These changes are the raw update to qemu-2.6.
[kvmfornfv.git] / qemu / hw / 9pfs / 9p-proxy.c
diff --git a/qemu/hw/9pfs/9p-proxy.c b/qemu/hw/9pfs/9p-proxy.c
new file mode 100644 (file)
index 0000000..00a4eb2
--- /dev/null
@@ -0,0 +1,1220 @@
+/*
+ * 9p Proxy callback
+ *
+ * Copyright IBM, Corp. 2011
+ *
+ * Authors:
+ * M. Mohan Kumar <mohan@in.ibm.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2.  See
+ * the COPYING file in the top-level directory.
+ */
+#include "qemu/osdep.h"
+#include <sys/socket.h>
+#include <sys/un.h>
+#include "9p.h"
+#include "qemu/cutils.h"
+#include "qemu/error-report.h"
+#include "fsdev/qemu-fsdev.h"
+#include "9p-proxy.h"
+
+typedef struct V9fsProxy {
+    int sockfd;
+    QemuMutex mutex;
+    struct iovec in_iovec;
+    struct iovec out_iovec;
+} V9fsProxy;
+
+/*
+ * Return received file descriptor on success in *status.
+ * errno is also returned on *status (which will be < 0)
+ * return < 0 on transport error.
+ */
+static int v9fs_receivefd(int sockfd, int *status)
+{
+    struct iovec iov;
+    struct msghdr msg;
+    struct cmsghdr *cmsg;
+    int retval, data, fd;
+    union MsgControl msg_control;
+
+    iov.iov_base = &data;
+    iov.iov_len = sizeof(data);
+
+    memset(&msg, 0, sizeof(msg));
+    msg.msg_iov = &iov;
+    msg.msg_iovlen = 1;
+    msg.msg_control = &msg_control;
+    msg.msg_controllen = sizeof(msg_control);
+
+    do {
+        retval = recvmsg(sockfd, &msg, 0);
+    } while (retval < 0 && errno == EINTR);
+    if (retval <= 0) {
+        return retval;
+    }
+    /*
+     * data is set to V9FS_FD_VALID, if ancillary data is sent.  If this
+     * request doesn't need ancillary data (fd) or an error occurred,
+     * data is set to negative errno value.
+     */
+    if (data != V9FS_FD_VALID) {
+        *status = data;
+        return 0;
+    }
+    /*
+     * File descriptor (fd) is sent in the ancillary data. Check if we
+     * indeed received it. One of the reasons to fail to receive it is if
+     * we exceeded the maximum number of file descriptors!
+     */
+    for (cmsg = CMSG_FIRSTHDR(&msg); cmsg; cmsg = CMSG_NXTHDR(&msg, cmsg)) {
+        if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)) ||
+            cmsg->cmsg_level != SOL_SOCKET ||
+            cmsg->cmsg_type != SCM_RIGHTS) {
+            continue;
+        }
+        fd = *((int *)CMSG_DATA(cmsg));
+        *status = fd;
+        return 0;
+    }
+    *status = -ENFILE;  /* Ancillary data sent but not received */
+    return 0;
+}
+
+static ssize_t socket_read(int sockfd, void *buff, size_t size)
+{
+    ssize_t retval, total = 0;
+
+    while (size) {
+        retval = read(sockfd, buff, size);
+        if (retval == 0) {
+            return -EIO;
+        }
+        if (retval < 0) {
+            if (errno == EINTR) {
+                continue;
+            }
+            return -errno;
+        }
+        size -= retval;
+        buff += retval;
+        total += retval;
+    }
+    return total;
+}
+
+/* Converts proxy_statfs to VFS statfs structure */
+static void prstatfs_to_statfs(struct statfs *stfs, ProxyStatFS *prstfs)
+{
+    memset(stfs, 0, sizeof(*stfs));
+    stfs->f_type = prstfs->f_type;
+    stfs->f_bsize = prstfs->f_bsize;
+    stfs->f_blocks = prstfs->f_blocks;
+    stfs->f_bfree = prstfs->f_bfree;
+    stfs->f_bavail = prstfs->f_bavail;
+    stfs->f_files = prstfs->f_files;
+    stfs->f_ffree = prstfs->f_ffree;
+    stfs->f_fsid.__val[0] = prstfs->f_fsid[0] & 0xFFFFFFFFU;
+    stfs->f_fsid.__val[1] = prstfs->f_fsid[1] >> 32 & 0xFFFFFFFFU;
+    stfs->f_namelen = prstfs->f_namelen;
+    stfs->f_frsize = prstfs->f_frsize;
+}
+
+/* Converts proxy_stat structure to VFS stat structure */
+static void prstat_to_stat(struct stat *stbuf, ProxyStat *prstat)
+{
+   memset(stbuf, 0, sizeof(*stbuf));
+   stbuf->st_dev = prstat->st_dev;
+   stbuf->st_ino = prstat->st_ino;
+   stbuf->st_nlink = prstat->st_nlink;
+   stbuf->st_mode = prstat->st_mode;
+   stbuf->st_uid = prstat->st_uid;
+   stbuf->st_gid = prstat->st_gid;
+   stbuf->st_rdev = prstat->st_rdev;
+   stbuf->st_size = prstat->st_size;
+   stbuf->st_blksize = prstat->st_blksize;
+   stbuf->st_blocks = prstat->st_blocks;
+   stbuf->st_atim.tv_sec = prstat->st_atim_sec;
+   stbuf->st_atim.tv_nsec = prstat->st_atim_nsec;
+   stbuf->st_mtime = prstat->st_mtim_sec;
+   stbuf->st_mtim.tv_nsec = prstat->st_mtim_nsec;
+   stbuf->st_ctime = prstat->st_ctim_sec;
+   stbuf->st_ctim.tv_nsec = prstat->st_ctim_nsec;
+}
+
+/*
+ * Response contains two parts
+ * {header, data}
+ * header.type == T_ERROR, data -> -errno
+ * header.type == T_SUCCESS, data -> response
+ * size of errno/response is given by header.size
+ * returns < 0, on transport error. response is
+ * valid only if status >= 0.
