hw/9pfs: Add CephFS support in VirtFS
diff mbox

Message ID 1456933303-1225-1-git-send-email-scaleqiao@gmail.com
State New
Headers show

Commit Message

Jevon Qiao March 2, 2016, 3:41 p.m. UTC
Ceph as a promising unified distributed storage system is widely used in the
world of OpenStack. OpenStack users deploying Ceph for block (Cinder) and
object (S3/Swift) are unsurprisingly looking at Manila and CephFS to round out
a unified storage solution. Since the typical hypervisor people are using is
Qemu/KVM, it is necessary to provide a high performance, easy to use, file
system service in it. VirtFS aims to offers paravirtualized system services and
simple passthrough for directories from host to guest, which currently only
support local file system, this patch wants to add CephFS support in VirtFS.

Signed-off-by: Jevon Qiao <scaleqiao@gmail.com>
---
 configure                         |  33 ++
 fsdev/qemu-fsdev.c                |   1 +
 fsdev/qemu-fsdev.h                |   3 +-
 hw/9pfs/9p-cephfs.c               | 739 ++++++++++++++++++++++++++++++++++++++
 hw/9pfs/Makefile.objs             |   3 +
 scripts/analyse-9p-simpletrace.py |  96 +++++
 trace-events                      |  33 ++
 7 files changed, 907 insertions(+), 1 deletion(-)
 create mode 100644 hw/9pfs/9p-cephfs.c

Comments

Jevon Qiao March 8, 2016, 12:51 a.m. UTC | #1
Any further comment on this patch?

