ret = setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, (char *)&opt, sizeof(opt));

    return ret;

}

/*

 * Return a socket discriptor to read/write objects.

 *

 * We cannot use this discriptor for other operations because

 * the block driver may be on waiting response from the server.

 */

static int get_sheep_fd(BDRVSheepdogState *s)

{

    int ret, fd;

    fd = connect_to_sdog(s->addr, s->port);

    if (fd < 0) {

        error_report("%s", strerror(errno));

        return -1;

    }

    socket_set_nonblock(fd);

    ret = set_nodelay(fd);

    if (ret) {

        error_report("%s", strerror(errno));

        closesocket(fd);

        return -1;

    }

    qemu_aio_set_fd_handler(fd, aio_read_response, NULL, aio_flush_request,

                            NULL, s);

    return fd;

}

/*

 * Parse a filename

 *

 * filename must be one of the following formats:

 *   1. [vdiname]

 *   2. [vdiname]:[snapid]

 *   3. [vdiname]:[tag]

 *   4. [hostname]:[port]:[vdiname]

 *   5. [hostname]:[port]:[vdiname]:[snapid]

 *   6. [hostname]:[port]:[vdiname]:[tag]

 *

 * You can boot from the snapshot images by specifying `snapid` or

 * `tag'.

 *

 * You can run VMs outside the Sheepdog cluster by specifying

 * `hostname' and `port' (experimental).

 */

static int parse_vdiname(BDRVSheepdogState *s, const char *filename,

                         char *vdi, uint32_t *snapid, char *tag)

{

    char *p, *q;

    int nr_sep;

    p = q = qemu_strdup(filename);

    /* count the number of separators */

    nr_sep = 0;

    while (*p) {

        if (*p == ':') {

            nr_sep++;

        }

        p++;

    }

    p = q;

    /* use the first two tokens as hostname and port number. */

    if (nr_sep >= 2) {

        s->addr = p;

        p = strchr(p, ':');

        *p++ = '\0';

        s->port = p;

        p = strchr(p, ':');

        *p++ = '\0';

    } else {

        s->addr = NULL;

        s->port = 0;

    }

    strncpy(vdi, p, SD_MAX_VDI_LEN);

    p = strchr(vdi, ':');

    if (p) {

        *p++ = '\0';

        *snapid = strtoul(p, NULL, 10);

        if (*snapid == 0) {

            strncpy(tag, p, SD_MAX_VDI_TAG_LEN);

        }

    } else {

        *snapid = CURRENT_VDI_ID; /* search current vdi */

    }

    if (s->addr == NULL) {

        qemu_free(q);

    }

    return 0;

}

static int find_vdi_name(BDRVSheepdogState *s, char *filename, uint32_t snapid,

                         char *tag, uint32_t *vid, int for_snapshot)

{

    int ret, fd;

    SheepdogVdiReq hdr;

    SheepdogVdiRsp *rsp = (SheepdogVdiRsp *)&hdr;

    unsigned int wlen, rlen = 0;

    char buf[SD_MAX_VDI_LEN + SD_MAX_VDI_TAG_LEN];

    fd = connect_to_sdog(s->addr, s->port);

    if (fd < 0) {

        return -1;

    }

    memset(buf, 0, sizeof(buf));

    strncpy(buf, filename, SD_MAX_VDI_LEN);

    strncpy(buf + SD_MAX_VDI_LEN, tag, SD_MAX_VDI_TAG_LEN);

    memset(&hdr, 0, sizeof(hdr));

    if (for_snapshot) {

        hdr.opcode = SD_OP_GET_VDI_INFO;

    } else {

        hdr.opcode = SD_OP_LOCK_VDI;

    }

    wlen = SD_MAX_VDI_LEN + SD_MAX_VDI_TAG_LEN;

    hdr.proto_ver = SD_PROTO_VER;

    hdr.data_length = wlen;

    hdr.snapid = snapid;

    hdr.flags = SD_FLAG_CMD_WRITE;

    ret = do_req(fd, (SheepdogReq *)&hdr, buf, &wlen, &rlen);

    if (ret) {

        ret = -1;

        goto out;

    }

    if (rsp->result != SD_RES_SUCCESS) {

        error_report("cannot get vdi info, %s, %s %d %s",

                     sd_strerror(rsp->result), filename, snapid, tag);

        ret = -1;

        goto out;

    }

    *vid = rsp->vdi_id;

    ret = 0;

out:

    closesocket(fd);

    return ret;

}

static int add_aio_request(BDRVSheepdogState *s, AIOReq *aio_req,

                           struct iovec *iov, int niov, int create,

                           enum AIOCBState aiocb_type)

