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/>.

 *

 * Contributions after 2012-01-13 are licensed under the terms of the

 * GNU GPL, version 2 or (at your option) any later version.

 */

#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_OP_FLUSH_VDI      0x16

#define SD_FLAG_CMD_WRITE    0x01

#define SD_FLAG_CMD_COW      0x02

#define SD_FLAG_CMD_CACHE    0x04

#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 identifier 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) aio_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;

    Coroutine *coroutine;

    void (*aio_done_func)(SheepdogAIOCB *);

    int canceled;

    int nr_pending;

};

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;

    uint8_t cache_enabled;

    char *addr;

    char *port;

    int fd;

    int flush_fd;

    CoMutex lock;

    Coroutine *co_send;

    Coroutine *co_recv;

    uint32_t aioreq_seq_num;

    QLIST_HEAD(inflight_aio_head, AIOReq) inflight_aio_head;

    QLIST_HEAD(pending_aio_head, AIOReq) pending_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 sd_co_rw_vector, we send the I/O requests to the server and

 *    link the requests to the inflight_list in the

 *    BDRVSheepdogState.  The function exits without waiting for

 *    receiving the response.

 *

 * 2. 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.  We switch back to sd_co_readv/writev after

 *    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 = g_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++;

    acb->nr_pending++;

    return aio_req;

}

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

{

    SheepdogAIOCB *acb = aio_req->aiocb;

    QLIST_REMOVE(aio_req, aio_siblings);

    g_free(aio_req);

    acb->nr_pending--;

}

static void coroutine_fn sd_finish_aiocb(SheepdogAIOCB *acb)

{

    if (!acb->canceled) {

        qemu_coroutine_enter(acb->coroutine, NULL);

    }

    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->ret = -EIO;

    qemu_coroutine_enter(acb->coroutine, NULL);

    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->coroutine = qemu_coroutine_self();

    acb->ret = 0;

    acb->nr_pending = 0;

    return acb;

}

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 -errno;

    }

    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 = -errno;

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

success:

    freeaddrinfo(res0);

    return fd;

}

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

                    unsigned int *wlen)

{

    int ret;

    ret = qemu_send_full(sockfd, hdr, sizeof(*hdr), 0);

    if (ret < sizeof(*hdr)) {

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

        return -errno;

    }

    ret = qemu_send_full(sockfd, data, *wlen, 0);

    if (ret < *wlen) {

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

        ret = -errno;

    }

    return ret;

}

static coroutine_fn int send_co_req(int sockfd, SheepdogReq *hdr, void *data,

                                    unsigned int *wlen)

{

    int ret;

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

    if (ret < sizeof(*hdr)) {

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

        return ret;

    }

    ret = qemu_co_send(sockfd, data, *wlen);

    if (ret < *wlen) {

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

    }

    return ret;

}

static coroutine_fn int do_co_req(int sockfd, SheepdogReq *hdr, void *data,

                                  unsigned int *wlen, unsigned int *rlen);

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

                  unsigned int *wlen, unsigned int *rlen)

{

    int ret;

    if (qemu_in_coroutine()) {

        return do_co_req(sockfd, hdr, data, wlen, rlen);

    }

    socket_set_block(sockfd);

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

    if (ret < 0) {

        goto out;

    }

    ret = qemu_recv_full(sockfd, hdr, sizeof(*hdr), 0);

    if (ret < sizeof(*hdr)) {

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

        ret = -errno;

        goto out;

    }

    if (*rlen > hdr->data_length) {

        *rlen = hdr->data_length;

    }

    if (*rlen) {

        ret = qemu_recv_full(sockfd, data, *rlen, 0);

        if (ret < *rlen) {

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

            ret = -errno;

            goto out;

        }

    }

    ret = 0;

out:

    socket_set_nonblock(sockfd);

    return ret;

}

static void restart_co_req(void *opaque)

{

    Coroutine *co = opaque;

    qemu_coroutine_enter(co, NULL);

}

static coroutine_fn int do_co_req(int sockfd, SheepdogReq *hdr, void *data,

                                  unsigned int *wlen, unsigned int *rlen)

{

    int ret;

    Coroutine *co;

    co = qemu_coroutine_self();

    qemu_aio_set_fd_handler(sockfd, NULL, restart_co_req, NULL, co);

    socket_set_block(sockfd);