Thanks,
Jevon
On 2/3/16 23:41, Jevon Qiao wrote:
> Ceph as a promising unified distributed storage system is widely used in the
> world of OpenStack. OpenStack users deploying Ceph for block (Cinder) and
> object (S3/Swift) are unsurprisingly looking at Manila and CephFS to round out
> a unified storage solution. Since the typical hypervisor people are using is
> Qemu/KVM, it is necessary to provide a high performance, easy to use, file
> system service in it. VirtFS aims to offers paravirtualized system services and
> simple passthrough for directories from host to guest, which currently only
> support local file system, this patch wants to add CephFS support in VirtFS.
>
> Signed-off-by: Jevon Qiao <scaleqiao@gmail.com>
> ---
>   configure                         |  33 ++
>   fsdev/qemu-fsdev.c                |   1 +
>   fsdev/qemu-fsdev.h                |   3 +-
>   hw/9pfs/9p-cephfs.c               | 739 ++++++++++++++++++++++++++++++++++++++
>   hw/9pfs/Makefile.objs             |   3 +
>   scripts/analyse-9p-simpletrace.py |  96 +++++
>   trace-events                      |  33 ++
>   7 files changed, 907 insertions(+), 1 deletion(-)
>   create mode 100644 hw/9pfs/9p-cephfs.c
>
> diff --git a/configure b/configure
> index 0c0472a..a2627be 100755
> --- a/configure
> +++ b/configure
> @@ -275,6 +275,7 @@ trace_backends="log"
>   trace_file="trace"
>   spice=""
>   rbd=""
> +cephfs=""
>   smartcard=""
>   libusb=""
>   usb_redir=""
> @@ -1019,6 +1020,10 @@ for opt do
>     ;;
>     --enable-rbd) rbd="yes"
>     ;;
> +  --disable-cephfs) cephfs="no"
> +  ;;
> +  --enable-cephfs) cephfs="yes"
> +  ;;
>     --disable-xfsctl) xfs="no"
>     ;;
>     --enable-xfsctl) xfs="yes"
> @@ -1345,6 +1350,7 @@ disabled with --disable-FEATURE, default is enabled if available:
>     vhost-net       vhost-net acceleration support
>     spice           spice
>     rbd             rados block device (rbd)
> +  cephfs          Ceph File System
>     libiscsi        iscsi support
>     libnfs          nfs support
>     smartcard       smartcard support (libcacard)
> @@ -3087,6 +3093,28 @@ EOF
>   fi
>   
>   ##########################################
> +# cephfs probe
> +if test "$cephfs" != "no" ; then
> +  cat > $TMPC <<EOF
> +#include <stdio.h>
> +#include <cephfs/libcephfs.h>
> +int main(void) {
> +    struct ceph_mount_info *cmount;
> +    ceph_create(&cmount, NULL);
> +    return 0;
> +}
> +EOF
> +  cephfs_libs="-lcephfs -lrados"
> +  if compile_prog "" "$cephfs_libs" ; then
> +    cephfs=yes
> +  else
> +    if test "$cephfs" = "yes" ; then
> +      feature_not_found "cephfs" "Install libcephfs/ceph devel"
> +    fi
> +    cephfs=no
> +  fi
> +fi
> +##########################################
>   # libssh2 probe
>   min_libssh2_version=1.2.8
>   if test "$libssh2" != "no" ; then
> @@ -4760,6 +4788,7 @@ else
>   echo "spice support     $spice"
>   fi
>   echo "rbd support       $rbd"
> +echo "cephfs support    $cephfs"
>   echo "xfsctl support    $xfs"
>   echo "smartcard support $smartcard"
>   echo "libusb            $libusb"
> @@ -5224,6 +5253,10 @@ if test "$rbd" = "yes" ; then
>     echo "RBD_CFLAGS=$rbd_cflags" >> $config_host_mak
>     echo "RBD_LIBS=$rbd_libs" >> $config_host_mak
>   fi
> +if test "$cephfs" = "yes" ; then
> +  echo "CONFIG_CEPHFS=m" >> $config_host_mak
> +  echo "CEPHFS_LIBS=$cephfs_libs" >> $config_host_mak
> +fi
>   
>   echo "CONFIG_COROUTINE_BACKEND=$coroutine" >> $config_host_mak
>   if test "$coroutine_pool" = "yes" ; then
> diff --git a/fsdev/qemu-fsdev.c b/fsdev/qemu-fsdev.c
> index bf7f0b0..7f07a2a 100644
> --- a/fsdev/qemu-fsdev.c
> +++ b/fsdev/qemu-fsdev.c
> @@ -27,6 +27,7 @@ static FsDriverTable FsDrivers[] = {
>   #endif
>       { .name = "synth", .ops = &synth_ops},
>       { .name = "proxy", .ops = &proxy_ops},
> +    { .name = "cephfs", .ops = &cephfs_ops},
>   };
>   
>   int qemu_fsdev_add(QemuOpts *opts)
> diff --git a/fsdev/qemu-fsdev.h b/fsdev/qemu-fsdev.h
> index 9fa45bf..86a17b8 100644
> --- a/fsdev/qemu-fsdev.h
> +++ b/fsdev/qemu-fsdev.h
> @@ -22,7 +22,7 @@
>    * fstype | ops
>    * -----------------
>    *  local | local_ops
> - *  .     |
> + *  cephfs| cephfs_ops
>    *  .     |
>    *  .     |
>    *  .     |
> @@ -45,4 +45,5 @@ extern FileOperations local_ops;
>   extern FileOperations handle_ops;
>   extern FileOperations synth_ops;
>   extern FileOperations proxy_ops;
> +extern FileOperations cephfs_ops;
>   #endif
> diff --git a/hw/9pfs/9p-cephfs.c b/hw/9pfs/9p-cephfs.c
> new file mode 100644
> index 0000000..f18ec89
> --- /dev/null
> +++ b/hw/9pfs/9p-cephfs.c
> @@ -0,0 +1,739 @@
> +/*
> + * Virtio 9p cephfs callback
> + *
> + * Copyright UnitedStack, Corp. 2016
> + *
> + * Authors:
> + *    Jevon Qiao <scaleqiao@gmail.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 "qemu/iov.h"
> +#include "9p.h"
> +#include "9p-xattr.h"
> +#include "trace.h"
> +#include <cephfs/libcephfs.h>
> +#include "fsdev/qemu-fsdev.h"   /* cephfs_ops */
> +#include <arpa/inet.h>
> +#include <pwd.h>
> +#include <grp.h>
> +#include <sys/socket.h>
> +#include <sys/un.h>
> +#include "qemu/xattr.h"
> +#include "qemu/error-report.h"
> +#include <libgen.h>
> +#include <unistd.h>
> +#include <linux/fs.h>
> +#ifdef CONFIG_LINUX_MAGIC_H
> +#include <linux/magic.h>
> +#endif
> +#include <sys/ioctl.h>
> +
> +#define CEPH_VER_LEN        32
> +#define MON_NAME_LEN        32
> +#define MON_SECRET_LEN      64
> +
> +#ifndef LIBCEPHFS_VERSION
> +#define LIBCEPHFS_VERSION(maj, min, extra) ((maj << 16) + (min << 8) + extra)
> +#define LIBCEPHFS_VERSION_CODE LIBCEPHFS_VERSION(0, 0, 0)
> +#endif
> +
> +struct cephfs_data {
> +    int	major, minor, patch;
> +    char ceph_version[CEPH_VER_LEN];
> +    struct  ceph_mount_info *cmount;
> +};
> +
> +/*
> + * Helper function for cephfs_preadv and cephfs_pwritev
> + */
> +inline static ssize_t preadv_pwritev(struct ceph_mount_info *cmount, int fd,
> +                              	    const struct iovec *iov, int iov_cnt,
> +                              	    off_t offset, bool do_write)
> +{
> +    ssize_t ret = 0;
> +    size_t i = 0;
> +    size_t len = 0;
> +    void *buf, *buftmp;
> +    size_t bufoffset = 0;
> +
> +    len = iov_size(iov, iov_cnt);
> +    buf = g_new0(uint8_t, len);
> +    buftmp = buf;
> +    if (do_write) {
> +        for (i = 0; i < iov_cnt; i++) {
> +            memcpy((buftmp + bufoffset), iov[i].iov_base, iov[i].iov_len);
> +            bufoffset += iov[i].iov_len;
> +        }
> +        ret = ceph_write(cmount, fd, buf, len, offset);
> +        if (ret <= 0) {
> +           errno = -ret;
> +           ret = -1;
> +        }
> +    } else {
> +        ret = ceph_read(cmount, fd, buf, len, offset);
> +        if (ret <= 0) {
> +            errno = -ret;
> +            ret = -1;
> +        } else {
> +            for (i = 0; i < iov_cnt; i++) {
> +                memcpy(iov[i].iov_base, (buftmp + bufoffset), iov[i].iov_len);
> +                bufoffset += iov[i].iov_len;
> +            }
> +        }
> +    }
> +
> +    free(buf);
> +    return ret;
> +}
> +
> +static int cephfs_update_file_cred(struct ceph_mount_info *cmount,
> +				   const char *name, FsCred *credp)
> +{
> +    int fd, ret;
> +    fd = ceph_open(cmount, name, O_NONBLOCK | O_NOFOLLOW, credp->fc_mode);
> +    if (fd < 0) {
> +        return fd;
> +    }
> +    ret = ceph_fchown(cmount, fd, credp->fc_uid, credp->fc_gid);
> +    if (ret < 0) {
> +        goto err_out;
> +    }
> +    ret = ceph_fchmod(cmount, fd, credp->fc_mode & 07777);
> +err_out:
> +    close(fd);
> +    return ret;
> +}
> +
> +static int cephfs_lstat(FsContext *fs_ctx, V9fsPath *fs_path,
> +                        struct stat *stbuf)
> +{
> +    int ret;
> +    char *path = fs_path->data;
> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    ret = ceph_lstat(cfsdata->cmount, path, stbuf);
> +    trace_cephfs_lstat_return(path, stbuf->st_mode, stbuf->st_uid, stbuf->st_gid, stbuf->st_size, ret);
> +    if (ret){
> +        errno = -ret;
> +        ret = -1;
> +    }
> +    return ret;
> +}
> +
> +static ssize_t cephfs_readlink(FsContext *fs_ctx, V9fsPath *fs_path,
> +                               char *buf, size_t bufsz)
> +{
> +    int ret;
> +    char *path = fs_path->data;
> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    ret = ceph_readlink(cfsdata->cmount, path, buf, bufsz);
> +    trace_cephfs_readlink_return(path, ret);
> +    return ret;
> +}
> +
> +static int cephfs_close(FsContext *ctx, V9fsFidOpenState *fs)
> +{
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    return ceph_close(cfsdata->cmount, fs->fd);
> +}
> +
> +static int cephfs_closedir(FsContext *ctx, V9fsFidOpenState *fs)
> +{
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    return ceph_closedir(cfsdata->cmount, (struct ceph_dir_result *)fs->dir);
> +}
> +
> +static int cephfs_open(FsContext *ctx, V9fsPath *fs_path,
> +                       int flags, V9fsFidOpenState *fs)
> +{
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    fs->fd = ceph_open(cfsdata->cmount, fs_path->data, flags, 0777);
> +    trace_cephfs_open_return(fs_path->data, flags, 0777, fs->fd);
> +    return fs->fd;
> +}
> +
> +static int cephfs_opendir(FsContext *ctx,
> +                          V9fsPath *fs_path, V9fsFidOpenState *fs)
> +{
> +    int ret;
> +    struct ceph_dir_result *result;
> +    struct cephfs_data *cfsdata = ctx->private;
> +    char *path = fs_path->data;
> +
> +    ret = ceph_opendir(cfsdata->cmount, path, &result);
> +    trace_cephfs_opendir_return(path, ret);
> +    if (ret) {
> +        fprintf(stderr, "ceph_opendir=%d\n", ret);
> +        return ret;
> +    }
> +    fs->dir = (DIR *)result;
> +    if (!fs->dir) {
> +        fprintf(stderr, "ceph_opendir return NULL for ceph_dir_result\n");
> +        return -1;
> +    }
> +    return 0;
> +}
> +
> +static void cephfs_rewinddir(FsContext *ctx, V9fsFidOpenState *fs)
> +{
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    trace_cephfs_rewinddir(fs->dir);
> +    return ceph_rewinddir(cfsdata->cmount, (struct ceph_dir_result *)fs->dir);
> +}
> +
> +static off_t cephfs_telldir(FsContext *ctx, V9fsFidOpenState *fs)
> +{
> +    int ret;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    trace_cephfs_telldir(fs->dir);
> +    ret = ceph_telldir(cfsdata->cmount, (struct ceph_dir_result *)fs->dir);
> +    return ret;
> +}
> +
> +static int cephfs_readdir_r(FsContext *ctx, V9fsFidOpenState *fs,
> +                            struct dirent *entry,
> +                            struct dirent **result)
> +{
> +    int ret;
> +    struct dirent *tmpent;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    tmpent = entry;
> +    ret = ceph_readdir_r(cfsdata->cmount, (struct ceph_dir_result *)fs->dir,
> +		    	entry);
> +    trace_cephfs_readdir_r_return(tmpent, entry, ret);
> +    if (ret > 0 && entry != NULL)
> +    {
> +        *result = entry;
> +    } else if (!ret)
> +    {
> +        *result = NULL;
> +        entry = tmpent;
> +    }
> +
> +    return ret;
> +}
> +
> +static void cephfs_seekdir(FsContext *ctx, V9fsFidOpenState *fs, off_t off)
> +{
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    trace_cephfs_seekdir(fs->dir, off);
> +    return ceph_seekdir(cfsdata->cmount, (struct ceph_dir_result*)fs->dir, off);
> +}
> +
> +static ssize_t cephfs_preadv(FsContext *ctx, V9fsFidOpenState *fs,
> +                             const struct iovec *iov,
> +                             int iovcnt, off_t offset)
> +{
> +    ssize_t ret = 0;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    trace_cephfs_preadv(iovcnt, iov_size(iov, iovcnt));
> +#if defined(LIBCEPHFS_VERSION) && LIBCEPHFS_VERSION_CODE >= LIBCEPHFS_VERSION(9, 0, 3)
> +    ret = ceph_preadv(cfsdata->cmount, fs->fd, iov, iovcnt, offset);
> +#else
> +    if (iovcnt > 1) {
> +	ret = preadv_pwritev(cfsdata->cmount, fs->fd, iov, iovcnt, offset, 0);
> +    } else if (iovcnt > 0) {
> +	ret = ceph_read(cfsdata->cmount, fs->fd, iov[0].iov_base,
> +			iov[0].iov_len, offset);
> +    }
> +#endif
> +    trace_cephfs_preadv_return(iovcnt, iov_size(iov, iovcnt), ret);
> +
> +    return ret;
> +}
> +
> +static ssize_t cephfs_pwritev(FsContext *ctx, V9fsFidOpenState *fs,
> +                              const struct iovec *iov,
> +                              int iovcnt, off_t offset)
> +{
> +    ssize_t ret = 0;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    trace_cephfs_pwritev(iovcnt, iov_size(iov, iovcnt), offset);
> +#if defined(LIBCEPHFS_VERSION) && LIBCEPHFS_VERSION_CODE >= LIBCEPHFS_VERSION(9, 0, 3)
> +    ret = ceph_pwritev(cfsdata->cmount, fs->fd, iov, iovcnt, offset);
> +#else
> +    if (iovcnt > 1) {
> +	ret = preadv_pwritev(cfsdata->cmount, fs->fd, iov, iovcnt, offset, 1);
> +    } else if (iovcnt > 0) {
> +	ret = ceph_write(cfsdata->cmount, fs->fd, iov[0].iov_base,
> +			iov[0].iov_len, offset);
> +    }
> +#endif
> +    trace_cephfs_pwritev_return(iovcnt, iov_size(iov, iovcnt), offset, ret);
> +
> +#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 cephfs_chmod(FsContext *fs_ctx, V9fsPath *fs_path, FsCred *credp)
> +{
> +    int  ret = -1;
> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    ret = ceph_chmod(cfsdata->cmount, fs_path->data, credp->fc_mode);
> +    trace_cephfs_chmod_return(fs_path->data, credp->fc_mode, ret);
> +    return ret;
> +}
> +
> +static int cephfs_mknod(FsContext *fs_ctx, V9fsPath *dir_path,
> +                       const char *name, FsCred *credp)
> +{
> +    int ret;
> +    V9fsString fullname;
> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    v9fs_string_init(&fullname);
> +    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
> +    ret = ceph_mknod(cfsdata->cmount, fullname.data, credp->fc_mode,
> +		    credp->fc_rdev);
> +    trace_cephfs_mknod_return(fullname.data, credp->fc_mode, credp->fc_rdev, ret);
> +
> +    v9fs_string_free(&fullname);
> +    return ret;
> +}
> +
> +static int cephfs_mkdir(FsContext *fs_ctx, V9fsPath *dir_path,
> +                       const char *name, FsCred *credp)
> +{
> +    int ret;
> +    V9fsString fullname;
> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    v9fs_string_init(&fullname);
> +    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
> +    ret = ceph_mkdir(cfsdata->cmount, fullname.data, credp->fc_mode);
> +    trace_cephfs_mkdir_return(fullname.data, credp->fc_mode, ret);
> +
> +    v9fs_string_free(&fullname);
> +    return ret;
> +}
> +
> +static int cephfs_fstat(FsContext *fs_ctx, int fid_type,
> +                        V9fsFidOpenState *fs, struct stat *stbuf)
> +{
> +    int fd = -1;
> +    int ret;
> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    if (fid_type == P9_FID_DIR) {
> +        fd = dirfd(fs->dir);
> +    } else {
> +        fd = fs->fd;
> +    }
> +    ret = ceph_fstat(cfsdata->cmount, fd, stbuf);
> +    trace_cephfs_fstat_return(fid_type, fd, stbuf->st_uid, stbuf->st_gid, stbuf->st_size, ret);
> +
> +    return ret;
> +}
> +
> +static int cephfs_open2(FsContext *fs_ctx, V9fsPath *dir_path, const char *name,
> +                        int flags, FsCred *credp, V9fsFidOpenState *fs)
> +{
> +    int fd = -1, ret = -1;
> +    V9fsString fullname;
> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    v9fs_string_init(&fullname);
> +    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
> +    fd = ceph_open(cfsdata->cmount, fullname.data, flags, credp->fc_mode);
> +    trace_cephfs_open2_return(fullname.data, flags, credp->fc_mode);
> +    if (fd >= 0) {
> +        /* After creating the file, need to set the cred */
> +        ret = cephfs_update_file_cred(cfsdata->cmount, name, credp);
> +        if (ret < 0) {
> +            ceph_close(cfsdata->cmount, fd);
> +            errno = -ret;
> +            fd = ret;
> +        } else {
> +            fs->fd = fd;
> +        }
> +    } else {
> +       errno = -fd;
> +    }
> +
> +    v9fs_string_free(&fullname);
> +    return fd;
> +}
> +
> +static int cephfs_symlink(FsContext *fs_ctx, const char *oldpath,
> +                          V9fsPath *dir_path, const char *name, FsCred *credp)
> +{
> +    int ret = -1;
> +    V9fsString fullname;
> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    v9fs_string_init(&fullname);
> +    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
> +    ret = ceph_symlink(cfsdata->cmount, oldpath, fullname.data);
> +    trace_cephfs_symlink_return(oldpath, fullname.data, ret);
> +
> +    v9fs_string_free(&fullname);
> +    return ret;
> +}
> +
> +static int cephfs_link(FsContext *ctx, V9fsPath *oldpath,
> +                       V9fsPath *dirpath, const char *name)
> +{
> +    int ret = -1;
> +    V9fsString newpath;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    v9fs_string_init(&newpath);
> +    v9fs_string_sprintf(&newpath, "%s/%s", dirpath->data, name);
> +    ret = ceph_link(cfsdata->cmount, oldpath->data, newpath.data);
> +    trace_cephfs_link_return(oldpath->data, newpath.data, ret);
> +
> +    v9fs_string_free(&newpath);
> +    return ret;
> +}
> +
> +static int cephfs_truncate(FsContext *ctx, V9fsPath *fs_path, off_t size)
> +{
> +    int ret = -1;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    ret = ceph_truncate(cfsdata->cmount, fs_path->data, size);
> +    trace_cephfs_truncate_return(fs_path->data, size, ret);
> +
> +    return ret;
> +}
> +
> +static int cephfs_rename(FsContext *ctx, const char *oldpath,
> +                         const char *newpath)
> +{
> +    int ret = -1;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    ret = ceph_rename(cfsdata->cmount, oldpath, newpath);
> +    trace_cephfs_rename_return(oldpath, newpath, ret);
> +
> +    return ret;
> +}
> +