+ */
+static int v9fs_receive_response(V9fsProxy *proxy, int type,
+                                 int *status, void *response)
+{
+    int retval;
+    ProxyHeader header;
+    struct iovec *reply = &proxy->in_iovec;
+
+    *status = 0;
+    reply->iov_len = 0;
+    retval = socket_read(proxy->sockfd, reply->iov_base, PROXY_HDR_SZ);
+    if (retval < 0) {
+        return retval;
+    }
+    reply->iov_len = PROXY_HDR_SZ;
+    proxy_unmarshal(reply, 0, "dd", &header.type, &header.size);
+    /*
+     * if response size > PROXY_MAX_IO_SZ, read the response but ignore it and
+     * return -ENOBUFS
+     */
+    if (header.size > PROXY_MAX_IO_SZ) {
+        int count;
+        while (header.size > 0) {
+            count = MIN(PROXY_MAX_IO_SZ, header.size);
+            count = socket_read(proxy->sockfd, reply->iov_base, count);
+            if (count < 0) {
+                return count;
+            }
+            header.size -= count;
+        }
+        *status = -ENOBUFS;
+        return 0;
+    }
+
+    retval = socket_read(proxy->sockfd,
+                         reply->iov_base + PROXY_HDR_SZ, header.size);
+    if (retval < 0) {
+        return retval;
+    }
+    reply->iov_len += header.size;
+    /* there was an error during processing request */
+    if (header.type == T_ERROR) {
+        int ret;
+        ret = proxy_unmarshal(reply, PROXY_HDR_SZ, "d", status);
+        if (ret < 0) {
+            *status = ret;
+        }
+        return 0;
+    }
+
+    switch (type) {
+    case T_LSTAT: {
+        ProxyStat prstat;
+        retval = proxy_unmarshal(reply, PROXY_HDR_SZ,
+                                 "qqqdddqqqqqqqqqq", &prstat.st_dev,
+                                 &prstat.st_ino, &prstat.st_nlink,
+                                 &prstat.st_mode, &prstat.st_uid,
+                                 &prstat.st_gid, &prstat.st_rdev,
+                                 &prstat.st_size, &prstat.st_blksize,
+                                 &prstat.st_blocks,
+                                 &prstat.st_atim_sec, &prstat.st_atim_nsec,
+                                 &prstat.st_mtim_sec, &prstat.st_mtim_nsec,
+                                 &prstat.st_ctim_sec, &prstat.st_ctim_nsec);
+        prstat_to_stat(response, &prstat);
+        break;
+    }
+    case T_STATFS: {
+        ProxyStatFS prstfs;
+        retval = proxy_unmarshal(reply, PROXY_HDR_SZ,
+                                 "qqqqqqqqqqq", &prstfs.f_type,
+                                 &prstfs.f_bsize, &prstfs.f_blocks,
+                                 &prstfs.f_bfree, &prstfs.f_bavail,
+                                 &prstfs.f_files, &prstfs.f_ffree,
+                                 &prstfs.f_fsid[0], &prstfs.f_fsid[1],
+                                 &prstfs.f_namelen, &prstfs.f_frsize);
+        prstatfs_to_statfs(response, &prstfs);
+        break;
+    }
+    case T_READLINK: {
+        V9fsString target;
+        v9fs_string_init(&target);
+        retval = proxy_unmarshal(reply, PROXY_HDR_SZ, "s", &target);
+        strcpy(response, target.data);
+        v9fs_string_free(&target);
+        break;
+    }
+    case T_LGETXATTR:
+    case T_LLISTXATTR: {
+        V9fsString xattr;
+        v9fs_string_init(&xattr);
+        retval = proxy_unmarshal(reply, PROXY_HDR_SZ, "s", &xattr);
+        memcpy(response, xattr.data, xattr.size);
+        v9fs_string_free(&xattr);
+        break;
+    }
+    case T_GETVERSION:
+        proxy_unmarshal(reply, PROXY_HDR_SZ, "q", response);
+        break;
+    default:
+        return -1;
+    }
+    if (retval < 0) {
+        *status  = retval;
+    }
+    return 0;
+}
+
+/*
+ * return < 0 on transport error.