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

    if (ret < 0) {

        goto out;

    }

    qemu_aio_set_fd_handler(sockfd, restart_co_req, NULL, NULL, co);

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

    if (ret < sizeof(*hdr)) {

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

        ret = -errno;

        goto out;

    }

    if (*rlen > hdr->data_length) {

        *rlen = hdr->data_length;

    }

    if (*rlen) {

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

        if (ret < *rlen) {

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

            ret = -errno;

            goto out;

        }

    }

    ret = 0;

out:

    qemu_aio_set_fd_handler(sockfd, NULL, NULL, NULL, NULL);

    socket_set_nonblock(sockfd);

    return ret;

}

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

                           struct iovec *iov, int niov, int create,

                           enum AIOCBState aiocb_type);

static AIOReq *find_pending_req(BDRVSheepdogState *s, uint64_t oid)

{

    AIOReq *aio_req;

    QLIST_FOREACH(aio_req, &s->pending_aio_head, aio_siblings) {

        if (aio_req->oid == oid) {

            return aio_req;

        }

    }

    return NULL;

}

/*

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

 * sends them.

 */

static void coroutine_fn send_pending_req(BDRVSheepdogState *s, uint64_t oid)

{

    AIOReq *aio_req;

    SheepdogAIOCB *acb;

    int ret;

    while ((aio_req = find_pending_req(s, oid)) != NULL) {

        acb = aio_req->aiocb;

        /* move aio_req from pending list to inflight one */

        QLIST_REMOVE(aio_req, aio_siblings);

        QLIST_INSERT_HEAD(&s->inflight_aio_head, aio_req, aio_siblings);

        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 (!acb->nr_pending) {

                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 coroutine_fn aio_read_response(void *opaque)

{

    SheepdogObjRsp rsp;

    BDRVSheepdogState *s = opaque;

    int fd = s->fd;

    int ret;

    AIOReq *aio_req = NULL;

    SheepdogAIOCB *acb;

    unsigned long idx;

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

        goto out;

    }

    /* read a header */

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

    if (ret < 0) {

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

        goto out;

    }

    /* find the right aio_req from the inflight aio list */

    QLIST_FOREACH(aio_req, &s->inflight_aio_head, aio_siblings) {

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

            break;

        }

    }

    if (!aio_req) {

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

        goto out;

    }

    acb = aio_req->aiocb;

    switch (acb->aiocb_type) {

    case AIOCB_WRITE_UDATA:

        /* this coroutine context is no longer suitable for co_recv

         * because we may send data to update vdi objects */

        s->co_recv = NULL;

        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));

        }

        break;

    case AIOCB_READ_UDATA:

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

                            aio_req->iov_offset);

        if (ret < 0) {

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

            goto out;

        }

        break;

    }

    if (rsp.result != SD_RES_SUCCESS) {

        acb->ret = -EIO;

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

    }

    free_aio_req(s, aio_req);

    if (!acb->nr_pending) {

        /*

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

         * we can switch back to sd_co_readv/writev now.

         */

        acb->aio_done_func(acb);

    }

out:

    s->co_recv = NULL;

}

static void co_read_response(void *opaque)

{

    BDRVSheepdogState *s = opaque;

    if (!s->co_recv) {

        s->co_recv = qemu_coroutine_create(aio_read_response);

    }

    qemu_coroutine_enter(s->co_recv, opaque);

}

static void co_write_request(void *opaque)

{

    BDRVSheepdogState *s = opaque;

    qemu_coroutine_enter(s->co_send, NULL);

}

static int aio_flush_request(void *opaque)

{

    BDRVSheepdogState *s = opaque;

    return !QLIST_EMPTY(&s->inflight_aio_head) ||

        !QLIST_EMPTY(&s->pending_aio_head);

}

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 fd;

    }

    socket_set_nonblock(fd);

    ret = set_nodelay(fd);

    if (ret) {

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

        closesocket(fd);

        return -errno;

    }

    qemu_aio_set_fd_handler(fd, co_read_response, NULL, aio_flush_request, 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 = g_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) {

        g_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 fd;

    }

    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) {

        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);

        if (rsp->result == SD_RES_NO_VDI) {

            ret = -ENOENT;

        } else {

            ret = -EIO;

        }

        goto out;

    }

    *vid = rsp->vdi_id;

    ret = 0;

out:

    closesocket(fd);

    return ret;