> +static int cephfs_chown(FsContext *fs_ctx, V9fsPath *fs_path, FsCred *credp)
> +{
> +    int ret = -1;
> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    ret = ceph_chown(cfsdata->cmount, fs_path->data, credp->fc_uid,
> +		    credp->fc_gid);
> +    trace_cephfs_chown_return(fs_path->data, credp->fc_uid, credp->fc_gid, ret);
> +
> +    return ret;
> +}
> +
> +static int cephfs_utimensat(FsContext *ctx, V9fsPath *fs_path,
> +                            const struct timespec *buf)
> +{
> +    int ret = -1;
> +
> +#ifdef CONFIG_UTIMENSAT
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    ret = ceph_utime(cfsdata->cmount, fs_path->data, (struct utimbuf *)buf);
> +    trace_cephfs_utimensat_return(fs_path->data, ret);
> +#else
> +    ret = -1;
> +    errno = ENOSYS;
> +#endif
> +
> +    return ret;
> +}
> +
> +static int cephfs_remove(FsContext *ctx, const char *path)
> +{
> +    errno = EOPNOTSUPP;
> +    return -1;
> +}
> +
> +static int cephfs_fsync(FsContext *ctx, int fid_type,
> +                        V9fsFidOpenState *fs, int datasync)
> +{
> +    int ret = -1, fd = -1;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    if (fid_type == P9_FID_DIR) {
> +        fd = dirfd(fs->dir);
> +    } else {
> +        fd = fs->fd;
> +    }
> +    ret = ceph_fsync(cfsdata->cmount, fd, datasync);
> +    trace_cephfs_fsync_return(fd, datasync, ret);
> +
> +    return ret;
> +}
> +
> +static int cephfs_statfs(FsContext *ctx, V9fsPath *fs_path,
> +                         struct statfs *stbuf)
> +{
> +    int ret;
> +    char *path = fs_path->data;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    ret = ceph_statfs(cfsdata->cmount, path, (struct statvfs*)stbuf);
> +    if (ret) {
> +        fprintf(stderr, "ceph_statfs=%d\n", ret);
> +    }
> +
> +    return ret;
> +}
> +
> +/*
> + * Get the extended attribute of normal file, if the path refer to a symbolic
> + * link, just return the extended attributes of the syslink rather than the
> + * attributes of the link itself.
> + */
> +static ssize_t cephfs_lgetxattr(FsContext *ctx, V9fsPath *fs_path,
> +                                const char *name, void *value, size_t size)
> +{
> +    int ret;
> +    char *path = fs_path->data;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    ret = ceph_lgetxattr(cfsdata->cmount, path, name, value, size);
> +    trace_cephfs_lgetxattr_return(path, name, ret);
> +
> +    return ret;
> +}
> +
> +static ssize_t cephfs_llistxattr(FsContext *ctx, V9fsPath *fs_path,
> +                                 void *value, size_t size)
> +{
> +    int ret = -1;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    ret = ceph_llistxattr(cfsdata->cmount, fs_path->data, value, size);
> +    trace_cephfs_llistxattr_return(fs_path->data, ret);
> +
> +    return ret;
> +}
> +
> +static int cephfs_lsetxattr(FsContext *ctx, V9fsPath *fs_path, const char *name,
> +                            void *value, size_t size, int flags)
> +{
> +    int ret = -1;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    ret = ceph_lsetxattr(cfsdata->cmount, fs_path->data, name, value, size,
> +	flags);
> +    trace_cephfs_lsetxattr_return(fs_path->data, name, flags, ret);
> +
> +    return ret;
> +}
> +
> +static int cephfs_lremovexattr(FsContext *ctx, V9fsPath *fs_path,
> +                               const char *name)
> +{
> +    int ret = -1;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    ret = ceph_lremovexattr(cfsdata->cmount, fs_path->data, name);
> +    trace_cephfs_lremovexattr_return(fs_path->data, name, ret);
> +
> +    return ret;
> +}
> +
> +static int cephfs_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 {
> +        /* if the path does not start from '/' */
> +        v9fs_string_sprintf((V9fsString *)target, "%s", name);
> +    }
> +
> +    /* Bump the size for including terminating NULL */
> +    target->size++;
> +    return 0;
> +}
> +
> +static int cephfs_renameat(FsContext *ctx, V9fsPath *olddir,
> +                           const char *old_name, V9fsPath *newdir,
> +                           const char *new_name)
> +{
> +    int ret = -1;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    ret = ceph_rename(cfsdata->cmount, old_name, new_name);
> +    trace_cephfs_renameat_return(old_name, new_name, ret);
> +
> +    return ret;
> +}
> +
> +static int cephfs_unlinkat(FsContext *ctx, V9fsPath *dir,
> +                           const char *name, int flags)
> +{
> +    int ret = 0;
> +    char *path = dir->data;
> +    struct stat fstat;
> +    V9fsString fullname;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    v9fs_string_init(&fullname);
> +    v9fs_string_sprintf(&fullname, "%s/%s", dir->data, name);
> +    path = fullname.data;
> +    /* determine which kind of file is being destroyed */
> +    ret = ceph_lstat(cfsdata->cmount, path, &fstat);
> +    if (!ret) {
> +        switch (fstat.st_mode & S_IFMT) {
> +        case S_IFDIR:
> +            ret = ceph_rmdir(cfsdata->cmount, path);
> +            break;
> +
> +        case S_IFBLK:
> +        case S_IFCHR:
> +        case S_IFIFO:
> +        case S_IFLNK:
> +        case S_IFREG:
> +        case S_IFSOCK:
> +            ret = ceph_unlink(cfsdata->cmount, path);
> +            break;
> +
> +        default:
> +            fprintf(stderr, "ceph_lstat unknown stmode\n");
> +            break;
> +        }
> +    } else {
> +        errno = -ret;
> +        ret = -1;
> +    }
> +    trace_cephfs_unlinkat_return(path, fstat.st_mode, ret);
> +
> +    v9fs_string_free(&fullname);
> +    return ret;
> +}
> +
> +/*
> + * Do two things in the init function:
> + * 1) Create a mount handle used by all cephfs interfaces.
> + * 2) Invoke ceph_mount() to initialize a link between the client and
> + *    ceph monitor
> + */
> +static int cephfs_init(FsContext *ctx)
> +{
> +    int ret;
> +    const char *ver = NULL;
> +    struct cephfs_data *data = g_malloc(sizeof(struct cephfs_data));
> +
> +    if (data == NULL) {
> +	errno = ENOMEM;
> +	return -1;
> +    }
> +    trace_cephfs_init(ctx->fs_root);
> +    memset(data, 0, sizeof(struct cephfs_data));
> +    ret = ceph_create(&data->cmount, NULL);
> +    if (ret) {
> +        fprintf(stderr, "ceph_create=%d\n", ret);
> +        goto err_out;
> +    }
> +
> +    ret = ceph_conf_read_file(data->cmount, NULL);
> +    if (ret) {
> +        fprintf(stderr, "ceph_conf_read_file=%d\n", ret);
> +        goto err_out;
> +    }
> +
> +    ret = ceph_mount(data->cmount, ctx->fs_root);
> +    if (ret) {
> +        fprintf(stderr, "ceph_mount=%d\n", ret);
> +        goto err_out;
> +    } else {
> +        ctx->private = data;
> +	/* CephFS does not support FS_IOC_GETVERSIO */
> +	ctx->exops.get_st_gen = NULL;
> +        goto out;
> +    }
> +
> +    ver = ceph_version(&data->major, &data->minor, &data->patch);
> +    memcpy(data->ceph_version, ver, strlen(ver) + 1);
> +
> +err_out:
> +    g_free(data);
> +out:
> +    return ret;
> +}
> +
> +static int cephfs_parse_opts(QemuOpts *opts, struct FsDriverEntry *fse)
> +{
> +    const char *sec_model = qemu_opt_get(opts, "security_model");
> +    const char *path = qemu_opt_get(opts, "path");
> +
> +    if (!sec_model) {
> +        fprintf(stderr, "Invalid argument security_model specified with "
> +		"cephfs fsdriver\n");
> +        return -1;
> +    }
> +
> +    if (!path) {
> +        fprintf(stderr, "fsdev: No path specified.\n");
> +        return -1;
> +    }
> +
> +    fse->path = g_strdup(path);
> +    return 0;
> +}
> +
> +FileOperations cephfs_ops = {
> +    .parse_opts   = cephfs_parse_opts,
> +    .init         = cephfs_init,
> +    .lstat        = cephfs_lstat,
> +    .readlink     = cephfs_readlink,
> +    .close        = cephfs_close,
> +    .closedir     = cephfs_closedir,
> +    .open         = cephfs_open,
> +    .opendir      = cephfs_opendir,
> +    .rewinddir    = cephfs_rewinddir,
> +    .telldir      = cephfs_telldir,
> +    .readdir_r    = cephfs_readdir_r,
> +    .seekdir      = cephfs_seekdir,
> +    .preadv       = cephfs_preadv,
> +    .pwritev      = cephfs_pwritev,
> +    .chmod        = cephfs_chmod,
> +    .mknod        = cephfs_mknod,
> +    .mkdir        = cephfs_mkdir,
> +    .fstat        = cephfs_fstat,
> +    .open2        = cephfs_open2,
> +    .symlink      = cephfs_symlink,
> +    .link         = cephfs_link,
> +    .truncate     = cephfs_truncate,
> +    .rename       = cephfs_rename,
> +    .chown        = cephfs_chown,
> +    .utimensat    = cephfs_utimensat,
> +    .remove       = cephfs_remove,
> +    .fsync        = cephfs_fsync,
> +    .statfs       = cephfs_statfs,
> +    .lgetxattr    = cephfs_lgetxattr,
> +    .llistxattr   = cephfs_llistxattr,
> +    .lsetxattr    = cephfs_lsetxattr,
> +    .lremovexattr = cephfs_lremovexattr,
> +    .name_to_path = cephfs_name_to_path,
> +    .renameat     = cephfs_renameat,
> +    .unlinkat     = cephfs_unlinkat,
> +};
> diff --git a/hw/9pfs/Makefile.objs b/hw/9pfs/Makefile.objs
> index da0ae0c..a77a6f4 100644
> --- a/hw/9pfs/Makefile.objs
> +++ b/hw/9pfs/Makefile.objs
> @@ -5,5 +5,8 @@ common-obj-y += coth.o cofs.o codir.o cofile.o
>   common-obj-y += coxattr.o 9p-synth.o
>   common-obj-$(CONFIG_OPEN_BY_HANDLE) +=  9p-handle.o
>   common-obj-y += 9p-proxy.o
> +common-obj-y += 9p-cephfs.o
>   
>   obj-y += virtio-9p-device.o
> +
> +9p-cephfs.o-libs := $(CEPHFS_LIBS)
> diff --git a/scripts/analyse-9p-simpletrace.py b/scripts/analyse-9p-simpletrace.py
> index 3c3dee4..fe0a496 100755
> --- a/scripts/analyse-9p-simpletrace.py
> +++ b/scripts/analyse-9p-simpletrace.py
> @@ -210,4 +210,100 @@ class VirtFSRequestTracker(simpletrace.Analyzer):
>           def v9fs_readlink_return(self, tag, id, target):
>                   print "RREADLINK (tag =", tag, ", target =", target, ")"
>   
> +	def cephfs_lstat_return(self, path, stmode, stuid, stgid, stsize, ret):
> +		print "RCEPHFSLSTAT (path =", path, ", stmode =", stmode, ", stuid =", stuid, ", stgid =", stgid, ", stsize =", stsize, ", ret =", ret, ")"
> +
> +	def cephfs_readlink_return(self, path, ret):
> +		print "RCEPHFSREADLINK (path =", path, ", ret =", ret, ")"
> +
> +	def cephfs_open_return(self, path, flags, mode, fd):
> +		print "RCEPHFSOPEN (path =", path, ", flags =", flags, ", mode =", mode, ", fd =", fd, ")"
> +
> +	def cephfs_opendir_return(self, path, ret):
> +		print "RCEPHFSOPENDIR (path =", path, ", ret =", ret, ")"
> +
> +	def cephfs_rewinddir(self, dir):
> +		print "TCEPHFSREWINDDIR (dir =", dir, ")"
> +
> +	def cephfs_telldir(self, dir):
> +		print "TCEPHFSTELLDIR (dir =", dir, ")"
> +
> +	def cephfs_readdir_r_return(self, tmpent, entry, ret):
> +		print "RCEPHFSREADDIRR (tmpent =", tmpent, ", entry =", entry, ", ret =", ret, ")"
> +
> +	def cephfs_seekdir(self, dir, off):
> +		print "TCEPHFSSEEKDIR (dir =", dir, ", off =", off, ")"
> +
> +	def cephfs_preadv(self, iovcnt, len):
> +		print "TCEPHFSPREADV (iovcnt=", iovcnt, ", len =", len, ")"
> +
> +	def cephfs_preadv_return(self, iovcnt, len, ret):
> +		print "RCEPHFSPREADV (iovcnt=", iovcnt, ", len =", len, ", ret = ", ret, ")"
> +
> +	def cephfs_pwritev(self, iovcnt, len, offset):
> +		print "TCEPHFSPWRITEV (iovcnt=", iovcnt, ", len =", len, ", offset =", offset, ")"
> +
> +	def cephfs_pwritev_return(self, iovcnt, len, offset, ret):
> +		print "RCEPHFSPWRITEV (iovcnt=", iovcnt, ", len =", len, ", offset =", offset, ", ret = ", ret, ")"
> +
> +	def cephfs_chmod(self, path, fcmode):
> +		print "TCEPHFSCHMOD (path =", path, ", fcmode =", fcmode, ")"
> +
> +	def cephfs_chmod_return(self, path, fcmode, ret):
> +		print "RCEPHFSCHMOD (path =", path, ", fcmode =", fcmode, ", ret =", ret, ")"
> +
> +	def cephfs_mknod_return(self, path, fcmode, fcrdev, ret):
> +		print "RCEPHFSMKNOD (path =", path, ", fcmode =", fcmode, ", fcrdev =", fcrdev, ", ret =", ret, ")"
> +
> +	def cephfs_mkdir_return(self, path, fcmode, ret):
> +		print "RCEPHFSMKDIR (path =", path, ", fcmode =", fcmode, ", ret =", ret, ")"
> +
> +	def cephfs_fstat_return(self, fidtype, fd, stuid, stgid, stsize, ret):
> +		print "RCEPHFSFSTAT (fidtype =", fidtype, ", fd =", fd, ", stuid =", stuid, ", stgid =", stgid, ", stsize =", stsize, ", ret =", ret, ")"
> +
> +	def cephfs_open2_return(self, path, flags, fcmode):
> +		print "RCEPHFSOPEN2 (path =", path, ", flags =", flags, "fcmode =", fcmode, ")"
> +
> +	def cephfs_symlink_return(self, oldpath, path, ret):
> +		print "RCEPHFSSYMLINK (oldpath =", oldpath, ", path =", path, ", ret =", ret, ")"
> +
> +	def cephfs_link_return(self, oldpath, path, ret):
> +		print "RCEPHFSLINK (oldpath =", oldpath, ", path =", path, ", ret =", ret, ")"
> +
> +	def cephfs_truncate_return(self, path, size, ret):
> +		print "RCEPHFSTRUNCATE (path =", path, ", size =", size, ", ret =", ret, ")"
> +
> +	def cephfs_rename_return(self, oldpath, newpath, ret):
> +		print "RCEPHFSRENAME (oldpath =", oldpath, ", newpath =", newpath, ", ret =", ret, ")"
> +
> +	def cephfs_chown_return(self, path, fcuid, fcgid, ret):
> +		print "RCEPHFSCHOWN (path =", path, ", fcuid =", fcuid, ", fcgid =", fcgid, ", ret =", ret, ")"
> +
> +	def cephfs_utimensat_return(self, path, ret):
> +		print "RCEPHFSUTIMENSAT (path =", path, ", ret =", ret, ")"
> +
> +	def cephfs_fsync_return(self, fd, datasync, ret):
> +		print "RCEPHFSFSYNC (fd =", fd, ", datasync =", datasync, ", ret =", ret, ")"
> +
> +	def cephfs_lgetxattr_return(self, path, name, ret):
> +		print "RCEPHFSLGETXATTR (path =", path, ", name =", name, ", ret =", ret, ")"
> +
> +	def cephfs_llistxattr_return(self, path, ret):
> +		print "RCEPHFSLLISTXATTR (path =", path, ", ret =", ret, ")"
> +
> +	def cephfs_lsetxattr_return(self, path, name, flags, ret):
> +		print "RCEPHFSLSETXATTR (path =", path, ", name =", name, ", flags =", flags, ", ret =", ret, ")"
> +
> +	def cephfs_lremovexattr_return(self, path, name, ret):
> +		print "RCEPHFSLREMOVEXATTR (path =", path, ", name =", name, ", ret =", ret, ")"
> +
> +	def cephfs_renameat_return(self, oldname, newname, ret):
> +		print "RCEPHFSRENAMEAT (oldname =", oldname, ", newname =", newname, ", ret =", ret, ")"
> +
> +	def cephfs_unlinkat_return(self, path, stmode, ret):
> +		print "RCEPHFSUNLINKAT (path =", path, ", stmode =", stmode, ", ret =", ret, ")"
> +
> +	def cephfs_init(self, path):
> +		print "RCEPHFSINIT (path =", path, ")"
> +
>   simpletrace.run(VirtFSRequestTracker())
> diff --git a/trace-events b/trace-events
> index 6fba6cc..11879d2 100644
> --- a/trace-events
> +++ b/trace-events
> @@ -1118,6 +1118,39 @@ v9fs_xattrcreate(uint16_t tag, uint8_t id, int32_t fid, char* name, int64_t size
>   v9fs_readlink(uint16_t tag, uint8_t id, int32_t fid) "tag %d id %d fid %d"
>   v9fs_readlink_return(uint16_t tag, uint8_t id, char* target) "tag %d id %d name %s"
>   
> +# hw/9pfs/9p-cephfs.c
> +cephfs_lstat_return(char *path, int stmode, int stuid, int stgid, int stsize, int ret) "path %s stmode %d stuid %d stgid %d stsize %d ret %d"
> +cephfs_readlink_return(char *path, int ret) "path %s ret %d"
> +cephfs_open_return(char *path, int flags, int mode, int fd) "path %s flags %d mode %d fd %d"
> +cephfs_opendir_return(char *path, int ret) "path %s ret %d"
> +cephfs_rewinddir(void *dir) "dir %p"
> +cephfs_telldir(void *dir) "dir %p"
> +cephfs_readdir_r_return(void *tmpent, void *entry, int ret) "tmpent %p entry %p ret %d"
> +cephfs_seekdir(void *dir, int off) "dir %p off %d"
> +cephfs_preadv(int iovcnt, int len) "iovcnt %d len %d"
> +cephfs_preadv_return(int iovcnt, int len, long ret) "iovcnt %d len %d ret %l"
> +cephfs_pwritev(int iovcnt, int len, int offset) "iovcnt %d len %d offset %d"
> +cephfs_pwritev_return(int iovcnt, int len, int offset, long ret) "iovcnt %d len %d offset %d ret %l"cephfs_chmod(char *path, int fcmode) "path %s fcmode %d"
> +cephfs_chmod_return(char *path, int fcmode, int ret) "path %s fcmode %d ret %d"
> +cephfs_mknod_return(char *path, int fcmode, uint32_t fcrdev, int ret) "path %s fcmode %d fcrdev %u ret %d"
> +cephfs_mkdir_return(char *path, int fcmode, int ret) " path %s fcmode %d ret %d"
> +cephfs_fstat_return(int fidtype, int fd, int stuid, int stgid, int stsize, int ret) "fidtype %d fd %d stuid %d stgid %d stsize %d ret %d"
> +cephfs_open2_return(char *path, int flags, int fcmode) "path %s flags %d fcmode %d"
> +cephfs_symlink_return(const char *oldpath, char *path, int ret) "oldpath %s path %s ret %d"
> +cephfs_link_return(char *oldpath, char *path, int ret) "oldpath %s path %s ret %d"
> +cephfs_truncate_return(char *path, int size, int ret) "path %s size %d ret %d"
> +cephfs_rename_return(const char *oldpath, const char *newpath, int ret) "oldpath %s newpath %s ret %d"
> +cephfs_chown_return(char *path, int fcuid, int fcgid, int ret) "path %s fcuid %d fcgid %d ret %d"
> +cephfs_utimensat_return(char *path, int ret) "path %s ret %d"
> +cephfs_fsync_return(int fd, int datasync, int ret) "fd %d datasync %d ret %d"
> +cephfs_lgetxattr_return(char *path, const char *name, int ret) "path %s name %s ret %d"
> +cephfs_llistxattr_return(char *path, int ret) "path %s ret %d"
> +cephfs_lsetxattr_return(char *path, const char *name, int flags, int ret) "path %s name %s flags %d ret %d"
> +cephfs_lremovexattr_return(char *path, const char *name, int ret) "path %s name %s ret %d"
> +cephfs_renameat_return(const char *oldname, const char *newname, int ret) "oldname %s newname %s ret %d"
> +cephfs_unlinkat_return(char *path, int stmode, int ret) "path %s stmode %d ret %d"
> +cephfs_init(char *path) "path %s"
> +
>   # target-sparc/mmu_helper.c
>   mmu_helper_dfault(uint64_t address, uint64_t context, int mmu_idx, uint32_t tl) "DFAULT at %"PRIx64" context %"PRIx64" mmu_idx=%d tl=%d"
>   mmu_helper_dprot(uint64_t address, uint64_t context, int mmu_idx, uint32_t tl) "DPROT at %"PRIx64" context %"PRIx64" mmu_idx=%d tl=%d"