{

    int nr_copies = s->inode.nr_copies;

    SheepdogObjReq hdr;

    unsigned int wlen;

    int ret;

    uint64_t oid = aio_req->oid;

    unsigned int datalen = aio_req->data_len;

    uint64_t offset = aio_req->offset;

    uint8_t flags = aio_req->flags;

    uint64_t old_oid = aio_req->base_oid;

    if (!nr_copies) {

        error_report("bug");

    }

    memset(&hdr, 0, sizeof(hdr));

    if (aiocb_type == AIOCB_READ_UDATA) {

        wlen = 0;

        hdr.opcode = SD_OP_READ_OBJ;

        hdr.flags = flags;

    } else if (create) {

        wlen = datalen;

        hdr.opcode = SD_OP_CREATE_AND_WRITE_OBJ;

        hdr.flags = SD_FLAG_CMD_WRITE | flags;

    } else {

        wlen = datalen;

        hdr.opcode = SD_OP_WRITE_OBJ;

        hdr.flags = SD_FLAG_CMD_WRITE | flags;

    }

    hdr.oid = oid;

    hdr.cow_oid = old_oid;

    hdr.copies = s->inode.nr_copies;

    hdr.data_length = datalen;

    hdr.offset = offset;

    hdr.id = aio_req->id;

    set_cork(s->fd, 1);

    /* send a header */

    ret = do_write(s->fd, &hdr, sizeof(hdr));

    if (ret) {

        error_report("failed to send a req, %s", strerror(errno));

        return -EIO;

    }

    if (wlen) {

        ret = do_writev(s->fd, iov, wlen, aio_req->iov_offset);

        if (ret) {

            error_report("failed to send a data, %s", strerror(errno));

            return -EIO;

        }

    }

    set_cork(s->fd, 0);

    return 0;

}

static int read_write_object(int fd, char *buf, uint64_t oid, int copies,

                             unsigned int datalen, uint64_t offset,

                             int write, int create)

{

    SheepdogObjReq hdr;

    SheepdogObjRsp *rsp = (SheepdogObjRsp *)&hdr;

    unsigned int wlen, rlen;

    int ret;

    memset(&hdr, 0, sizeof(hdr));

    if (write) {

        wlen = datalen;

        rlen = 0;

        hdr.flags = SD_FLAG_CMD_WRITE;

        if (create) {

            hdr.opcode = SD_OP_CREATE_AND_WRITE_OBJ;

        } else {

            hdr.opcode = SD_OP_WRITE_OBJ;

        }

    } else {

        wlen = 0;

        rlen = datalen;

        hdr.opcode = SD_OP_READ_OBJ;

    }

    hdr.oid = oid;

    hdr.data_length = datalen;

    hdr.offset = offset;

    hdr.copies = copies;

    ret = do_req(fd, (SheepdogReq *)&hdr, buf, &wlen, &rlen);

    if (ret) {

        error_report("failed to send a request to the sheep");

        return -1;

    }

    switch (rsp->result) {

    case SD_RES_SUCCESS:

        return 0;

    default:

        error_report("%s", sd_strerror(rsp->result));

        return -1;

    }

}

static int read_object(int fd, char *buf, uint64_t oid, int copies,

                       unsigned int datalen, uint64_t offset)

{

    return read_write_object(fd, buf, oid, copies, datalen, offset, 0, 0);

}

static int write_object(int fd, char *buf, uint64_t oid, int copies,

                        unsigned int datalen, uint64_t offset, int create)

{

    return read_write_object(fd, buf, oid, copies, datalen, offset, 1, create);

}

static int sd_open(BlockDriverState *bs, const char *filename, int flags)

{

    int ret, fd;

    uint32_t vid = 0;

    BDRVSheepdogState *s = bs->opaque;

    char vdi[SD_MAX_VDI_LEN], tag[SD_MAX_VDI_TAG_LEN];

    uint32_t snapid;

    char *buf = NULL;

    strstart(filename, "sheepdog:", (const char **)&filename);

    QLIST_INIT(&s->outstanding_aio_head);

    s->fd = -1;

    memset(vdi, 0, sizeof(vdi));

    memset(tag, 0, sizeof(tag));

    if (parse_vdiname(s, filename, vdi, &snapid, tag) < 0) {

        goto out;

    }

    s->fd = get_sheep_fd(s);

    if (s->fd < 0) {

        goto out;

    }

    ret = find_vdi_name(s, vdi, snapid, tag, &vid, 0);

    if (ret) {

        goto out;