+ * *status is valid only if return >= 0
+ */
+static int v9fs_receive_status(V9fsProxy *proxy,
+                               struct iovec *reply, int *status)
+{
+    int retval;
+    ProxyHeader header;
+
+    *status = 0;
+    reply->iov_len = 0;
+    retval = socket_read(proxy->sockfd, reply->iov_base, PROXY_HDR_SZ);
+    if (retval < 0) {
+        return retval;
+    }
+    reply->iov_len = PROXY_HDR_SZ;
+    proxy_unmarshal(reply, 0, "dd", &header.type, &header.size);
+    if (header.size != sizeof(int)) {
+        *status = -ENOBUFS;
+        return 0;
+    }
+    retval = socket_read(proxy->sockfd,
+                         reply->iov_base + PROXY_HDR_SZ, header.size);
+    if (retval < 0) {
+        return retval;
+    }
+    reply->iov_len += header.size;
+    proxy_unmarshal(reply, PROXY_HDR_SZ, "d", status);
+    return 0;
+}
+
+/*
+ * Proxy->header and proxy->request written to socket by QEMU process.
+ * This request read by proxy helper process
+ * returns 0 on success and -errno on error
+ */
+static int v9fs_request(V9fsProxy *proxy, int type,
+                        void *response, const char *fmt, ...)
+{
+    dev_t rdev;
+    va_list ap;
+    int size = 0;
+    int retval = 0;
+    uint64_t offset;
+    ProxyHeader header = { 0, 0};
+    struct timespec spec[2];
+    int flags, mode, uid, gid;
+    V9fsString *name, *value;
+    V9fsString *path, *oldpath;
+    struct iovec *iovec = NULL, *reply = NULL;
+
+    qemu_mutex_lock(&proxy->mutex);
+
+    if (proxy->sockfd == -1) {
+        retval = -EIO;
+        goto err_out;
+    }
+    iovec = &proxy->out_iovec;
+    reply = &proxy->in_iovec;
+    va_start(ap, fmt);
+    switch (type) {
+    case T_OPEN:
+        path = va_arg(ap, V9fsString *);
+        flags = va_arg(ap, int);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sd", path, flags);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_OPEN;
+        }
+        break;
+    case T_CREATE:
+        path = va_arg(ap, V9fsString *);
+        flags = va_arg(ap, int);
+        mode = va_arg(ap, int);
+        uid = va_arg(ap, int);
+        gid = va_arg(ap, int);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sdddd", path,
+                                    flags, mode, uid, gid);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_CREATE;
+        }
+        break;
+    case T_MKNOD:
+        path = va_arg(ap, V9fsString *);
+        mode = va_arg(ap, int);
+        rdev = va_arg(ap, long int);
+        uid = va_arg(ap, int);
+        gid = va_arg(ap, int);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ddsdq",
+                                    uid, gid, path, mode, rdev);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_MKNOD;
+        }
+        break;
+    case T_MKDIR:
+        path = va_arg(ap, V9fsString *);
+        mode = va_arg(ap, int);
+        uid = va_arg(ap, int);
+        gid = va_arg(ap, int);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ddsd",
+                                    uid, gid, path, mode);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_MKDIR;
+        }
+        break;
+    case T_SYMLINK:
+        oldpath = va_arg(ap, V9fsString *);
+        path = va_arg(ap, V9fsString *);
+        uid = va_arg(ap, int);
+        gid = va_arg(ap, int);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ddss",
+                                    uid, gid, oldpath, path);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_SYMLINK;
+        }
+        break;
+    case T_LINK:
+        oldpath = va_arg(ap, V9fsString *);
+        path = va_arg(ap, V9fsString *);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ss",
+                                    oldpath, path);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_LINK;
+        }
+        break;
+    case T_LSTAT:
+        path = va_arg(ap, V9fsString *);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "s", path);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_LSTAT;
+        }
+        break;
+    case T_READLINK:
+        path = va_arg(ap, V9fsString *);
+        size = va_arg(ap, int);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sd", path, size);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_READLINK;
+        }
+        break;
+    case T_STATFS:
+        path = va_arg(ap, V9fsString *);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "s", path);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_STATFS;
+        }
+        break;
+    case T_CHMOD:
+        path = va_arg(ap, V9fsString *);
+        mode = va_arg(ap, int);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sd", path, mode);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_CHMOD;
+        }
+        break;
+    case T_CHOWN:
+        path = va_arg(ap, V9fsString *);
+        uid = va_arg(ap, int);
+        gid = va_arg(ap, int);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sdd", path, uid, gid);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_CHOWN;
+        }
+        break;
+    case T_TRUNCATE:
+        path = va_arg(ap, V9fsString *);
+        offset = va_arg(ap, uint64_t);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sq", path, offset);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_TRUNCATE;
+        }
+        break;
+    case T_UTIME:
+        path = va_arg(ap, V9fsString *);
+        spec[0].tv_sec = va_arg(ap, long);
+        spec[0].tv_nsec = va_arg(ap, long);
+        spec[1].tv_sec = va_arg(ap, long);
+        spec[1].tv_nsec = va_arg(ap, long);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sqqqq", path,
+                                    spec[0].tv_sec, spec[1].tv_nsec,
+                                    spec[1].tv_sec, spec[1].tv_nsec);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_UTIME;
+        }
+        break;
+    case T_RENAME:
+        oldpath = va_arg(ap, V9fsString *);
+        path = va_arg(ap, V9fsString *);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ss", oldpath, path);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_RENAME;
+        }
+        break;
+    case T_REMOVE:
+        path = va_arg(ap, V9fsString *);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "s", path);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_REMOVE;
+        }
+        break;
+    case T_LGETXATTR:
+        size = va_arg(ap, int);
+        path = va_arg(ap, V9fsString *);
+        name = va_arg(ap, V9fsString *);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ,
+                                    "dss", size, path, name);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_LGETXATTR;
+        }
+        break;
+    case T_LLISTXATTR:
+        size = va_arg(ap, int);
+        path = va_arg(ap, V9fsString *);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ds", size, path);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_LLISTXATTR;
+        }
+        break;
+    case T_LSETXATTR:
+        path = va_arg(ap, V9fsString *);
+        name = va_arg(ap, V9fsString *);
+        value = va_arg(ap, V9fsString *);
+        size = va_arg(ap, int);
+        flags = va_arg(ap, int);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sssdd",
+                                    path, name, value, size, flags);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_LSETXATTR;
+        }
+        break;
+    case T_LREMOVEXATTR:
+        path = va_arg(ap, V9fsString *);
+        name = va_arg(ap, V9fsString *);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ss", path, name);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_LREMOVEXATTR;
+        }
+        break;
+    case T_GETVERSION:
+        path = va_arg(ap, V9fsString *);
+        retval = proxy_marshal(iovec, PROXY_HDR_SZ, "s", path);
+        if (retval > 0) {
+            header.size = retval;
+            header.type = T_GETVERSION;
+        }
+        break;
+    default:
+        error_report("Invalid type %d", type);
+        retval = -EINVAL;
+        break;
+    }
+    va_end(ap);
+
+    if (retval < 0) {
+        goto err_out;
+    }
+
+    /* marshal the header details */
+    proxy_marshal(iovec, 0, "dd", header.type, header.size);
+    header.size += PROXY_HDR_SZ;
+
+    retval = qemu_write_full(proxy->sockfd, iovec->iov_base, header.size);
+    if (retval != header.size) {
+        goto close_error;
+    }
+
+    switch (type) {
+    case T_OPEN:
+    case T_CREATE:
+        /*
+         * A file descriptor is returned as response for
+         * T_OPEN,T_CREATE on success
+         */
+        if (v9fs_receivefd(proxy->sockfd, &retval) < 0) {
+            goto close_error;
+        }
+        break;
+    case T_MKNOD:
+    case T_MKDIR:
+    case T_SYMLINK:
+    case T_LINK:
+    case T_CHMOD:
+    case T_CHOWN:
+    case T_RENAME:
+    case T_TRUNCATE:
+    case T_UTIME:
+    case T_REMOVE:
+    case T_LSETXATTR:
+    case T_LREMOVEXATTR:
+        if (v9fs_receive_status(proxy, reply, &retval) < 0) {
+            goto close_error;
+        }
+        break;
+    case T_LSTAT:
+    case T_READLINK:
+    case T_STATFS:
+    case T_GETVERSION:
+        if (v9fs_receive_response(proxy, type, &retval, response) < 0) {
+            goto close_error;
+        }
+        break;
+    case T_LGETXATTR:
+    case T_LLISTXATTR:
+        if (!size) {
+            if (v9fs_receive_status(proxy, reply, &retval) < 0) {
+                goto close_error;
+            }
+        } else {
+            if (v9fs_receive_response(proxy, type, &retval, response) < 0) {
+                goto close_error;
+            }
+        }
+        break;
+    }
+
+err_out:
+    qemu_mutex_unlock(&proxy->mutex);
+    return retval;
+
+close_error:
+    close(proxy->sockfd);
+    proxy->sockfd = -1;
+    qemu_mutex_unlock(&proxy->mutex);
+    return -EIO;
+}
+
+static int proxy_lstat(FsContext *fs_ctx, V9fsPath *fs_path, struct stat *stbuf)
+{
+    int retval;
+    retval = v9fs_request(fs_ctx->private, T_LSTAT, stbuf, "s", fs_path);
+    if (retval < 0) {
+        errno = -retval;
+        return -1;
+    }
+    return retval;
+}
+
+static ssize_t proxy_readlink(FsContext *fs_ctx, V9fsPath *fs_path,
+                              char *buf, size_t bufsz)
+{
+    int retval;
+    retval = v9fs_request(fs_ctx->private, T_READLINK, buf, "sd",
+                          fs_path, bufsz);
+    if (retval < 0) {
+        errno = -retval;
+        return -1;
+    }
+    return strlen(buf);
+}
+
+static int proxy_close(FsContext *ctx, V9fsFidOpenState *fs)
+{
+    return close(fs->fd);
+}
+
+static int proxy_closedir(FsContext *ctx, V9fsFidOpenState *fs)
+{
+    return closedir(fs->dir);
+}
+
+static int proxy_open(FsContext *ctx, V9fsPath *fs_path,
+                      int flags, V9fsFidOpenState *fs)
+{
+    fs->fd = v9fs_request(ctx->private, T_OPEN, NULL, "sd", fs_path, flags);
+    if (fs->fd < 0) {
+        errno = -fs->fd;
+        fs->fd = -1;
+    }
+    return fs->fd;
+}
+
+static int proxy_opendir(FsContext *ctx,
+                         V9fsPath *fs_path, V9fsFidOpenState *fs)
+{
+    int serrno, fd;
+
+    fs->dir = NULL;
+    fd = v9fs_request(ctx->private, T_OPEN, NULL, "sd", fs_path, O_DIRECTORY);
+    if (fd < 0) {
+        errno = -fd;
+        return -1;
+    }
+    fs->dir = fdopendir(fd);
+    if (!fs->dir) {
+        serrno = errno;
+        close(fd);
+        errno = serrno;
+        return -1;
+    }
+    return 0;
+}
+
+static void proxy_rewinddir(FsContext *ctx, V9fsFidOpenState *fs)
+{
+    rewinddir(fs->dir);
+}
+
+static off_t proxy_telldir(FsContext *ctx, V9fsFidOpenState *fs)
+{
+    return telldir(fs->dir);
+}
+
+static int proxy_readdir_r(FsContext *ctx, V9fsFidOpenState *fs,
+                           struct dirent *entry,
+                           struct dirent **result)
+{
+    return readdir_r(fs->dir, entry, result);
+}
+
+static void proxy_seekdir(FsContext *ctx, V9fsFidOpenState *fs, off_t off)
+{
+    seekdir(fs->dir, off);
+}
+
+static ssize_t proxy_preadv(FsContext *ctx, V9fsFidOpenState *fs,
+                            const struct iovec *iov,
+                            int iovcnt, off_t offset)
+{
+    ssize_t ret;
+#ifdef CONFIG_PREADV
+    ret = preadv(fs->fd, iov, iovcnt, offset);
+#else
+    ret = lseek(fs->fd, offset, SEEK_SET);
+    if (ret >= 0) {
+        ret = readv(fs->fd, iov, iovcnt);
+    }
+#endif
+    return ret;
+}
+
+static ssize_t proxy_pwritev(FsContext *ctx, V9fsFidOpenState *fs,
+                             const struct iovec *iov,
+                             int iovcnt, off_t offset)
+{
+    ssize_t ret;
+
+#ifdef CONFIG_PREADV
+    ret = pwritev(fs->fd, iov, iovcnt, offset);
+#else
+    ret = lseek(fs->fd, offset, SEEK_SET);
+    if (ret >= 0) {
+        ret = writev(fs->fd, iov, iovcnt);
+    }
+#endif
+#ifdef CONFIG_SYNC_FILE_RANGE
+    if (ret > 0 && ctx->export_flags & V9FS_IMMEDIATE_WRITEOUT) {
+        /*
+         * Initiate a writeback. This is not a data integrity sync.
+         * We want to ensure that we don't leave dirty pages in the cache
+         * after write when writeout=immediate is sepcified.