--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Greg Kurz March 9, 2016, 9:59 a.m. UTC | #2
On Tue, 8 Mar 2016 08:51:23 +0800
Jevon Qiao <scaleqiao@gmail.com> wrote:

> Any further comment on this patch?
> 

First comment is:

$ git show | ./scripts/checkpatch.pl - | grep ^total
total: 44 errors, 7 warnings, 975 lines checked

I'm now reading the patch and will come back with other comments.

> Thanks,
> Jevon
> On 2/3/16 23:41, Jevon Qiao wrote:
> > Ceph as a promising unified distributed storage system is widely used in the
> > world of OpenStack. OpenStack users deploying Ceph for block (Cinder) and
> > object (S3/Swift) are unsurprisingly looking at Manila and CephFS to round out
> > a unified storage solution. Since the typical hypervisor people are using is
> > Qemu/KVM, it is necessary to provide a high performance, easy to use, file
> > system service in it. VirtFS aims to offers paravirtualized system services and
> > simple passthrough for directories from host to guest, which currently only
> > support local file system, this patch wants to add CephFS support in VirtFS.
> >
> > Signed-off-by: Jevon Qiao <scaleqiao@gmail.com>
> > ---
> >   configure                         |  33 ++
> >   fsdev/qemu-fsdev.c                |   1 +
> >   fsdev/qemu-fsdev.h                |   3 +-
> >   hw/9pfs/9p-cephfs.c               | 739 ++++++++++++++++++++++++++++++++++++++
> >   hw/9pfs/Makefile.objs             |   3 +
> >   scripts/analyse-9p-simpletrace.py |  96 +++++
> >   trace-events                      |  33 ++
> >   7 files changed, 907 insertions(+), 1 deletion(-)
> >   create mode 100644 hw/9pfs/9p-cephfs.c
> >
> > diff --git a/configure b/configure
> > index 0c0472a..a2627be 100755
> > --- a/configure
> > +++ b/configure
> > @@ -275,6 +275,7 @@ trace_backends="log"
> >   trace_file="trace"
> >   spice=""
> >   rbd=""
> > +cephfs=""
> >   smartcard=""
> >   libusb=""
> >   usb_redir=""
> > @@ -1019,6 +1020,10 @@ for opt do
> >     ;;
> >     --enable-rbd) rbd="yes"
> >     ;;
> > +  --disable-cephfs) cephfs="no"
> > +  ;;
> > +  --enable-cephfs) cephfs="yes"
> > +  ;;
> >     --disable-xfsctl) xfs="no"
> >     ;;
> >     --enable-xfsctl) xfs="yes"
> > @@ -1345,6 +1350,7 @@ disabled with --disable-FEATURE, default is enabled if available:
> >     vhost-net       vhost-net acceleration support
> >     spice           spice
> >     rbd             rados block device (rbd)
> > +  cephfs          Ceph File System
> >     libiscsi        iscsi support
> >     libnfs          nfs support
> >     smartcard       smartcard support (libcacard)
> > @@ -3087,6 +3093,28 @@ EOF
> >   fi
> >   
> >   ##########################################
> > +# cephfs probe
> > +if test "$cephfs" != "no" ; then
> > +  cat > $TMPC <<EOF
> > +#include <stdio.h>
> > +#include <cephfs/libcephfs.h>
> > +int main(void) {
> > +    struct ceph_mount_info *cmount;
> > +    ceph_create(&cmount, NULL);
> > +    return 0;
> > +}
> > +EOF
> > +  cephfs_libs="-lcephfs -lrados"
> > +  if compile_prog "" "$cephfs_libs" ; then
> > +    cephfs=yes
> > +  else
> > +    if test "$cephfs" = "yes" ; then
> > +      feature_not_found "cephfs" "Install libcephfs/ceph devel"
> > +    fi
> > +    cephfs=no
> > +  fi
> > +fi
> > +##########################################
> >   # libssh2 probe
> >   min_libssh2_version=1.2.8
> >   if test "$libssh2" != "no" ; then
> > @@ -4760,6 +4788,7 @@ else
> >   echo "spice support     $spice"
> >   fi
> >   echo "rbd support       $rbd"
> > +echo "cephfs support    $cephfs"
> >   echo "xfsctl support    $xfs"
> >   echo "smartcard support $smartcard"
> >   echo "libusb            $libusb"
> > @@ -5224,6 +5253,10 @@ if test "$rbd" = "yes" ; then
> >     echo "RBD_CFLAGS=$rbd_cflags" >> $config_host_mak
> >     echo "RBD_LIBS=$rbd_libs" >> $config_host_mak
> >   fi
> > +if test "$cephfs" = "yes" ; then
> > +  echo "CONFIG_CEPHFS=m" >> $config_host_mak
> > +  echo "CEPHFS_LIBS=$cephfs_libs" >> $config_host_mak
> > +fi
> >   
> >   echo "CONFIG_COROUTINE_BACKEND=$coroutine" >> $config_host_mak
> >   if test "$coroutine_pool" = "yes" ; then
> > diff --git a/fsdev/qemu-fsdev.c b/fsdev/qemu-fsdev.c
> > index bf7f0b0..7f07a2a 100644
> > --- a/fsdev/qemu-fsdev.c
> > +++ b/fsdev/qemu-fsdev.c
> > @@ -27,6 +27,7 @@ static FsDriverTable FsDrivers[] = {
> >   #endif
> >       { .name = "synth", .ops = &synth_ops},
> >       { .name = "proxy", .ops = &proxy_ops},
> > +    { .name = "cephfs", .ops = &cephfs_ops},
> >   };
> >   
> >   int qemu_fsdev_add(QemuOpts *opts)
> > diff --git a/fsdev/qemu-fsdev.h b/fsdev/qemu-fsdev.h
> > index 9fa45bf..86a17b8 100644
> > --- a/fsdev/qemu-fsdev.h
> > +++ b/fsdev/qemu-fsdev.h
> > @@ -22,7 +22,7 @@
> >    * fstype | ops
> >    * -----------------
> >    *  local | local_ops
> > - *  .     |
> > + *  cephfs| cephfs_ops
> >    *  .     |
> >    *  .     |
> >    *  .     |
> > @@ -45,4 +45,5 @@ extern FileOperations local_ops;
> >   extern FileOperations handle_ops;
> >   extern FileOperations synth_ops;
> >   extern FileOperations proxy_ops;
> > +extern FileOperations cephfs_ops;
> >   #endif
> > diff --git a/hw/9pfs/9p-cephfs.c b/hw/9pfs/9p-cephfs.c
> > new file mode 100644
> > index 0000000..f18ec89
> > --- /dev/null
> > +++ b/hw/9pfs/9p-cephfs.c
> > @@ -0,0 +1,739 @@
> > +/*
> > + * Virtio 9p cephfs callback
> > + *
> > + * Copyright UnitedStack, Corp. 2016
> > + *
> > + * Authors:
> > + *    Jevon Qiao <scaleqiao@gmail.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 "qemu/iov.h"
> > +#include "9p.h"
> > +#include "9p-xattr.h"
> > +#include "trace.h"
> > +#include <cephfs/libcephfs.h>
> > +#include "fsdev/qemu-fsdev.h"   /* cephfs_ops */
> > +#include <arpa/inet.h>
> > +#include <pwd.h>
> > +#include <grp.h>
> > +#include <sys/socket.h>
> > +#include <sys/un.h>
> > +#include "qemu/xattr.h"
> > +#include "qemu/error-report.h"
> > +#include <libgen.h>
> > +#include <unistd.h>
> > +#include <linux/fs.h>
> > +#ifdef CONFIG_LINUX_MAGIC_H
> > +#include <linux/magic.h>
> > +#endif
> > +#include <sys/ioctl.h>
> > +
> > +#define CEPH_VER_LEN        32
> > +#define MON_NAME_LEN        32
> > +#define MON_SECRET_LEN      64
> > +
> > +#ifndef LIBCEPHFS_VERSION
> > +#define LIBCEPHFS_VERSION(maj, min, extra) ((maj << 16) + (min << 8) + extra)
> > +#define LIBCEPHFS_VERSION_CODE LIBCEPHFS_VERSION(0, 0, 0)
> > +#endif
> > +
> > +struct cephfs_data {
> > +    int	major, minor, patch;
> > +    char ceph_version[CEPH_VER_LEN];
> > +    struct  ceph_mount_info *cmount;
> > +};
> > +
> > +/*
> > + * Helper function for cephfs_preadv and cephfs_pwritev
> > + */
> > +inline static ssize_t preadv_pwritev(struct ceph_mount_info *cmount, int fd,
> > +                              	    const struct iovec *iov, int iov_cnt,
> > +                              	    off_t offset, bool do_write)
> > +{
> > +    ssize_t ret = 0;
> > +    size_t i = 0;
> > +    size_t len = 0;
> > +    void *buf, *buftmp;
> > +    size_t bufoffset = 0;
> > +
> > +    len = iov_size(iov, iov_cnt);
> > +    buf = g_new0(uint8_t, len);
> > +    buftmp = buf;
> > +    if (do_write) {
> > +        for (i = 0; i < iov_cnt; i++) {
> > +            memcpy((buftmp + bufoffset), iov[i].iov_base, iov[i].iov_len);
> > +            bufoffset += iov[i].iov_len;
> > +        }
> > +        ret = ceph_write(cmount, fd, buf, len, offset);
> > +        if (ret <= 0) {
> > +           errno = -ret;
> > +           ret = -1;
> > +        }
> > +    } else {
> > +        ret = ceph_read(cmount, fd, buf, len, offset);
> > +        if (ret <= 0) {
> > +            errno = -ret;
> > +            ret = -1;
> > +        } else {
> > +            for (i = 0; i < iov_cnt; i++) {
> > +                memcpy(iov[i].iov_base, (buftmp + bufoffset), iov[i].iov_len);
> > +                bufoffset += iov[i].iov_len;
> > +            }
> > +        }
> > +    }
> > +
> > +    free(buf);
> > +    return ret;
> > +}
> > +
> > +static int cephfs_update_file_cred(struct ceph_mount_info *cmount,
> > +				   const char *name, FsCred *credp)
> > +{
> > +    int fd, ret;
> > +    fd = ceph_open(cmount, name, O_NONBLOCK | O_NOFOLLOW, credp->fc_mode);
> > +    if (fd < 0) {
> > +        return fd;
> > +    }
> > +    ret = ceph_fchown(cmount, fd, credp->fc_uid, credp->fc_gid);
> > +    if (ret < 0) {
> > +        goto err_out;
> > +    }
> > +    ret = ceph_fchmod(cmount, fd, credp->fc_mode & 07777);
> > +err_out:
> > +    close(fd);
> > +    return ret;
> > +}
> > +
> > +static int cephfs_lstat(FsContext *fs_ctx, V9fsPath *fs_path,
> > +                        struct stat *stbuf)
> > +{
> > +    int ret;
> > +    char *path = fs_path->data;
> > +    struct cephfs_data *cfsdata = fs_ctx->private;
> > +
> > +    ret = ceph_lstat(cfsdata->cmount, path, stbuf);
> > +    trace_cephfs_lstat_return(path, stbuf->st_mode, stbuf->st_uid, stbuf->st_gid, stbuf->st_size, ret);
> > +    if (ret){
> > +        errno = -ret;
> > +        ret = -1;
> > +    }
> > +    return ret;
> > +}
> > +
> > +static ssize_t cephfs_readlink(FsContext *fs_ctx, V9fsPath *fs_path,
> > +                               char *buf, size_t bufsz)
> > +{
> > +    int ret;
> > +    char *path = fs_path->data;
> > +    struct cephfs_data *cfsdata = fs_ctx->private;
> > +
> > +    ret = ceph_readlink(cfsdata->cmount, path, buf, bufsz);
> > +    trace_cephfs_readlink_return(path, ret);
> > +    return ret;
> > +}
> > +
> > +static int cephfs_close(FsContext *ctx, V9fsFidOpenState *fs)
> > +{
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    return ceph_close(cfsdata->cmount, fs->fd);
> > +}
> > +
> > +static int cephfs_closedir(FsContext *ctx, V9fsFidOpenState *fs)
> > +{
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    return ceph_closedir(cfsdata->cmount, (struct ceph_dir_result *)fs->dir);
> > +}
> > +
> > +static int cephfs_open(FsContext *ctx, V9fsPath *fs_path,
> > +                       int flags, V9fsFidOpenState *fs)
> > +{
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    fs->fd = ceph_open(cfsdata->cmount, fs_path->data, flags, 0777);
> > +    trace_cephfs_open_return(fs_path->data, flags, 0777, fs->fd);
> > +    return fs->fd;
> > +}
> > +
> > +static int cephfs_opendir(FsContext *ctx,
> > +                          V9fsPath *fs_path, V9fsFidOpenState *fs)
> > +{
> > +    int ret;
> > +    struct ceph_dir_result *result;
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +    char *path = fs_path->data;
> > +
> > +    ret = ceph_opendir(cfsdata->cmount, path, &result);
> > +    trace_cephfs_opendir_return(path, ret);
> > +    if (ret) {
> > +        fprintf(stderr, "ceph_opendir=%d\n", ret);
> > +        return ret;
> > +    }
> > +    fs->dir = (DIR *)result;
> > +    if (!fs->dir) {
> > +        fprintf(stderr, "ceph_opendir return NULL for ceph_dir_result\n");
> > +        return -1;
> > +    }
> > +    return 0;
> > +}
> > +
> > +static void cephfs_rewinddir(FsContext *ctx, V9fsFidOpenState *fs)
> > +{
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    trace_cephfs_rewinddir(fs->dir);
> > +    return ceph_rewinddir(cfsdata->cmount, (struct ceph_dir_result *)fs->dir);
> > +}
> > +
> > +static off_t cephfs_telldir(FsContext *ctx, V9fsFidOpenState *fs)
> > +{
> > +    int ret;
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    trace_cephfs_telldir(fs->dir);
> > +    ret = ceph_telldir(cfsdata->cmount, (struct ceph_dir_result *)fs->dir);
> > +    return ret;
> > +}
> > +
> > +static int cephfs_readdir_r(FsContext *ctx, V9fsFidOpenState *fs,
> > +                            struct dirent *entry,
> > +                            struct dirent **result)
> > +{
> > +    int ret;
> > +    struct dirent *tmpent;
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    tmpent = entry;
> > +    ret = ceph_readdir_r(cfsdata->cmount, (struct ceph_dir_result *)fs->dir,
> > +		    	entry);
> > +    trace_cephfs_readdir_r_return(tmpent, entry, ret);
> > +    if (ret > 0 && entry != NULL)
> > +    {
> > +        *result = entry;
> > +    } else if (!ret)
> > +    {
> > +        *result = NULL;
> > +        entry = tmpent;
> > +    }
> > +
> > +    return ret;
> > +}
> > +
> > +static void cephfs_seekdir(FsContext *ctx, V9fsFidOpenState *fs, off_t off)
> > +{
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    trace_cephfs_seekdir(fs->dir, off);
> > +    return ceph_seekdir(cfsdata->cmount, (struct ceph_dir_result*)fs->dir, off);
> > +}
> > +
> > +static ssize_t cephfs_preadv(FsContext *ctx, V9fsFidOpenState *fs,
> > +                             const struct iovec *iov,
> > +                             int iovcnt, off_t offset)
> > +{
> > +    ssize_t ret = 0;
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    trace_cephfs_preadv(iovcnt, iov_size(iov, iovcnt));
> > +#if defined(LIBCEPHFS_VERSION) && LIBCEPHFS_VERSION_CODE >= LIBCEPHFS_VERSION(9, 0, 3)
> > +    ret = ceph_preadv(cfsdata->cmount, fs->fd, iov, iovcnt, offset);
> > +#else
> > +    if (iovcnt > 1) {
> > +	ret = preadv_pwritev(cfsdata->cmount, fs->fd, iov, iovcnt, offset, 0);
> > +    } else if (iovcnt > 0) {
> > +	ret = ceph_read(cfsdata->cmount, fs->fd, iov[0].iov_base,
> > +			iov[0].iov_len, offset);
> > +    }
> > +#endif
> > +    trace_cephfs_preadv_return(iovcnt, iov_size(iov, iovcnt), ret);
> > +
> > +    return ret;
> > +}
> > +
> > +static ssize_t cephfs_pwritev(FsContext *ctx, V9fsFidOpenState *fs,
> > +                              const struct iovec *iov,
> > +                              int iovcnt, off_t offset)
> > +{
> > +    ssize_t ret = 0;
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    trace_cephfs_pwritev(iovcnt, iov_size(iov, iovcnt), offset);
> > +#if defined(LIBCEPHFS_VERSION) && LIBCEPHFS_VERSION_CODE >= LIBCEPHFS_VERSION(9, 0, 3)
> > +    ret = ceph_pwritev(cfsdata->cmount, fs->fd, iov, iovcnt, offset);
> > +#else
> > +    if (iovcnt > 1) {
> > +	ret = preadv_pwritev(cfsdata->cmount, fs->fd, iov, iovcnt, offset, 1);
> > +    } else if (iovcnt > 0) {
> > +	ret = ceph_write(cfsdata->cmount, fs->fd, iov[0].iov_base,
> > +			iov[0].iov_len, offset);
> > +    }
> > +#endif
> > +    trace_cephfs_pwritev_return(iovcnt, iov_size(iov, iovcnt), offset, ret);
> > +
> > +#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 cephfs_chmod(FsContext *fs_ctx, V9fsPath *fs_path, FsCred *credp)
> > +{
> > +    int  ret = -1;
> > +    struct cephfs_data *cfsdata = fs_ctx->private;
> > +
> > +    ret = ceph_chmod(cfsdata->cmount, fs_path->data, credp->fc_mode);
> > +    trace_cephfs_chmod_return(fs_path->data, credp->fc_mode, ret);
> > +    return ret;
> > +}
> > +
> > +static int cephfs_mknod(FsContext *fs_ctx, V9fsPath *dir_path,
> > +                       const char *name, FsCred *credp)
> > +{
> > +    int ret;
> > +    V9fsString fullname;
> > +    struct cephfs_data *cfsdata = fs_ctx->private;
> > +
> > +    v9fs_string_init(&fullname);
> > +    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
> > +    ret = ceph_mknod(cfsdata->cmount, fullname.data, credp->fc_mode,
> > +		    credp->fc_rdev);
> > +    trace_cephfs_mknod_return(fullname.data, credp->fc_mode, credp->fc_rdev, ret);
> > +
> > +    v9fs_string_free(&fullname);
> > +    return ret;
> > +}
> > +
> > +static int cephfs_mkdir(FsContext *fs_ctx, V9fsPath *dir_path,
> > +                       const char *name, FsCred *credp)
> > +{
> > +    int ret;
> > +    V9fsString fullname;
> > +    struct cephfs_data *cfsdata = fs_ctx->private;
> > +
> > +    v9fs_string_init(&fullname);
> > +    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
> > +    ret = ceph_mkdir(cfsdata->cmount, fullname.data, credp->fc_mode);
> > +    trace_cephfs_mkdir_return(fullname.data, credp->fc_mode, ret);
> > +
> > +    v9fs_string_free(&fullname);
> > +    return ret;
> > +}
> > +
> > +static int cephfs_fstat(FsContext *fs_ctx, int fid_type,
> > +                        V9fsFidOpenState *fs, struct stat *stbuf)
> > +{
> > +    int fd = -1;
> > +    int ret;
> > +    struct cephfs_data *cfsdata = fs_ctx->private;
> > +
> > +    if (fid_type == P9_FID_DIR) {
> > +        fd = dirfd(fs->dir);
> > +    } else {
> > +        fd = fs->fd;
> > +    }
> > +    ret = ceph_fstat(cfsdata->cmount, fd, stbuf);
> > +    trace_cephfs_fstat_return(fid_type, fd, stbuf->st_uid, stbuf->st_gid, stbuf->st_size, ret);
> > +
> > +    return ret;
> > +}
> > +
> > +static int cephfs_open2(FsContext *fs_ctx, V9fsPath *dir_path, const char *name,
> > +                        int flags, FsCred *credp, V9fsFidOpenState *fs)
> > +{
> > +    int fd = -1, ret = -1;
> > +    V9fsString fullname;
> > +    struct cephfs_data *cfsdata = fs_ctx->private;
> > +
> > +    v9fs_string_init(&fullname);
> > +    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
> > +    fd = ceph_open(cfsdata->cmount, fullname.data, flags, credp->fc_mode);
> > +    trace_cephfs_open2_return(fullname.data, flags, credp->fc_mode);
> > +    if (fd >= 0) {
> > +        /* After creating the file, need to set the cred */
> > +        ret = cephfs_update_file_cred(cfsdata->cmount, name, credp);
> > +        if (ret < 0) {
> > +            ceph_close(cfsdata->cmount, fd);
> > +            errno = -ret;
> > +            fd = ret;
> > +        } else {
> > +            fs->fd = fd;
> > +        }
> > +    } else {
> > +       errno = -fd;
> > +    }
> > +
> > +    v9fs_string_free(&fullname);
> > +    return fd;
> > +}
> > +
> > +static int cephfs_symlink(FsContext *fs_ctx, const char *oldpath,
> > +                          V9fsPath *dir_path, const char *name, FsCred *credp)
> > +{
> > +    int ret = -1;
> > +    V9fsString fullname;
> > +    struct cephfs_data *cfsdata = fs_ctx->private;
> > +
> > +    v9fs_string_init(&fullname);
> > +    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
> > +    ret = ceph_symlink(cfsdata->cmount, oldpath, fullname.data);
> > +    trace_cephfs_symlink_return(oldpath, fullname.data, ret);
> > +
> > +    v9fs_string_free(&fullname);
> > +    return ret;
> > +}
> > +
> > +static int cephfs_link(FsContext *ctx, V9fsPath *oldpath,
> > +                       V9fsPath *dirpath, const char *name)
> > +{
> > +    int ret = -1;
> > +    V9fsString newpath;
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    v9fs_string_init(&newpath);
> > +    v9fs_string_sprintf(&newpath, "%s/%s", dirpath->data, name);
> > +    ret = ceph_link(cfsdata->cmount, oldpath->data, newpath.data);
> > +    trace_cephfs_link_return(oldpath->data, newpath.data, ret);
> > +
> > +    v9fs_string_free(&newpath);
> > +    return ret;
> > +}
> > +
> > +static int cephfs_truncate(FsContext *ctx, V9fsPath *fs_path, off_t size)
> > +{
> > +    int ret = -1;
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    ret = ceph_truncate(cfsdata->cmount, fs_path->data, size);
> > +    trace_cephfs_truncate_return(fs_path->data, size, ret);
> > +
> > +    return ret;
> > +}
> > +
> > +static int cephfs_rename(FsContext *ctx, const char *oldpath,