    }

    if (snapid) {

        dprintf("%" PRIx32 " snapshot inode was open.\n", vid);

        s->is_snapshot = 1;

    }

    fd = connect_to_sdog(s->addr, s->port);

    if (fd < 0) {

        error_report("failed to connect");

        goto out;

    }

    buf = qemu_malloc(SD_INODE_SIZE);

    ret = read_object(fd, buf, vid_to_vdi_oid(vid), 0, SD_INODE_SIZE, 0);

    closesocket(fd);

    if (ret) {

        goto out;

    }

    memcpy(&s->inode, buf, sizeof(s->inode));

    s->min_dirty_data_idx = UINT32_MAX;

    s->max_dirty_data_idx = 0;

    bs->total_sectors = s->inode.vdi_size / SECTOR_SIZE;

    strncpy(s->name, vdi, sizeof(s->name));

    qemu_free(buf);

    return 0;

out:

    qemu_aio_set_fd_handler(s->fd, NULL, NULL, NULL, NULL, NULL);

    if (s->fd >= 0) {

        closesocket(s->fd);

    }

    qemu_free(buf);

    return -1;

}

static int do_sd_create(char *filename, int64_t vdi_size,

                        uint32_t base_vid, uint32_t *vdi_id, int snapshot,

                        const char *addr, const char *port)

{

    SheepdogVdiReq hdr;

    SheepdogVdiRsp *rsp = (SheepdogVdiRsp *)&hdr;

    int fd, ret;

    unsigned int wlen, rlen = 0;

    char buf[SD_MAX_VDI_LEN];

    fd = connect_to_sdog(addr, port);

    if (fd < 0) {

        return -EIO;

    }

    memset(buf, 0, sizeof(buf));

    strncpy(buf, filename, SD_MAX_VDI_LEN);

    memset(&hdr, 0, sizeof(hdr));

    hdr.opcode = SD_OP_NEW_VDI;

    hdr.base_vdi_id = base_vid;

    wlen = SD_MAX_VDI_LEN;

    hdr.flags = SD_FLAG_CMD_WRITE;

    hdr.snapid = snapshot;

    hdr.data_length = wlen;

    hdr.vdi_size = vdi_size;

    ret = do_req(fd, (SheepdogReq *)&hdr, buf, &wlen, &rlen);

    closesocket(fd);

    if (ret) {

        return -EIO;

    }

    if (rsp->result != SD_RES_SUCCESS) {

        error_report("%s, %s", sd_strerror(rsp->result), filename);

        return -EIO;

    }

    if (vdi_id) {

        *vdi_id = rsp->vdi_id;

    }

    return 0;

}

static int sd_prealloc(const char *filename)

{

    BlockDriverState *bs = NULL;

    uint32_t idx, max_idx;

    int64_t vdi_size;

    void *buf = qemu_mallocz(SD_DATA_OBJ_SIZE);

    int ret;

    ret = bdrv_file_open(&bs, filename, BDRV_O_RDWR);

    if (ret < 0) {

        goto out;

    }

    vdi_size = bdrv_getlength(bs);

    if (vdi_size < 0) {

        ret = vdi_size;

        goto out;

    }

    max_idx = DIV_ROUND_UP(vdi_size, SD_DATA_OBJ_SIZE);

    for (idx = 0; idx < max_idx; idx++) {

        /*

         * The created image can be a cloned image, so we need to read

         * a data from the source image.

         */

        ret = bdrv_pread(bs, idx * SD_DATA_OBJ_SIZE, buf, SD_DATA_OBJ_SIZE);

        if (ret < 0) {

            goto out;

        }

        ret = bdrv_pwrite(bs, idx * SD_DATA_OBJ_SIZE, buf, SD_DATA_OBJ_SIZE);

        if (ret < 0) {

            goto out;

        }

    }

out:

    if (bs) {

        bdrv_delete(bs);

    }

    qemu_free(buf);

    return ret;

}

static int sd_create(const char *filename, QEMUOptionParameter *options)

{

    int ret;

    uint32_t vid = 0, base_vid = 0;

    int64_t vdi_size = 0;

    char *backing_file = NULL;

    BDRVSheepdogState s;

    char vdi[SD_MAX_VDI_LEN], tag[SD_MAX_VDI_TAG_LEN];

    uint32_t snapid;

    int prealloc = 0;

    const char *vdiname;

    strstart(filename, "sheepdog:", &vdiname);

    memset(&s, 0, sizeof(s));

    memset(vdi, 0, sizeof(vdi));

    memset(tag, 0, sizeof(tag));

    if (parse_vdiname(&s, vdiname, vdi, &snapid, tag) < 0) {

        error_report("invalid filename");