+         */
+        sync_file_range(fs->fd, offset, ret,
+                        SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE);
+    }
+#endif
+    return ret;
+}
+
+static int proxy_chmod(FsContext *fs_ctx, V9fsPath *fs_path, FsCred *credp)
+{
+    int retval;
+    retval = v9fs_request(fs_ctx->private, T_CHMOD, NULL, "sd",
+                          fs_path, credp->fc_mode);
+    if (retval < 0) {
+        errno = -retval;
+    }
+    return retval;
+}
+
+static int proxy_mknod(FsContext *fs_ctx, V9fsPath *dir_path,
+                       const char *name, FsCred *credp)
+{
+    int retval;
+    V9fsString fullname;
+
+    v9fs_string_init(&fullname);
+    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
+
+    retval = v9fs_request(fs_ctx->private, T_MKNOD, NULL, "sdqdd",
+                          &fullname, credp->fc_mode, credp->fc_rdev,
+                          credp->fc_uid, credp->fc_gid);
+    v9fs_string_free(&fullname);
+    if (retval < 0) {
+        errno = -retval;
+        retval = -1;
+    }
+    return retval;
+}
+
+static int proxy_mkdir(FsContext *fs_ctx, V9fsPath *dir_path,
+                       const char *name, FsCred *credp)
+{
+    int retval;
+    V9fsString fullname;
+
+    v9fs_string_init(&fullname);
+    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
+
+    retval = v9fs_request(fs_ctx->private, T_MKDIR, NULL, "sddd", &fullname,
+                          credp->fc_mode, credp->fc_uid, credp->fc_gid);
+    v9fs_string_free(&fullname);
+    if (retval < 0) {
+        errno = -retval;
+        retval = -1;
+    }
+    v9fs_string_free(&fullname);
+    return retval;
+}
+
+static int proxy_fstat(FsContext *fs_ctx, int fid_type,
+                       V9fsFidOpenState *fs, struct stat *stbuf)
+{
+    int fd;
+
+    if (fid_type == P9_FID_DIR) {
+        fd = dirfd(fs->dir);
+    } else {
+        fd = fs->fd;
+    }
+    return fstat(fd, stbuf);
+}
+
+static int proxy_open2(FsContext *fs_ctx, V9fsPath *dir_path, const char *name,
+                       int flags, FsCred *credp, V9fsFidOpenState *fs)
+{
+    V9fsString fullname;
+
+    v9fs_string_init(&fullname);
+    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
+
+    fs->fd = v9fs_request(fs_ctx->private, T_CREATE, NULL, "sdddd",
+                          &fullname, flags, credp->fc_mode,
+                          credp->fc_uid, credp->fc_gid);
+    v9fs_string_free(&fullname);
+    if (fs->fd < 0) {
+        errno = -fs->fd;
+        fs->fd = -1;
+    }
+    return fs->fd;
+}
+
+static int proxy_symlink(FsContext *fs_ctx, const char *oldpath,
+                         V9fsPath *dir_path, const char *name, FsCred *credp)
+{
+    int retval;
+    V9fsString fullname, target;
+
+    v9fs_string_init(&fullname);
+    v9fs_string_init(&target);
+
+    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
+    v9fs_string_sprintf(&target, "%s", oldpath);
+
+    retval = v9fs_request(fs_ctx->private, T_SYMLINK, NULL, "ssdd",
+                          &target, &fullname, credp->fc_uid, credp->fc_gid);
+    v9fs_string_free(&fullname);
+    v9fs_string_free(&target);
+    if (retval < 0) {
+        errno = -retval;
+        retval = -1;
+    }
+    return retval;
+}
+
+static int proxy_link(FsContext *ctx, V9fsPath *oldpath,
+                      V9fsPath *dirpath, const char *name)
+{
+    int retval;
+    V9fsString newpath;
+
+    v9fs_string_init(&newpath);
+    v9fs_string_sprintf(&newpath, "%s/%s", dirpath->data, name);
+
+    retval = v9fs_request(ctx->private, T_LINK, NULL, "ss", oldpath, &newpath);
+    v9fs_string_free(&newpath);
+    if (retval < 0) {
+        errno = -retval;
+        retval = -1;
+    }
+    return retval;
+}
+
+static int proxy_truncate(FsContext *ctx, V9fsPath *fs_path, off_t size)
+{
+    int retval;
+
+    retval = v9fs_request(ctx->private, T_TRUNCATE, NULL, "sq", fs_path, size);
+    if (retval < 0) {
+        errno = -retval;
+        return -1;
+    }
+    return 0;
+}
+
+static int proxy_rename(FsContext *ctx, const char *oldpath,
+                        const char *newpath)
+{
+    int retval;
+    V9fsString oldname, newname;
+
+    v9fs_string_init(&oldname);
+    v9fs_string_init(&newname);
+
+    v9fs_string_sprintf(&oldname, "%s", oldpath);
+    v9fs_string_sprintf(&newname, "%s", newpath);
+    retval = v9fs_request(ctx->private, T_RENAME, NULL, "ss",
+                          &oldname, &newname);
+    v9fs_string_free(&oldname);
+    v9fs_string_free(&newname);
+    if (retval < 0) {
+        errno = -retval;
+    }
+    return retval;
+}
+
+static int proxy_chown(FsContext *fs_ctx, V9fsPath *fs_path, FsCred *credp)
+{
+    int retval;
+    retval = v9fs_request(fs_ctx->private, T_CHOWN, NULL, "sdd",