> > +                         const char *newpath)
> > +{
> > +    int ret = -1;
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    ret = ceph_rename(cfsdata->cmount, oldpath, newpath);
> > +    trace_cephfs_rename_return(oldpath, newpath, ret);
> > +
> > +    return ret;
> > +}
> > +
> > +static int cephfs_chown(FsContext *fs_ctx, V9fsPath *fs_path, FsCred *credp)
> > +{
> > +    int ret = -1;
> > +    struct cephfs_data *cfsdata = fs_ctx->private;
> > +
> > +    ret = ceph_chown(cfsdata->cmount, fs_path->data, credp->fc_uid,
> > +		    credp->fc_gid);
> > +    trace_cephfs_chown_return(fs_path->data, credp->fc_uid, credp->fc_gid, ret);
> > +
> > +    return ret;
> > +}
> > +
> > +static int cephfs_utimensat(FsContext *ctx, V9fsPath *fs_path,
> > +                            const struct timespec *buf)
> > +{
> > +    int ret = -1;
> > +
> > +#ifdef CONFIG_UTIMENSAT
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    ret = ceph_utime(cfsdata->cmount, fs_path->data, (struct utimbuf *)buf);
> > +    trace_cephfs_utimensat_return(fs_path->data, ret);
> > +#else
> > +    ret = -1;
> > +    errno = ENOSYS;
> > +#endif
> > +
> > +    return ret;
> > +}
> > +
> > +static int cephfs_remove(FsContext *ctx, const char *path)
> > +{
> > +    errno = EOPNOTSUPP;
> > +    return -1;
> > +}
> > +
> > +static int cephfs_fsync(FsContext *ctx, int fid_type,
> > +                        V9fsFidOpenState *fs, int datasync)
> > +{
> > +    int ret = -1, fd = -1;
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    if (fid_type == P9_FID_DIR) {
> > +        fd = dirfd(fs->dir);
> > +    } else {
> > +        fd = fs->fd;
> > +    }
> > +    ret = ceph_fsync(cfsdata->cmount, fd, datasync);
> > +    trace_cephfs_fsync_return(fd, datasync, ret);
> > +
> > +    return ret;
> > +}
> > +
> > +static int cephfs_statfs(FsContext *ctx, V9fsPath *fs_path,
> > +                         struct statfs *stbuf)
> > +{
> > +    int ret;
> > +    char *path = fs_path->data;
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    ret = ceph_statfs(cfsdata->cmount, path, (struct statvfs*)stbuf);
> > +    if (ret) {
> > +        fprintf(stderr, "ceph_statfs=%d\n", ret);
> > +    }
> > +
> > +    return ret;
> > +}
> > +
> > +/*
> > + * Get the extended attribute of normal file, if the path refer to a symbolic
> > + * link, just return the extended attributes of the syslink rather than the
> > + * attributes of the link itself.
> > + */
> > +static ssize_t cephfs_lgetxattr(FsContext *ctx, V9fsPath *fs_path,
> > +                                const char *name, void *value, size_t size)
> > +{
> > +    int ret;
> > +    char *path = fs_path->data;
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    ret = ceph_lgetxattr(cfsdata->cmount, path, name, value, size);
> > +    trace_cephfs_lgetxattr_return(path, name, ret);
> > +
> > +    return ret;
> > +}
> > +
> > +static ssize_t cephfs_llistxattr(FsContext *ctx, V9fsPath *fs_path,
> > +                                 void *value, size_t size)
> > +{
> > +    int ret = -1;
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    ret = ceph_llistxattr(cfsdata->cmount, fs_path->data, value, size);
> > +    trace_cephfs_llistxattr_return(fs_path->data, ret);
> > +
> > +    return ret;
> > +}
> > +
> > +static int cephfs_lsetxattr(FsContext *ctx, V9fsPath *fs_path, const char *name,
> > +                            void *value, size_t size, int flags)
> > +{
> > +    int ret = -1;
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    ret = ceph_lsetxattr(cfsdata->cmount, fs_path->data, name, value, size,
> > +	flags);
> > +    trace_cephfs_lsetxattr_return(fs_path->data, name, flags, ret);
> > +
> > +    return ret;
> > +}
> > +
> > +static int cephfs_lremovexattr(FsContext *ctx, V9fsPath *fs_path,
> > +                               const char *name)
> > +{
> > +    int ret = -1;
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    ret = ceph_lremovexattr(cfsdata->cmount, fs_path->data, name);
> > +    trace_cephfs_lremovexattr_return(fs_path->data, name, ret);
> > +
> > +    return ret;
> > +}
> > +
> > +static int cephfs_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 {
> > +        /* if the path does not start from '/' */
> > +        v9fs_string_sprintf((V9fsString *)target, "%s", name);
> > +    }
> > +
> > +    /* Bump the size for including terminating NULL */
> > +    target->size++;
> > +    return 0;
> > +}
> > +
> > +static int cephfs_renameat(FsContext *ctx, V9fsPath *olddir,
> > +                           const char *old_name, V9fsPath *newdir,
> > +                           const char *new_name)
> > +{
> > +    int ret = -1;
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    ret = ceph_rename(cfsdata->cmount, old_name, new_name);
> > +    trace_cephfs_renameat_return(old_name, new_name, ret);
> > +
> > +    return ret;
> > +}
> > +
> > +static int cephfs_unlinkat(FsContext *ctx, V9fsPath *dir,
> > +                           const char *name, int flags)
> > +{
> > +    int ret = 0;
> > +    char *path = dir->data;
> > +    struct stat fstat;
> > +    V9fsString fullname;
> > +    struct cephfs_data *cfsdata = ctx->private;
> > +
> > +    v9fs_string_init(&fullname);
> > +    v9fs_string_sprintf(&fullname, "%s/%s", dir->data, name);
> > +    path = fullname.data;
> > +    /* determine which kind of file is being destroyed */
> > +    ret = ceph_lstat(cfsdata->cmount, path, &fstat);
> > +    if (!ret) {
> > +        switch (fstat.st_mode & S_IFMT) {
> > +        case S_IFDIR:
> > +            ret = ceph_rmdir(cfsdata->cmount, path);
> > +            break;
> > +
> > +        case S_IFBLK:
> > +        case S_IFCHR:
> > +        case S_IFIFO:
> > +        case S_IFLNK:
> > +        case S_IFREG:
> > +        case S_IFSOCK:
> > +            ret = ceph_unlink(cfsdata->cmount, path);
> > +            break;
> > +
> > +        default:
> > +            fprintf(stderr, "ceph_lstat unknown stmode\n");
> > +            break;
> > +        }
> > +    } else {
> > +        errno = -ret;
> > +        ret = -1;
> > +    }
> > +    trace_cephfs_unlinkat_return(path, fstat.st_mode, ret);
> > +
> > +    v9fs_string_free(&fullname);
> > +    return ret;
> > +}
> > +
> > +/*
> > + * Do two things in the init function:
> > + * 1) Create a mount handle used by all cephfs interfaces.
> > + * 2) Invoke ceph_mount() to initialize a link between the client and
> > + *    ceph monitor
> > + */
> > +static int cephfs_init(FsContext *ctx)
> > +{
> > +    int ret;
> > +    const char *ver = NULL;
> > +    struct cephfs_data *data = g_malloc(sizeof(struct cephfs_data));
> > +
> > +    if (data == NULL) {
> > +	errno = ENOMEM;
> > +	return -1;
> > +    }
> > +    trace_cephfs_init(ctx->fs_root);
> > +    memset(data, 0, sizeof(struct cephfs_data));
> > +    ret = ceph_create(&data->cmount, NULL);
> > +    if (ret) {
> > +        fprintf(stderr, "ceph_create=%d\n", ret);
> > +        goto err_out;
> > +    }
> > +
> > +    ret = ceph_conf_read_file(data->cmount, NULL);
> > +    if (ret) {
> > +        fprintf(stderr, "ceph_conf_read_file=%d\n", ret);
> > +        goto err_out;
> > +    }
> > +
> > +    ret = ceph_mount(data->cmount, ctx->fs_root);
> > +    if (ret) {
> > +        fprintf(stderr, "ceph_mount=%d\n", ret);
> > +        goto err_out;
> > +    } else {
> > +        ctx->private = data;
> > +	/* CephFS does not support FS_IOC_GETVERSIO */
> > +	ctx->exops.get_st_gen = NULL;
> > +        goto out;
> > +    }
> > +
> > +    ver = ceph_version(&data->major, &data->minor, &data->patch);
> > +    memcpy(data->ceph_version, ver, strlen(ver) + 1);
> > +
> > +err_out:
> > +    g_free(data);
> > +out:
> > +    return ret;
> > +}
> > +
> > +static int cephfs_parse_opts(QemuOpts *opts, struct FsDriverEntry *fse)
> > +{
> > +    const char *sec_model = qemu_opt_get(opts, "security_model");
> > +    const char *path = qemu_opt_get(opts, "path");
> > +
> > +    if (!sec_model) {
> > +        fprintf(stderr, "Invalid argument security_model specified with "
> > +		"cephfs fsdriver\n");
> > +        return -1;
> > +    }
> > +
> > +    if (!path) {
> > +        fprintf(stderr, "fsdev: No path specified.\n");
> > +        return -1;
> > +    }
> > +
> > +    fse->path = g_strdup(path);
> > +    return 0;
> > +}
> > +
> > +FileOperations cephfs_ops = {
> > +    .parse_opts   = cephfs_parse_opts,
> > +    .init         = cephfs_init,
> > +    .lstat        = cephfs_lstat,
> > +    .readlink     = cephfs_readlink,
> > +    .close        = cephfs_close,
> > +    .closedir     = cephfs_closedir,
> > +    .open         = cephfs_open,
> > +    .opendir      = cephfs_opendir,
> > +    .rewinddir    = cephfs_rewinddir,
> > +    .telldir      = cephfs_telldir,
> > +    .readdir_r    = cephfs_readdir_r,
> > +    .seekdir      = cephfs_seekdir,
> > +    .preadv       = cephfs_preadv,
> > +    .pwritev      = cephfs_pwritev,
> > +    .chmod        = cephfs_chmod,
> > +    .mknod        = cephfs_mknod,
> > +    .mkdir        = cephfs_mkdir,
> > +    .fstat        = cephfs_fstat,
> > +    .open2        = cephfs_open2,
> > +    .symlink      = cephfs_symlink,
> > +    .link         = cephfs_link,
> > +    .truncate     = cephfs_truncate,
> > +    .rename       = cephfs_rename,
> > +    .chown        = cephfs_chown,
> > +    .utimensat    = cephfs_utimensat,
> > +    .remove       = cephfs_remove,
> > +    .fsync        = cephfs_fsync,
> > +    .statfs       = cephfs_statfs,
> > +    .lgetxattr    = cephfs_lgetxattr,
> > +    .llistxattr   = cephfs_llistxattr,
> > +    .lsetxattr    = cephfs_lsetxattr,
> > +    .lremovexattr = cephfs_lremovexattr,
> > +    .name_to_path = cephfs_name_to_path,
> > +    .renameat     = cephfs_renameat,
> > +    .unlinkat     = cephfs_unlinkat,
> > +};
> > diff --git a/hw/9pfs/Makefile.objs b/hw/9pfs/Makefile.objs
> > index da0ae0c..a77a6f4 100644
> > --- a/hw/9pfs/Makefile.objs
> > +++ b/hw/9pfs/Makefile.objs
> > @@ -5,5 +5,8 @@ common-obj-y += coth.o cofs.o codir.o cofile.o
> >   common-obj-y += coxattr.o 9p-synth.o
> >   common-obj-$(CONFIG_OPEN_BY_HANDLE) +=  9p-handle.o
> >   common-obj-y += 9p-proxy.o
> > +common-obj-y += 9p-cephfs.o
> >   
> >   obj-y += virtio-9p-device.o
> > +
> > +9p-cephfs.o-libs := $(CEPHFS_LIBS)
> > diff --git a/scripts/analyse-9p-simpletrace.py b/scripts/analyse-9p-simpletrace.py
> > index 3c3dee4..fe0a496 100755
> > --- a/scripts/analyse-9p-simpletrace.py
> > +++ b/scripts/analyse-9p-simpletrace.py
> > @@ -210,4 +210,100 @@ class VirtFSRequestTracker(simpletrace.Analyzer):
> >           def v9fs_readlink_return(self, tag, id, target):
> >                   print "RREADLINK (tag =", tag, ", target =", target, ")"
> >   
> > +	def cephfs_lstat_return(self, path, stmode, stuid, stgid, stsize, ret):
> > +		print "RCEPHFSLSTAT (path =", path, ", stmode =", stmode, ", stuid =", stuid, ", stgid =", stgid, ", stsize =", stsize, ", ret =", ret, ")"
> > +
> > +	def cephfs_readlink_return(self, path, ret):
> > +		print "RCEPHFSREADLINK (path =", path, ", ret =", ret, ")"
> > +
> > +	def cephfs_open_return(self, path, flags, mode, fd):
> > +		print "RCEPHFSOPEN (path =", path, ", flags =", flags, ", mode =", mode, ", fd =", fd, ")"
> > +
> > +	def cephfs_opendir_return(self, path, ret):
> > +		print "RCEPHFSOPENDIR (path =", path, ", ret =", ret, ")"
> > +
> > +	def cephfs_rewinddir(self, dir):
> > +		print "TCEPHFSREWINDDIR (dir =", dir, ")"
> > +
> > +	def cephfs_telldir(self, dir):
> > +		print "TCEPHFSTELLDIR (dir =", dir, ")"
> > +
> > +	def cephfs_readdir_r_return(self, tmpent, entry, ret):
> > +		print "RCEPHFSREADDIRR (tmpent =", tmpent, ", entry =", entry, ", ret =", ret, ")"
> > +
> > +	def cephfs_seekdir(self, dir, off):
> > +		print "TCEPHFSSEEKDIR (dir =", dir, ", off =", off, ")"
> > +
> > +	def cephfs_preadv(self, iovcnt, len):
> > +		print "TCEPHFSPREADV (iovcnt=", iovcnt, ", len =", len, ")"
> > +
> > +	def cephfs_preadv_return(self, iovcnt, len, ret):
> > +		print "RCEPHFSPREADV (iovcnt=", iovcnt, ", len =", len, ", ret = ", ret, ")"
> > +
> > +	def cephfs_pwritev(self, iovcnt, len, offset):
> > +		print "TCEPHFSPWRITEV (iovcnt=", iovcnt, ", len =", len, ", offset =", offset, ")"
> > +
> > +	def cephfs_pwritev_return(self, iovcnt, len, offset, ret):
> > +		print "RCEPHFSPWRITEV (iovcnt=", iovcnt, ", len =", len, ", offset =", offset, ", ret = ", ret, ")"
> > +
> > +	def cephfs_chmod(self, path, fcmode):
> > +		print "TCEPHFSCHMOD (path =", path, ", fcmode =", fcmode, ")"
> > +
> > +	def cephfs_chmod_return(self, path, fcmode, ret):
> > +		print "RCEPHFSCHMOD (path =", path, ", fcmode =", fcmode, ", ret =", ret, ")"
> > +
> > +	def cephfs_mknod_return(self, path, fcmode, fcrdev, ret):
> > +		print "RCEPHFSMKNOD (path =", path, ", fcmode =", fcmode, ", fcrdev =", fcrdev, ", ret =", ret, ")"
> > +
> > +	def cephfs_mkdir_return(self, path, fcmode, ret):
> > +		print "RCEPHFSMKDIR (path =", path, ", fcmode =", fcmode, ", ret =", ret, ")"
> > +
> > +	def cephfs_fstat_return(self, fidtype, fd, stuid, stgid, stsize, ret):
> > +		print "RCEPHFSFSTAT (fidtype =", fidtype, ", fd =", fd, ", stuid =", stuid, ", stgid =", stgid, ", stsize =", stsize, ", ret =", ret, ")"
> > +
> > +	def cephfs_open2_return(self, path, flags, fcmode):
> > +		print "RCEPHFSOPEN2 (path =", path, ", flags =", flags, "fcmode =", fcmode, ")"
> > +
> > +	def cephfs_symlink_return(self, oldpath, path, ret):
> > +		print "RCEPHFSSYMLINK (oldpath =", oldpath, ", path =", path, ", ret =", ret, ")"
> > +
> > +	def cephfs_link_return(self, oldpath, path, ret):
> > +		print "RCEPHFSLINK (oldpath =", oldpath, ", path =", path, ", ret =", ret, ")"
> > +
> > +	def cephfs_truncate_return(self, path, size, ret):
> > +		print "RCEPHFSTRUNCATE (path =", path, ", size =", size, ", ret =", ret, ")"
> > +
> > +	def cephfs_rename_return(self, oldpath, newpath, ret):
> > +		print "RCEPHFSRENAME (oldpath =", oldpath, ", newpath =", newpath, ", ret =", ret, ")"
> > +
> > +	def cephfs_chown_return(self, path, fcuid, fcgid, ret):
> > +		print "RCEPHFSCHOWN (path =", path, ", fcuid =", fcuid, ", fcgid =", fcgid, ", ret =", ret, ")"
> > +
> > +	def cephfs_utimensat_return(self, path, ret):
> > +		print "RCEPHFSUTIMENSAT (path =", path, ", ret =", ret, ")"
> > +
> > +	def cephfs_fsync_return(self, fd, datasync, ret):
> > +		print "RCEPHFSFSYNC (fd =", fd, ", datasync =", datasync, ", ret =", ret, ")"
> > +
> > +	def cephfs_lgetxattr_return(self, path, name, ret):
> > +		print "RCEPHFSLGETXATTR (path =", path, ", name =", name, ", ret =", ret, ")"
> > +
> > +	def cephfs_llistxattr_return(self, path, ret):
> > +		print "RCEPHFSLLISTXATTR (path =", path, ", ret =", ret, ")"
> > +
> > +	def cephfs_lsetxattr_return(self, path, name, flags, ret):
> > +		print "RCEPHFSLSETXATTR (path =", path, ", name =", name, ", flags =", flags, ", ret =", ret, ")"
> > +
> > +	def cephfs_lremovexattr_return(self, path, name, ret):
> > +		print "RCEPHFSLREMOVEXATTR (path =", path, ", name =", name, ", ret =", ret, ")"
> > +
> > +	def cephfs_renameat_return(self, oldname, newname, ret):
> > +		print "RCEPHFSRENAMEAT (oldname =", oldname, ", newname =", newname, ", ret =", ret, ")"
> > +
> > +	def cephfs_unlinkat_return(self, path, stmode, ret):
> > +		print "RCEPHFSUNLINKAT (path =", path, ", stmode =", stmode, ", ret =", ret, ")"
> > +
> > +	def cephfs_init(self, path):
> > +		print "RCEPHFSINIT (path =", path, ")"
> > +
> >   simpletrace.run(VirtFSRequestTracker())
> > diff --git a/trace-events b/trace-events
> > index 6fba6cc..11879d2 100644
> > --- a/trace-events
> > +++ b/trace-events
> > @@ -1118,6 +1118,39 @@ v9fs_xattrcreate(uint16_t tag, uint8_t id, int32_t fid, char* name, int64_t size
> >   v9fs_readlink(uint16_t tag, uint8_t id, int32_t fid) "tag %d id %d fid %d"
> >   v9fs_readlink_return(uint16_t tag, uint8_t id, char* target) "tag %d id %d name %s"
> >   
> > +# hw/9pfs/9p-cephfs.c
> > +cephfs_lstat_return(char *path, int stmode, int stuid, int stgid, int stsize, int ret) "path %s stmode %d stuid %d stgid %d stsize %d ret %d"
> > +cephfs_readlink_return(char *path, int ret) "path %s ret %d"
> > +cephfs_open_return(char *path, int flags, int mode, int fd) "path %s flags %d mode %d fd %d"
> > +cephfs_opendir_return(char *path, int ret) "path %s ret %d"
> > +cephfs_rewinddir(void *dir) "dir %p"
> > +cephfs_telldir(void *dir) "dir %p"
> > +cephfs_readdir_r_return(void *tmpent, void *entry, int ret) "tmpent %p entry %p ret %d"
> > +cephfs_seekdir(void *dir, int off) "dir %p off %d"
> > +cephfs_preadv(int iovcnt, int len) "iovcnt %d len %d"
> > +cephfs_preadv_return(int iovcnt, int len, long ret) "iovcnt %d len %d ret %l"
> > +cephfs_pwritev(int iovcnt, int len, int offset) "iovcnt %d len %d offset %d"
> > +cephfs_pwritev_return(int iovcnt, int len, int offset, long ret) "iovcnt %d len %d offset %d ret %l"cephfs_chmod(char *path, int fcmode) "path %s fcmode %d"
> > +cephfs_chmod_return(char *path, int fcmode, int ret) "path %s fcmode %d ret %d"
> > +cephfs_mknod_return(char *path, int fcmode, uint32_t fcrdev, int ret) "path %s fcmode %d fcrdev %u ret %d"
> > +cephfs_mkdir_return(char *path, int fcmode, int ret) " path %s fcmode %d ret %d"
> > +cephfs_fstat_return(int fidtype, int fd, int stuid, int stgid, int stsize, int ret) "fidtype %d fd %d stuid %d stgid %d stsize %d ret %d"
> > +cephfs_open2_return(char *path, int flags, int fcmode) "path %s flags %d fcmode %d"
> > +cephfs_symlink_return(const char *oldpath, char *path, int ret) "oldpath %s path %s ret %d"
> > +cephfs_link_return(char *oldpath, char *path, int ret) "oldpath %s path %s ret %d"
> > +cephfs_truncate_return(char *path, int size, int ret) "path %s size %d ret %d"
> > +cephfs_rename_return(const char *oldpath, const char *newpath, int ret) "oldpath %s newpath %s ret %d"
> > +cephfs_chown_return(char *path, int fcuid, int fcgid, int ret) "path %s fcuid %d fcgid %d ret %d"
> > +cephfs_utimensat_return(char *path, int ret) "path %s ret %d"
> > +cephfs_fsync_return(int fd, int datasync, int ret) "fd %d datasync %d ret %d"
> > +cephfs_lgetxattr_return(char *path, const char *name, int ret) "path %s name %s ret %d"
> > +cephfs_llistxattr_return(char *path, int ret) "path %s ret %d"
> > +cephfs_lsetxattr_return(char *path, const char *name, int flags, int ret) "path %s name %s flags %d ret %d"
> > +cephfs_lremovexattr_return(char *path, const char *name, int ret) "path %s name %s ret %d"
> > +cephfs_renameat_return(const char *oldname, const char *newname, int ret) "oldname %s newname %s ret %d"
> > +cephfs_unlinkat_return(char *path, int stmode, int ret) "path %s stmode %d ret %d"
> > +cephfs_init(char *path) "path %s"
> > +
> >   # target-sparc/mmu_helper.c
> >   mmu_helper_dfault(uint64_t address, uint64_t context, int mmu_idx, uint32_t tl) "DFAULT at %"PRIx64" context %"PRIx64" mmu_idx=%d tl=%d"
> >   mmu_helper_dprot(uint64_t address, uint64_t context, int mmu_idx, uint32_t tl) "DPROT at %"PRIx64" context %"PRIx64" mmu_idx=%d tl=%d"  
> 
> 