}

static int coroutine_fn 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;

    }

    if (s->cache_enabled) {

        hdr.flags |= SD_FLAG_CMD_CACHE;

    }

    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;

    qemu_co_mutex_lock(&s->lock);

    s->co_send = qemu_coroutine_self();

    qemu_aio_set_fd_handler(s->fd, co_read_response, co_write_request,

                            aio_flush_request, s);

    socket_set_cork(s->fd, 1);

    /* send a header */

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

    if (ret < 0) {

        qemu_co_mutex_unlock(&s->lock);

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

        return -errno;

    }

    if (wlen) {

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

        if (ret < 0) {

            qemu_co_mutex_unlock(&s->lock);

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

            return -errno;

        }

    }

    socket_set_cork(s->fd, 0);

    qemu_aio_set_fd_handler(s->fd, co_read_response, NULL,

                            aio_flush_request, s);

    qemu_co_mutex_unlock(&s->lock);

    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, uint8_t cache)

{

    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;

    }

    if (cache) {

        hdr.flags |= SD_FLAG_CMD_CACHE;

    }

    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 ret;

    }

    switch (rsp->result) {

    case SD_RES_SUCCESS:

        return 0;

    default:

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

        return -EIO;

    }

}

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

                       unsigned int datalen, uint64_t offset, uint8_t cache)

{

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

                             cache);

}

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

                        unsigned int datalen, uint64_t offset, int create,

                        uint8_t cache)

{

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

                             cache);

}

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->inflight_aio_head);

    QLIST_INIT(&s->pending_aio_head);

    s->fd = -1;

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

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

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

        ret = -EINVAL;

        goto out;

    }

    s->fd = get_sheep_fd(s);

    if (s->fd < 0) {

        ret = s->fd;

        goto out;

    }

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

    if (ret) {

        goto out;

    }

    if (flags & BDRV_O_CACHE_WB) {

        s->cache_enabled = 1;

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

        if (s->flush_fd < 0) {

            error_report("failed to connect");

            ret = s->flush_fd;

            goto out;

        }

    }

    if (snapid || tag[0] != '\0') {

        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");

        ret = fd;

        goto out;

    }

    buf = g_malloc(SD_INODE_SIZE);

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

                      s->cache_enabled);

    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_co_mutex_init(&s->lock);

    g_free(buf);

    return 0;

out:

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

    if (s->fd >= 0) {

        closesocket(s->fd);

    }

    g_free(buf);

    return ret;

}

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 fd;

    }

    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 ret;

    }

    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 = g_malloc0(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);

    }

    g_free(buf);

    return ret;

}

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

{

    int ret = 0;

    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;

    s = g_malloc0(sizeof(BDRVSheepdogState));

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

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

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

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

        error_report("invalid filename");

        ret = -EINVAL;

        goto out;

    }

    while (options && options->name) {

        if (!strcmp(options->name, BLOCK_OPT_SIZE)) {

            vdi_size = options->value.n;

        } else if (!strcmp(options->name, BLOCK_OPT_BACKING_FILE)) {

            backing_file = options->value.s;

        } else if (!strcmp(options->name, BLOCK_OPT_PREALLOC)) {

            if (!options->value.s || !strcmp(options->value.s, "off")) {

                prealloc = 0;

            } else if (!strcmp(options->value.s, "full")) {

                prealloc = 1;

            } else {

                error_report("Invalid preallocation mode: '%s'",

                             options->value.s);

                ret = -EINVAL;

                goto out;

            }

        }

        options++;

    }

    if (vdi_size > SD_MAX_VDI_SIZE) {

        error_report("too big image size");

        ret = -EINVAL;

        goto out;

    }

    if (backing_file) {

        BlockDriverState *bs;

        BDRVSheepdogState *s;

        BlockDriver *drv;

        /* Currently, only Sheepdog backing image is supported. */

        drv = bdrv_find_protocol(backing_file);

        if (!drv || strcmp(drv->protocol_name, "sheepdog") != 0) {

            error_report("backing_file must be a sheepdog image");

            ret = -EINVAL;

            goto out;

        }

        ret = bdrv_file_open(&bs, backing_file, 0);

        if (ret < 0) {

            goto out;

        }

        s = bs->opaque;

        if (!is_snapshot(&s->inode)) {

            error_report("cannot clone from a non snapshot vdi");