+                          fs_path, credp->fc_uid, credp->fc_gid);
+    if (retval < 0) {
+        errno = -retval;
+    }
+    return retval;
+}
+
+static int proxy_utimensat(FsContext *s, V9fsPath *fs_path,
+                           const struct timespec *buf)
+{
+    int retval;
+    retval = v9fs_request(s->private, T_UTIME, NULL, "sqqqq",
+                          fs_path,
+                          buf[0].tv_sec, buf[0].tv_nsec,
+                          buf[1].tv_sec, buf[1].tv_nsec);
+    if (retval < 0) {
+        errno = -retval;
+    }
+    return retval;
+}
+
+static int proxy_remove(FsContext *ctx, const char *path)
+{
+    int retval;
+    V9fsString name;
+    v9fs_string_init(&name);
+    v9fs_string_sprintf(&name, "%s", path);
+    retval = v9fs_request(ctx->private, T_REMOVE, NULL, "s", &name);
+    v9fs_string_free(&name);
+    if (retval < 0) {
+        errno = -retval;
+    }
+    return retval;
+}
+
+static int proxy_fsync(FsContext *ctx, int fid_type,
+                       V9fsFidOpenState *fs, int datasync)
+{
+    int fd;
+
+    if (fid_type == P9_FID_DIR) {
+        fd = dirfd(fs->dir);
+    } else {
+        fd = fs->fd;
+    }
+
+    if (datasync) {
+        return qemu_fdatasync(fd);
+    } else {
+        return fsync(fd);
+    }
+}
+
+static int proxy_statfs(FsContext *s, V9fsPath *fs_path, struct statfs *stbuf)
+{
+    int retval;
+    retval = v9fs_request(s->private, T_STATFS, stbuf, "s", fs_path);
+    if (retval < 0) {
+        errno = -retval;
+        return -1;
+    }
+    return retval;
+}
+
+static ssize_t proxy_lgetxattr(FsContext *ctx, V9fsPath *fs_path,
+                               const char *name, void *value, size_t size)
+{
+    int retval;
+    V9fsString xname;
+
+    v9fs_string_init(&xname);
+    v9fs_string_sprintf(&xname, "%s", name);
+    retval = v9fs_request(ctx->private, T_LGETXATTR, value, "dss", size,
+                          fs_path, &xname);
+    v9fs_string_free(&xname);
+    if (retval < 0) {
+        errno = -retval;
+    }
+    return retval;
+}
+
+static ssize_t proxy_llistxattr(FsContext *ctx, V9fsPath *fs_path,
+                                void *value, size_t size)
+{
+    int retval;
+    retval = v9fs_request(ctx->private, T_LLISTXATTR, value, "ds", size,
+                        fs_path);
+    if (retval < 0) {
+        errno = -retval;
+    }
+    return retval;
+}
+
+static int proxy_lsetxattr(FsContext *ctx, V9fsPath *fs_path, const char *name,
+                           void *value, size_t size, int flags)
+{
+    int retval;
+    V9fsString xname, xvalue;
+
+    v9fs_string_init(&xname);
+    v9fs_string_sprintf(&xname, "%s", name);
+
+    v9fs_string_init(&xvalue);
+    xvalue.size = size;
+    xvalue.data = g_malloc(size);
+    memcpy(xvalue.data, value, size);
+
+    retval = v9fs_request(ctx->private, T_LSETXATTR, value, "sssdd",
+                          fs_path, &xname, &xvalue, size, flags);
+    v9fs_string_free(&xname);
+    v9fs_string_free(&xvalue);
+    if (retval < 0) {
+        errno = -retval;
+    }
+    return retval;
+}
+
+static int proxy_lremovexattr(FsContext *ctx, V9fsPath *fs_path,
+                              const char *name)
+{
+    int retval;
+    V9fsString xname;
+
+    v9fs_string_init(&xname);
+    v9fs_string_sprintf(&xname, "%s", name);
+    retval = v9fs_request(ctx->private, T_LREMOVEXATTR, NULL, "ss",
+                          fs_path, &xname);
+    v9fs_string_free(&xname);
+    if (retval < 0) {
+        errno = -retval;
+    }
+    return retval;
+}
+
+static int proxy_name_to_path(FsContext *ctx, V9fsPath *dir_path,
+                              const char *name, V9fsPath *target)
+{
+    if (dir_path) {
+        v9fs_string_sprintf((V9fsString *)target, "%s/%s",
+                            dir_path->data, name);
+    } else {
+        v9fs_string_sprintf((V9fsString *)target, "%s", name);
+    }
+    /* Bump the size for including terminating NULL */
+    target->size++;
+    return 0;
+}
+
+static int proxy_renameat(FsContext *ctx, V9fsPath *olddir,
+                          const char *old_name, V9fsPath *newdir,
+                          const char *new_name)
+{
+    int ret;
+    V9fsString old_full_name, new_full_name;
+
+    v9fs_string_init(&old_full_name);
+    v9fs_string_init(&new_full_name);
+
+    v9fs_string_sprintf(&old_full_name, "%s/%s", olddir->data, old_name);
+    v9fs_string_sprintf(&new_full_name, "%s/%s", newdir->data, new_name);
+
+    ret = proxy_rename(ctx, old_full_name.data, new_full_name.data);
+    v9fs_string_free(&old_full_name);
+    v9fs_string_free(&new_full_name);
+    return ret;
+}
+
+static int proxy_unlinkat(FsContext *ctx, V9fsPath *dir,