--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Greg Kurz March 9, 2016, 7:02 p.m. UTC | #3
On Wed,  2 Mar 2016 23:41:43 +0800
Jevon Qiao <scaleqiao@gmail.com> wrote:

> Ceph as a promising unified distributed storage system is widely used in the
> world of OpenStack. OpenStack users deploying Ceph for block (Cinder) and
> object (S3/Swift) are unsurprisingly looking at Manila and CephFS to round out
> a unified storage solution. Since the typical hypervisor people are using is
> Qemu/KVM, it is necessary to provide a high performance, easy to use, file
> system service in it. VirtFS aims to offers paravirtualized system services and
> simple passthrough for directories from host to guest, which currently only
> support local file system, this patch wants to add CephFS support in VirtFS.
> 
> Signed-off-by: Jevon Qiao <scaleqiao@gmail.com>
> ---

Please fix all the formatting errors and warnings spit by checkpatch.pl, as
mentioned previously.

I also have comments, see below.

>  configure                         |  33 ++
>  fsdev/qemu-fsdev.c                |   1 +
>  fsdev/qemu-fsdev.h                |   3 +-
>  hw/9pfs/9p-cephfs.c               | 739 ++++++++++++++++++++++++++++++++++++++
>  hw/9pfs/Makefile.objs             |   3 +
>  scripts/analyse-9p-simpletrace.py |  96 +++++
>  trace-events                      |  33 ++
>  7 files changed, 907 insertions(+), 1 deletion(-)
>  create mode 100644 hw/9pfs/9p-cephfs.c
> 
> diff --git a/configure b/configure
> index 0c0472a..a2627be 100755
> --- a/configure
> +++ b/configure
> @@ -275,6 +275,7 @@ trace_backends="log"
>  trace_file="trace"
>  spice=""
>  rbd=""
> +cephfs=""
>  smartcard=""
>  libusb=""
>  usb_redir=""
> @@ -1019,6 +1020,10 @@ for opt do
>    ;;
>    --enable-rbd) rbd="yes"
>    ;;
> +  --disable-cephfs) cephfs="no"
> +  ;;
> +  --enable-cephfs) cephfs="yes"
> +  ;;
>    --disable-xfsctl) xfs="no"
>    ;;
>    --enable-xfsctl) xfs="yes"
> @@ -1345,6 +1350,7 @@ disabled with --disable-FEATURE, default is enabled if available:
>    vhost-net       vhost-net acceleration support
>    spice           spice
>    rbd             rados block device (rbd)
> +  cephfs          Ceph File System 
>    libiscsi        iscsi support
>    libnfs          nfs support
>    smartcard       smartcard support (libcacard)
> @@ -3087,6 +3093,28 @@ EOF
>  fi
> 
>  ##########################################
> +# cephfs probe
> +if test "$cephfs" != "no" ; then
> +  cat > $TMPC <<EOF
> +#include <stdio.h>
> +#include <cephfs/libcephfs.h>
> +int main(void) {
> +    struct ceph_mount_info *cmount;
> +    ceph_create(&cmount, NULL);
> +    return 0;
> +}
> +EOF
> +  cephfs_libs="-lcephfs -lrados"

I don't think -lrados is needed here.

> +  if compile_prog "" "$cephfs_libs" ; then
> +    cephfs=yes
> +  else
> +    if test "$cephfs" = "yes" ; then
> +      feature_not_found "cephfs" "Install libcephfs/ceph devel"
> +    fi
> +    cephfs=no
> +  fi
> +fi
> +##########################################
>  # libssh2 probe
>  min_libssh2_version=1.2.8
>  if test "$libssh2" != "no" ; then
> @@ -4760,6 +4788,7 @@ else
>  echo "spice support     $spice"
>  fi
>  echo "rbd support       $rbd"
> +echo "cephfs support    $cephfs"
>  echo "xfsctl support    $xfs"
>  echo "smartcard support $smartcard"
>  echo "libusb            $libusb"
> @@ -5224,6 +5253,10 @@ if test "$rbd" = "yes" ; then
>    echo "RBD_CFLAGS=$rbd_cflags" >> $config_host_mak
>    echo "RBD_LIBS=$rbd_libs" >> $config_host_mak
>  fi
> +if test "$cephfs" = "yes" ; then
> +  echo "CONFIG_CEPHFS=m" >> $config_host_mak
> +  echo "CEPHFS_LIBS=$cephfs_libs" >> $config_host_mak
> +fi
> 
>  echo "CONFIG_COROUTINE_BACKEND=$coroutine" >> $config_host_mak
>  if test "$coroutine_pool" = "yes" ; then
> diff --git a/fsdev/qemu-fsdev.c b/fsdev/qemu-fsdev.c
> index bf7f0b0..7f07a2a 100644
> --- a/fsdev/qemu-fsdev.c
> +++ b/fsdev/qemu-fsdev.c
> @@ -27,6 +27,7 @@ static FsDriverTable FsDrivers[] = {
>  #endif
>      { .name = "synth", .ops = &synth_ops},
>      { .name = "proxy", .ops = &proxy_ops},
> +    { .name = "cephfs", .ops = &cephfs_ops},
>  };
> 
>  int qemu_fsdev_add(QemuOpts *opts)
> diff --git a/fsdev/qemu-fsdev.h b/fsdev/qemu-fsdev.h
> index 9fa45bf..86a17b8 100644
> --- a/fsdev/qemu-fsdev.h
> +++ b/fsdev/qemu-fsdev.h
> @@ -22,7 +22,7 @@
>   * fstype | ops
>   * -----------------
>   *  local | local_ops
> - *  .     |
> + *  cephfs| cephfs_ops
>   *  .     |
>   *  .     |
>   *  .     |
> @@ -45,4 +45,5 @@ extern FileOperations local_ops;
>  extern FileOperations handle_ops;
>  extern FileOperations synth_ops;
>  extern FileOperations proxy_ops;
> +extern FileOperations cephfs_ops;
>  #endif
> diff --git a/hw/9pfs/9p-cephfs.c b/hw/9pfs/9p-cephfs.c
> new file mode 100644
> index 0000000..f18ec89
> --- /dev/null
> +++ b/hw/9pfs/9p-cephfs.c
> @@ -0,0 +1,739 @@
> +/*
> + * Virtio 9p cephfs callback

s/Virtio // since transport isn't involved here.