+                          const char *name, int flags)
+{
+    int ret;
+    V9fsString fullname;
+    v9fs_string_init(&fullname);
+
+    v9fs_string_sprintf(&fullname, "%s/%s", dir->data, name);
+    ret = proxy_remove(ctx, fullname.data);
+    v9fs_string_free(&fullname);
+
+    return ret;
+}
+
+static int proxy_ioc_getversion(FsContext *fs_ctx, V9fsPath *path,
+                                mode_t st_mode, uint64_t *st_gen)
+{
+    int err;
+
+    /* Do not try to open special files like device nodes, fifos etc
+     * we can get fd for regular files and directories only
+     */
+    if (!S_ISREG(st_mode) && !S_ISDIR(st_mode)) {
+        errno = ENOTTY;
+        return -1;
+    }
+    err = v9fs_request(fs_ctx->private, T_GETVERSION, st_gen, "s", path);
+    if (err < 0) {
+        errno = -err;
+        err = -1;
+    }
+    return err;
+}
+
+static int connect_namedsocket(const char *path)
+{
+    int sockfd, size;
+    struct sockaddr_un helper;
+
+    if (strlen(path) >= sizeof(helper.sun_path)) {
+        error_report("Socket name too long");
+        return -1;
+    }
+    sockfd = socket(AF_UNIX, SOCK_STREAM, 0);
+    if (sockfd < 0) {
+        error_report("Failed to create socket: %s", strerror(errno));
+        return -1;
+    }
+    strcpy(helper.sun_path, path);
+    helper.sun_family = AF_UNIX;
+    size = strlen(helper.sun_path) + sizeof(helper.sun_family);
+    if (connect(sockfd, (struct sockaddr *)&helper, size) < 0) {
+        error_report("Failed to connect to %s: %s", path, strerror(errno));
+        close(sockfd);
+        return -1;
+    }
+
+    /* remove the socket for security reasons */
+    unlink(path);
+    return sockfd;
+}
+
+static int proxy_parse_opts(QemuOpts *opts, struct FsDriverEntry *fs)
+{
+    const char *socket = qemu_opt_get(opts, "socket");
+    const char *sock_fd = qemu_opt_get(opts, "sock_fd");
+
+    if (!socket && !sock_fd) {
+        error_report("Must specify either socket or sock_fd");
+        return -1;
+    }
+    if (socket && sock_fd) {
+        error_report("Both socket and sock_fd options specified");
+        return -1;
+    }
+    if (socket) {
+        fs->path = g_strdup(socket);
+        fs->export_flags = V9FS_PROXY_SOCK_NAME;
+    } else {
+        fs->path = g_strdup(sock_fd);
+        fs->export_flags = V9FS_PROXY_SOCK_FD;
+    }
+    return 0;
+}
+
+static int proxy_init(FsContext *ctx)
+{
+    V9fsProxy *proxy = g_malloc(sizeof(V9fsProxy));
+    int sock_id;
+
+    if (ctx->export_flags & V9FS_PROXY_SOCK_NAME) {
+        sock_id = connect_namedsocket(ctx->fs_root);
+    } else {
+        sock_id = atoi(ctx->fs_root);
+        if (sock_id < 0) {
+            error_report("Socket descriptor not initialized");
+        }
+    }
+    if (sock_id < 0) {
+        g_free(proxy);
+        return -1;
+    }
+    g_free(ctx->fs_root);
+    ctx->fs_root = NULL;
+
+    proxy->in_iovec.iov_base  = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
+    proxy->in_iovec.iov_len   = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
+    proxy->out_iovec.iov_base = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
+    proxy->out_iovec.iov_len  = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
+
+    ctx->private = proxy;
+    proxy->sockfd = sock_id;
+    qemu_mutex_init(&proxy->mutex);
+
+    ctx->export_flags |= V9FS_PATHNAME_FSCONTEXT;
+    ctx->exops.get_st_gen = proxy_ioc_getversion;
+    return 0;
+}
+
+FileOperations proxy_ops = {
+    .parse_opts   = proxy_parse_opts,
+    .init         = proxy_init,
+    .lstat        = proxy_lstat,
+    .readlink     = proxy_readlink,
+    .close        = proxy_close,
+    .closedir     = proxy_closedir,
+    .open         = proxy_open,
+    .opendir      = proxy_opendir,
+    .rewinddir    = proxy_rewinddir,
+    .telldir      = proxy_telldir,
+    .readdir_r    = proxy_readdir_r,
+    .seekdir      = proxy_seekdir,
+    .preadv       = proxy_preadv,
+    .pwritev      = proxy_pwritev,
+    .chmod        = proxy_chmod,
+    .mknod        = proxy_mknod,
+    .mkdir        = proxy_mkdir,
+    .fstat        = proxy_fstat,
+    .open2        = proxy_open2,
+    .symlink      = proxy_symlink,
+    .link         = proxy_link,
+    .truncate     = proxy_truncate,
+    .rename       = proxy_rename,
+    .chown        = proxy_chown,
+    .utimensat    = proxy_utimensat,
+    .remove       = proxy_remove,
+    .fsync        = proxy_fsync,
+    .statfs       = proxy_statfs,
+    .lgetxattr    = proxy_lgetxattr,
+    .llistxattr   = proxy_llistxattr,
+    .lsetxattr    = proxy_lsetxattr,
+    .lremovexattr = proxy_lremovexattr,
+    .name_to_path = proxy_name_to_path,
+    .renameat     = proxy_renameat,
+    .unlinkat     = proxy_unlinkat,
+};