> + *
> + * Copyright UnitedStack, Corp. 2016
> + *
> + * Authors:
> + *    Jevon Qiao <scaleqiao@gmail.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 "qemu/iov.h"
> +#include "9p.h"
> +#include "9p-xattr.h"
> +#include "trace.h"
> +#include <cephfs/libcephfs.h>
> +#include "fsdev/qemu-fsdev.h"   /* cephfs_ops */
> +#include <arpa/inet.h>
> +#include <pwd.h>
> +#include <grp.h>
> +#include <sys/socket.h>
> +#include <sys/un.h>
> +#include "qemu/xattr.h"
> +#include "qemu/error-report.h"
> +#include <libgen.h>
> +#include <unistd.h>
> +#include <linux/fs.h>
> +#ifdef CONFIG_LINUX_MAGIC_H
> +#include <linux/magic.h>
> +#endif
> +#include <sys/ioctl.h>
> +
> +#define CEPH_VER_LEN        32
> +#define MON_NAME_LEN        32
> +#define MON_SECRET_LEN      64
> +
> +#ifndef LIBCEPHFS_VERSION
> +#define LIBCEPHFS_VERSION(maj, min, extra) ((maj << 16) + (min << 8) + extra)
> +#define LIBCEPHFS_VERSION_CODE LIBCEPHFS_VERSION(0, 0, 0)
> +#endif
> +
> +struct cephfs_data {
> +    int	major, minor, patch;
> +    char ceph_version[CEPH_VER_LEN];
> +    struct  ceph_mount_info *cmount;
> +};
> +
> +/*
> + * Helper function for cephfs_preadv and cephfs_pwritev
> + */
> +inline static ssize_t preadv_pwritev(struct ceph_mount_info *cmount, int fd,
> +                              	    const struct iovec *iov, int iov_cnt,
> +                              	    off_t offset, bool do_write)

Why inline ? I'd rather leave this to the compiler.

> +{
> +    ssize_t ret = 0;
> +    size_t i = 0;
> +    size_t len = 0;

These variables are assigned a value in all branches: no need to initialize
them to 0.

> +    void *buf, *buftmp;
> +    size_t bufoffset = 0;
> +
> +    len = iov_size(iov, iov_cnt);
> +    buf = g_new0(uint8_t, len);
> +    buftmp = buf;
> +    if (do_write) {
> +        for (i = 0; i < iov_cnt; i++) {
> +            memcpy((buftmp + bufoffset), iov[i].iov_base, iov[i].iov_len);
> +            bufoffset += iov[i].iov_len;
> +        }
> +        ret = ceph_write(cmount, fd, buf, len, offset);
> +        if (ret <= 0) {
> +           errno = -ret;
> +           ret = -1;
> +        }
> +    } else {
> +        ret = ceph_read(cmount, fd, buf, len, offset);
> +        if (ret <= 0) {
> +            errno = -ret;
> +            ret = -1;
> +        } else {
> +            for (i = 0; i < iov_cnt; i++) {
> +                memcpy(iov[i].iov_base, (buftmp + bufoffset), iov[i].iov_len);
> +                bufoffset += iov[i].iov_len;
> +            }
> +        }
> +    }
> +

Since all the meaningful code is different for read and write, I'm not
convinced of the interest of having a single helper... Please split.

> +    free(buf);
> +    return ret;
> +}
> +
> +static int cephfs_update_file_cred(struct ceph_mount_info *cmount,
> +				   const char *name, FsCred *credp)
> +{
> +    int fd, ret;
> +    fd = ceph_open(cmount, name, O_NONBLOCK | O_NOFOLLOW, credp->fc_mode);
> +    if (fd < 0) {
> +        return fd;
> +    }
> +    ret = ceph_fchown(cmount, fd, credp->fc_uid, credp->fc_gid);
> +    if (ret < 0) {
> +        goto err_out;
> +    }
> +    ret = ceph_fchmod(cmount, fd, credp->fc_mode & 07777);
> +err_out:
> +    close(fd);
> +    return ret;
> +}
> +
> +static int cephfs_lstat(FsContext *fs_ctx, V9fsPath *fs_path,
> +                        struct stat *stbuf)
> +{
> +    int ret;
> +    char *path = fs_path->data;
> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    ret = ceph_lstat(cfsdata->cmount, path, stbuf);
> +    trace_cephfs_lstat_return(path, stbuf->st_mode, stbuf->st_uid, stbuf->st_gid, stbuf->st_size, ret);
> +    if (ret){

I prefer ret < 0, as you already do it in most places.

> +        errno = -ret; 
> +        ret = -1;

s/ret = -1/return -1/

> +    }
> +    return ret;

s/return ret/return 0/

> +}
> +
> +static ssize_t cephfs_readlink(FsContext *fs_ctx, V9fsPath *fs_path,
> +                               char *buf, size_t bufsz)
> +{
> +    int ret;
> +    char *path = fs_path->data;
> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    ret = ceph_readlink(cfsdata->cmount, path, buf, bufsz);
> +    trace_cephfs_readlink_return(path, ret);
> +    return ret;

ceph_readlink() returns a negative errno, unlike the readlink() call
from the C library.

You need the same "if (ret < 0)" thing to set errno as in cephfs_lstat(),
and the function should return -1.

> +}
> +
> +static int cephfs_close(FsContext *ctx, V9fsFidOpenState *fs)
> +{
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    return ceph_close(cfsdata->cmount, fs->fd);

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_closedir(FsContext *ctx, V9fsFidOpenState *fs)
> +{
> +    struct cephfs_data *cfsdata = ctx->private;
> +   
> +    return ceph_closedir(cfsdata->cmount, (struct ceph_dir_result *)fs->dir);

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_open(FsContext *ctx, V9fsPath *fs_path,
> +                       int flags, V9fsFidOpenState *fs)
> +{
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    fs->fd = ceph_open(cfsdata->cmount, fs_path->data, flags, 0777);
> +    trace_cephfs_open_return(fs_path->data, flags, 0777, fs->fd);
> +    return fs->fd;

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_opendir(FsContext *ctx,
> +                          V9fsPath *fs_path, V9fsFidOpenState *fs)
> +{
> +    int ret;
> +    struct ceph_dir_result *result;
> +    struct cephfs_data *cfsdata = ctx->private;
> +    char *path = fs_path->data;
> +  
> +    ret = ceph_opendir(cfsdata->cmount, path, &result);
> +    trace_cephfs_opendir_return(path, ret);
> +    if (ret) {
> +        fprintf(stderr, "ceph_opendir=%d\n", ret);

Please use error_report() and print a meaningful message... at
least strerror(-ret).

> +        return ret;

Set errno and return -1 on error.

> +    }
> +    fs->dir = (DIR *)result;
> +    if (!fs->dir) {
> +        fprintf(stderr, "ceph_opendir return NULL for ceph_dir_result\n");

Hmm... is this a message for the QEMU user, so that she can fix something and
retry ? I suspect it is more for debugging purposes, in which case I'd rather
add a result argument to trace_cephfs_opendir_return() above.

And BTW, can ceph_opendir() return success without filling the structure ?

> +        return -1;
> +    }
> +    return 0;
> +}
> + 
> +static void cephfs_rewinddir(FsContext *ctx, V9fsFidOpenState *fs)
> +{
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    trace_cephfs_rewinddir(fs->dir);
> +    return ceph_rewinddir(cfsdata->cmount, (struct ceph_dir_result *)fs->dir);

Set errno and return -1 on error.

> +}
> +
> +static off_t cephfs_telldir(FsContext *ctx, V9fsFidOpenState *fs)
> +{
> +    int ret;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    trace_cephfs_telldir(fs->dir);
> +    ret = ceph_telldir(cfsdata->cmount, (struct ceph_dir_result *)fs->dir);
> +    return ret;

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_readdir_r(FsContext *ctx, V9fsFidOpenState *fs,
> +                            struct dirent *entry,
> +                            struct dirent **result)
> +{
> +    int ret;
> +    struct dirent *tmpent;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    tmpent = entry;
> +    ret = ceph_readdir_r(cfsdata->cmount, (struct ceph_dir_result *)fs->dir,
> +		    	entry);
> +    trace_cephfs_readdir_r_return(tmpent, entry, ret);

Hmm... I don't see the point here since tmpent == entry...

> +    if (ret > 0 && entry != NULL)
> +    {
> +        *result = entry;
> +    } else if (!ret)
> +    {
> +        *result = NULL;
> +        entry = tmpent;

This looks even weirder since entry has no users...

> +    }
> +    
> +    return ret;

This function should behave like the original readdir_r() function from the
C library, but it doesn't.

According to the the libcephfs.h header:

 * @returns 1 if the next entry was filled in, 0 if the end of the directory stream was reached,
 *          and a negative error code on failure.
 */
int ceph_readdir_r(struct ceph_mount_info *cmount, struct ceph_dir_result *dirp, struct dirent *de);

and the readdir_r() manual page says:

   The  readdir_r() function returns 0 on success.  On error, it returns a
   positive error number (listed under ERRORS).  If the end of the  direc?
   tory  stream  is  reached,  readdir_r()  returns 0, and returns NULL in
   *result.

If ceph_readdir_r() returns 1, we should return 0 instead of 1.

If ceph_readdir_r() returns 0, we should also return 0 and nullify *result,
while your patch doesn't update *result.

If ceph_readdir_r() returns a negative value ret, we should return -ret
instead of ret.

The code to set errno is also missing.

> +}
> +
> +static void cephfs_seekdir(FsContext *ctx, V9fsFidOpenState *fs, off_t off)
> +{
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    trace_cephfs_seekdir(fs->dir, off);
> +    return ceph_seekdir(cfsdata->cmount, (struct ceph_dir_result*)fs->dir, off);

This is a void function, return is not needed.

> +}
> +
> +static ssize_t cephfs_preadv(FsContext *ctx, V9fsFidOpenState *fs,
> +                             const struct iovec *iov,
> +                             int iovcnt, off_t offset)
> +{
> +    ssize_t ret = 0;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    trace_cephfs_preadv(iovcnt, iov_size(iov, iovcnt));
> +#if defined(LIBCEPHFS_VERSION) && LIBCEPHFS_VERSION_CODE >= LIBCEPHFS_VERSION(9, 0, 3) 
> +    ret = ceph_preadv(cfsdata->cmount, fs->fd, iov, iovcnt, offset); 
> +#else

According to the preadv() manual page:

   EINVAL The  vector  count  iovcnt is less than zero or greater than the
          permitted maximum.

so we should do the same. Since this is a sanity check, it is better placed
before functional checks.

if (iovcnt < 0) {
    errno = EINVAL;
    ret = -1;
} else ...

> +    if (iovcnt > 1) {
> +	ret = preadv_pwritev(cfsdata->cmount, fs->fd, iov, iovcnt, offset, 0);
> +    } else if (iovcnt > 0) {

I'd rather make it explicitly iovcnt == 1.

> +	ret = ceph_read(cfsdata->cmount, fs->fd, iov[0].iov_base,
> +			iov[0].iov_len, offset);
> +    }
> +#endif

Ok, so IIUC, only newer versions of libcephfs have ceph_preadv().

I suggest you move all the code in the #else part to a local
ceph_preadv() implementation that only gets compiled when
libcephfs is too old. A bit like this:

#if defined(LIBCEPHFS_VERSION) && LIBCEPHFS_VERSION_CODE >= LIBCEPHFS_VERSION(9, 0, 3) 
#define HAVE_CEPH_PREADV 1
#endif

#ifndef HAVE_CEPH_PREADV
static int ceph_preadv(struct ceph_mount_info *cmount, int fd, const struct iovec *iov,
                       int iovcnt, int64_t offset)
{
	...
}
#endif

The cephfs_preadv() function then just needs to call ceph_preadv().

> +    trace_cephfs_preadv_return(iovcnt, iov_size(iov, iovcnt), ret);
> +

Set errno and return -1 on error.

> +    return ret;
> +}
> +
> +static ssize_t cephfs_pwritev(FsContext *ctx, V9fsFidOpenState *fs,
> +                              const struct iovec *iov,
> +                              int iovcnt, off_t offset)
> +{
> +    ssize_t ret = 0;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    trace_cephfs_pwritev(iovcnt, iov_size(iov, iovcnt), offset);
> +#if defined(LIBCEPHFS_VERSION) && LIBCEPHFS_VERSION_CODE >= LIBCEPHFS_VERSION(9, 0, 3) 
> +    ret = ceph_pwritev(cfsdata->cmount, fs->fd, iov, iovcnt, offset);
> +#else
> +    if (iovcnt > 1) {
> +	ret = preadv_pwritev(cfsdata->cmount, fs->fd, iov, iovcnt, offset, 1);
> +    } else if (iovcnt > 0) {
> +	ret = ceph_write(cfsdata->cmount, fs->fd, iov[0].iov_base,
> +			iov[0].iov_len, offset);
> +    }
> +#endif
> +    trace_cephfs_pwritev_return(iovcnt, iov_size(iov, iovcnt), offset, ret);
> +
> +#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;

Same remarks as for cephfs_preadv().

> +}
> +
> +static int cephfs_chmod(FsContext *fs_ctx, V9fsPath *fs_path, FsCred *credp)
> +{
> +    int  ret = -1;

Initialization isn't needed.

> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    ret = ceph_chmod(cfsdata->cmount, fs_path->data, credp->fc_mode);
> +    trace_cephfs_chmod_return(fs_path->data, credp->fc_mode, ret);
> +    return ret;

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_mknod(FsContext *fs_ctx, V9fsPath *dir_path,
> +                       const char *name, FsCred *credp)
> +{
> +    int ret;
> +    V9fsString fullname;
> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    v9fs_string_init(&fullname);
> +    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
> +    ret = ceph_mknod(cfsdata->cmount, fullname.data, credp->fc_mode,
> +		    credp->fc_rdev);
> +    trace_cephfs_mknod_return(fullname.data, credp->fc_mode, credp->fc_rdev, ret);
> +
> +    v9fs_string_free(&fullname);
> +    return ret;

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_mkdir(FsContext *fs_ctx, V9fsPath *dir_path,
> +                       const char *name, FsCred *credp)
> +{
> +    int ret;
> +    V9fsString fullname;
> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    v9fs_string_init(&fullname);
> +    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
> +    ret = ceph_mkdir(cfsdata->cmount, fullname.data, credp->fc_mode);
> +    trace_cephfs_mkdir_return(fullname.data, credp->fc_mode, ret);
> +
> +    v9fs_string_free(&fullname);
> +    return ret;

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_fstat(FsContext *fs_ctx, int fid_type,
> +                        V9fsFidOpenState *fs, struct stat *stbuf)
> +{
> +    int fd = -1;

Initialization isn't needed.

> +    int ret;
> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    if (fid_type == P9_FID_DIR) {
> +        fd = dirfd(fs->dir);
> +    } else {
> +        fd = fs->fd;
> +    }
> +    ret = ceph_fstat(cfsdata->cmount, fd, stbuf);
> +    trace_cephfs_fstat_return(fid_type, fd, stbuf->st_uid, stbuf->st_gid, stbuf->st_size, ret);
> +
> +    return ret;

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_open2(FsContext *fs_ctx, V9fsPath *dir_path, const char *name,
> +                        int flags, FsCred *credp, V9fsFidOpenState *fs)
> +{
> +    int fd = -1, ret = -1;

Initialization isn't needed.

> +    V9fsString fullname;
> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    v9fs_string_init(&fullname);
> +    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
> +    fd = ceph_open(cfsdata->cmount, fullname.data, flags, credp->fc_mode);
> +    trace_cephfs_open2_return(fullname.data, flags, credp->fc_mode);
> +    if (fd >= 0) {
> +        /* After creating the file, need to set the cred */
> +        ret = cephfs_update_file_cred(cfsdata->cmount, name, credp);
> +        if (ret < 0) {
> +            ceph_close(cfsdata->cmount, fd);
> +            errno = -ret;
> +            fd = ret;

s/fd = ret/fd = -1/

> +        } else {
> +            fs->fd = fd;
> +        }
> +    } else {
> +       errno = -fd;

and

fd = -1;

> +    }
> +
> +    v9fs_string_free(&fullname);
> +    return fd;
> +}
> +
> +static int cephfs_symlink(FsContext *fs_ctx, const char *oldpath,
> +                          V9fsPath *dir_path, const char *name, FsCred *credp)
> +{
> +    int ret = -1;

Initialization isn't needed.

> +    V9fsString fullname;
> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    v9fs_string_init(&fullname);
> +    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
> +    ret = ceph_symlink(cfsdata->cmount, oldpath, fullname.data);
> +    trace_cephfs_symlink_return(oldpath, fullname.data, ret);
> +
> +    v9fs_string_free(&fullname);

Set errno and return -1 on error.

> +    return ret;
> +}
> +
> +static int cephfs_link(FsContext *ctx, V9fsPath *oldpath,
> +                       V9fsPath *dirpath, const char *name)
> +{
> +    int ret = -1;

Initialization isn't needed.

> +    V9fsString newpath;
> +    struct cephfs_data *cfsdata = ctx->private;
> +    
> +    v9fs_string_init(&newpath);
> +    v9fs_string_sprintf(&newpath, "%s/%s", dirpath->data, name);
> +    ret = ceph_link(cfsdata->cmount, oldpath->data, newpath.data);
> +    trace_cephfs_link_return(oldpath->data, newpath.data, ret);
> +
> +    v9fs_string_free(&newpath); 
> +    return ret;

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_truncate(FsContext *ctx, V9fsPath *fs_path, off_t size)
> +{
> +    int ret = -1;

Initialization isn't needed.

> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    ret = ceph_truncate(cfsdata->cmount, fs_path->data, size);
> +    trace_cephfs_truncate_return(fs_path->data, size, ret);
> +
> +    return ret;

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_rename(FsContext *ctx, const char *oldpath,
> +                         const char *newpath)
> +{
> +    int ret = -1;

Initialization isn't needed.

> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    ret = ceph_rename(cfsdata->cmount, oldpath, newpath);
> +    trace_cephfs_rename_return(oldpath, newpath, ret);
> +
> +    return ret;

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_chown(FsContext *fs_ctx, V9fsPath *fs_path, FsCred *credp)
> +{
> +    int ret = -1;

Initialization isn't needed.

> +    struct cephfs_data *cfsdata = fs_ctx->private;
> +
> +    ret = ceph_chown(cfsdata->cmount, fs_path->data, credp->fc_uid,
> +		    credp->fc_gid);
> +    trace_cephfs_chown_return(fs_path->data, credp->fc_uid, credp->fc_gid, ret);
> +
> +    return ret;

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_utimensat(FsContext *ctx, V9fsPath *fs_path,
> +                            const struct timespec *buf)
> +{
> +    int ret = -1;

Initialization isn't needed.

> +
> +#ifdef CONFIG_UTIMENSAT
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    ret = ceph_utime(cfsdata->cmount, fs_path->data, (struct utimbuf *)buf);
> +    trace_cephfs_utimensat_return(fs_path->data, ret);
> +#else
> +    ret = -1;
> +    errno = ENOSYS;
> +#endif
> +
> +    return ret;

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_remove(FsContext *ctx, const char *path)
> +{
> +    errno = EOPNOTSUPP;
> +    return -1;
> +}
> +
> +static int cephfs_fsync(FsContext *ctx, int fid_type,
> +                        V9fsFidOpenState *fs, int datasync)
> +{
> +    int ret = -1, fd = -1;

Initialization isn't needed.

> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    if (fid_type == P9_FID_DIR) {
> +        fd = dirfd(fs->dir);
> +    } else {
> +        fd = fs->fd;
> +    }
> +    ret = ceph_fsync(cfsdata->cmount, fd, datasync);
> +    trace_cephfs_fsync_return(fd, datasync, ret);
> +
> +    return ret;

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_statfs(FsContext *ctx, V9fsPath *fs_path,
> +                         struct statfs *stbuf)
> +{
> +    int ret;
> +    char *path = fs_path->data;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    ret = ceph_statfs(cfsdata->cmount, path, (struct statvfs*)stbuf);
> +    if (ret) {
> +        fprintf(stderr, "ceph_statfs=%d\n", ret); 

error_report() with a meaningful message.

> +    }
> +
> +    return ret;

Set errno and return -1 on error.

> +}
> +
> +/*
> + * Get the extended attribute of normal file, if the path refer to a symbolic
> + * link, just return the extended attributes of the syslink rather than the
> + * attributes of the link itself.
> + */
> +static ssize_t cephfs_lgetxattr(FsContext *ctx, V9fsPath *fs_path,
> +                                const char *name, void *value, size_t size)
> +{
> +    int ret;
> +    char *path = fs_path->data;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    ret = ceph_lgetxattr(cfsdata->cmount, path, name, value, size);
> +    trace_cephfs_lgetxattr_return(path, name, ret);
> +
> +    return ret;

Set errno and return -1 on error.

> +}
> +
> +static ssize_t cephfs_llistxattr(FsContext *ctx, V9fsPath *fs_path,
> +                                 void *value, size_t size)
> +{
> +    int ret = -1;

Initialization isn't needed.

> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    ret = ceph_llistxattr(cfsdata->cmount, fs_path->data, value, size);
> +    trace_cephfs_llistxattr_return(fs_path->data, ret);
> +
> +    return ret;

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_lsetxattr(FsContext *ctx, V9fsPath *fs_path, const char *name,
> +                            void *value, size_t size, int flags)
> +{
> +    int ret = -1;

Initialization isn't needed.

> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    ret = ceph_lsetxattr(cfsdata->cmount, fs_path->data, name, value, size,
> +	flags);
> +    trace_cephfs_lsetxattr_return(fs_path->data, name, flags, ret);
> +
> +    return ret;

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_lremovexattr(FsContext *ctx, V9fsPath *fs_path,
> +                               const char *name)
> +{
> +    int ret = -1;

Initialization isn't needed.

> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    ret = ceph_lremovexattr(cfsdata->cmount, fs_path->data, name);
> +    trace_cephfs_lremovexattr_return(fs_path->data, name, ret);
> +
> +    return ret;

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_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 {
> +        /* if the path does not start from '/' */
> +        v9fs_string_sprintf((V9fsString *)target, "%s", name);
> +    }
> +
> +    /* Bump the size for including terminating NULL */ 
> +    target->size++;
> +    return 0;
> +}
> +
> +static int cephfs_renameat(FsContext *ctx, V9fsPath *olddir,
> +                           const char *old_name, V9fsPath *newdir,
> +                           const char *new_name)
> +{
> +    int ret = -1;

Initialization isn't needed.

> +    struct cephfs_data *cfsdata = ctx->private;
> +    
> +    ret = ceph_rename(cfsdata->cmount, old_name, new_name);
> +    trace_cephfs_renameat_return(old_name, new_name, ret);
> +
> +    return ret;

Set errno and return -1 on error.

> +}
> +
> +static int cephfs_unlinkat(FsContext *ctx, V9fsPath *dir,
> +                           const char *name, int flags)
> +{
> +    int ret = 0;

Initialization isn't needed.

> +    char *path = dir->data;
> +    struct stat fstat;
> +    V9fsString fullname;
> +    struct cephfs_data *cfsdata = ctx->private;
> +
> +    v9fs_string_init(&fullname);
> +    v9fs_string_sprintf(&fullname, "%s/%s", dir->data, name);
> +    path = fullname.data;
> +    /* determine which kind of file is being destroyed */ 
> +    ret = ceph_lstat(cfsdata->cmount, path, &fstat);
> +    if (!ret) {
> +        switch (fstat.st_mode & S_IFMT) {
> +        case S_IFDIR:
> +            ret = ceph_rmdir(cfsdata->cmount, path);
> +            break;
> +
> +        case S_IFBLK:
> +        case S_IFCHR:
> +        case S_IFIFO:
> +        case S_IFLNK:
> +        case S_IFREG:
> +        case S_IFSOCK:
> +            ret = ceph_unlink(cfsdata->cmount, path);
> +            break;
> +
> +        default:
> +            fprintf(stderr, "ceph_lstat unknown stmode\n");

error_report()

> +            break;
> +        }
> +    } else {
> +        errno = -ret;
> +        ret = -1;
> +    }
> +    trace_cephfs_unlinkat_return(path, fstat.st_mode, ret);
> +
> +    v9fs_string_free(&fullname);
> +    return ret;
> +}
> +
> +/*
> + * Do two things in the init function:
> + * 1) Create a mount handle used by all cephfs interfaces.
> + * 2) Invoke ceph_mount() to initialize a link between the client and 
> + *    ceph monitor
> + */
> +static int cephfs_init(FsContext *ctx)
> +{
> +    int ret;
> +    const char *ver = NULL;
> +    struct cephfs_data *data = g_malloc(sizeof(struct cephfs_data));
> +
> +    if (data == NULL) {
> +	errno = ENOMEM;
> +	return -1;
> +    }
> +    trace_cephfs_init(ctx->fs_root);
> +    memset(data, 0, sizeof(struct cephfs_data));
> +    ret = ceph_create(&data->cmount, NULL);
> +    if (ret) {
> +        fprintf(stderr, "ceph_create=%d\n", ret);

error_report()

> +        goto err_out;
> +    }
> +
> +    ret = ceph_conf_read_file(data->cmount, NULL);
> +    if (ret) {
> +        fprintf(stderr, "ceph_conf_read_file=%d\n", ret);

error_report()

> +        goto err_out;
> +    }
> +
> +    ret = ceph_mount(data->cmount, ctx->fs_root);
> +    if (ret) {
> +        fprintf(stderr, "ceph_mount=%d\n", ret);

error_report()

> +        goto err_out;
> +    } else {
> +        ctx->private = data;
> +	/* CephFS does not support FS_IOC_GETVERSIO */ 
> +	ctx->exops.get_st_gen = NULL;
> +        goto out;
> +    }
> +
> +    ver = ceph_version(&data->major, &data->minor, &data->patch);
> +    memcpy(data->ceph_version, ver, strlen(ver) + 1);
> +    
> +err_out:
> +    g_free(data);
> +out:
> +    return ret;
> +}
> +
> +static int cephfs_parse_opts(QemuOpts *opts, struct FsDriverEntry *fse)
> +{
> +    const char *sec_model = qemu_opt_get(opts, "security_model");
> +    const char *path = qemu_opt_get(opts, "path");
> +
> +    if (!sec_model) {
> +        fprintf(stderr, "Invalid argument security_model specified with "
> +		"cephfs fsdriver\n");
> +        return -1;
> +    }
> +
> +    if (!path) {
> +        fprintf(stderr, "fsdev: No path specified.\n");
> +        return -1;
> +    }
> +
> +    fse->path = g_strdup(path);
> +    return 0;
> +}
> +
> +FileOperations cephfs_ops = {
> +    .parse_opts   = cephfs_parse_opts,
> +    .init         = cephfs_init,
> +    .lstat        = cephfs_lstat,
> +    .readlink     = cephfs_readlink,
> +    .close        = cephfs_close,
> +    .closedir     = cephfs_closedir,
> +    .open         = cephfs_open,
> +    .opendir      = cephfs_opendir,
> +    .rewinddir    = cephfs_rewinddir,
> +    .telldir      = cephfs_telldir,
> +    .readdir_r    = cephfs_readdir_r,
> +    .seekdir      = cephfs_seekdir,
> +    .preadv       = cephfs_preadv,
> +    .pwritev      = cephfs_pwritev,
> +    .chmod        = cephfs_chmod,
> +    .mknod        = cephfs_mknod,
> +    .mkdir        = cephfs_mkdir,
> +    .fstat        = cephfs_fstat,
> +    .open2        = cephfs_open2,
> +    .symlink      = cephfs_symlink,
> +    .link         = cephfs_link,
> +    .truncate     = cephfs_truncate,
> +    .rename       = cephfs_rename,
> +    .chown        = cephfs_chown,
> +    .utimensat    = cephfs_utimensat,
> +    .remove       = cephfs_remove,
> +    .fsync        = cephfs_fsync,
> +    .statfs       = cephfs_statfs,
> +    .lgetxattr    = cephfs_lgetxattr,
> +    .llistxattr   = cephfs_llistxattr,
> +    .lsetxattr    = cephfs_lsetxattr,
> +    .lremovexattr = cephfs_lremovexattr,
> +    .name_to_path = cephfs_name_to_path,
> +    .renameat     = cephfs_renameat,
> +    .unlinkat     = cephfs_unlinkat,
> +};
> diff --git a/hw/9pfs/Makefile.objs b/hw/9pfs/Makefile.objs
> index da0ae0c..a77a6f4 100644
> --- a/hw/9pfs/Makefile.objs
> +++ b/hw/9pfs/Makefile.objs
> @@ -5,5 +5,8 @@ common-obj-y += coth.o cofs.o codir.o cofile.o
>  common-obj-y += coxattr.o 9p-synth.o
>  common-obj-$(CONFIG_OPEN_BY_HANDLE) +=  9p-handle.o
>  common-obj-y += 9p-proxy.o
> +common-obj-y += 9p-cephfs.o
> 
>  obj-y += virtio-9p-device.o
> +
> +9p-cephfs.o-libs := $(CEPHFS_LIBS)
> diff --git a/scripts/analyse-9p-simpletrace.py b/scripts/analyse-9p-simpletrace.py
> index 3c3dee4..fe0a496 100755
> --- a/scripts/analyse-9p-simpletrace.py
> +++ b/scripts/analyse-9p-simpletrace.py

Even if both spellings are correct according to http://www.thefreedictionary.com/,
it seems that the QEMU code only uses analyze currently. Please rename for consistency.

> @@ -210,4 +210,100 @@ class VirtFSRequestTracker(simpletrace.Analyzer):
>          def v9fs_readlink_return(self, tag, id, target):
>                  print "RREADLINK (tag =", tag, ", target =", target, ")"
> 
> +	def cephfs_lstat_return(self, path, stmode, stuid, stgid, stsize, ret):
> +		print "RCEPHFSLSTAT (path =", path, ", stmode =", stmode, ", stuid =", stuid, ", stgid =", stgid, ", stsize =", stsize, ", ret =", ret, ")"
> +
> +	def cephfs_readlink_return(self, path, ret):
> +		print "RCEPHFSREADLINK (path =", path, ", ret =", ret, ")"
> +
> +	def cephfs_open_return(self, path, flags, mode, fd):
> +		print "RCEPHFSOPEN (path =", path, ", flags =", flags, ", mode =", mode, ", fd =", fd, ")"
> +
> +	def cephfs_opendir_return(self, path, ret):
> +		print "RCEPHFSOPENDIR (path =", path, ", ret =", ret, ")"
> +
> +	def cephfs_rewinddir(self, dir):
> +		print "TCEPHFSREWINDDIR (dir =", dir, ")"
> +
> +	def cephfs_telldir(self, dir):
> +		print "TCEPHFSTELLDIR (dir =", dir, ")"
> +
> +	def cephfs_readdir_r_return(self, tmpent, entry, ret):
> +		print "RCEPHFSREADDIRR (tmpent =", tmpent, ", entry =", entry, ", ret =", ret, ")"
> +
> +	def cephfs_seekdir(self, dir, off):
> +		print "TCEPHFSSEEKDIR (dir =", dir, ", off =", off, ")"
> +
> +	def cephfs_preadv(self, iovcnt, len):
> +		print "TCEPHFSPREADV (iovcnt=", iovcnt, ", len =", len, ")"
> +
> +	def cephfs_preadv_return(self, iovcnt, len, ret):
> +		print "RCEPHFSPREADV (iovcnt=", iovcnt, ", len =", len, ", ret = ", ret, ")"
> +
> +	def cephfs_pwritev(self, iovcnt, len, offset):
> +		print "TCEPHFSPWRITEV (iovcnt=", iovcnt, ", len =", len, ", offset =", offset, ")"
> +
> +	def cephfs_pwritev_return(self, iovcnt, len, offset, ret):
> +		print "RCEPHFSPWRITEV (iovcnt=", iovcnt, ", len =", len, ", offset =", offset, ", ret = ", ret, ")"
> +
> +	def cephfs_chmod(self, path, fcmode):
> +		print "TCEPHFSCHMOD (path =", path, ", fcmode =", fcmode, ")"
> +
> +	def cephfs_chmod_return(self, path, fcmode, ret):
> +		print "RCEPHFSCHMOD (path =", path, ", fcmode =", fcmode, ", ret =", ret, ")"
> +
> +	def cephfs_mknod_return(self, path, fcmode, fcrdev, ret):
> +		print "RCEPHFSMKNOD (path =", path, ", fcmode =", fcmode, ", fcrdev =", fcrdev, ", ret =", ret, ")"
> +
> +	def cephfs_mkdir_return(self, path, fcmode, ret):
> +		print "RCEPHFSMKDIR (path =", path, ", fcmode =", fcmode, ", ret =", ret, ")"
> +
> +	def cephfs_fstat_return(self, fidtype, fd, stuid, stgid, stsize, ret):
> +		print "RCEPHFSFSTAT (fidtype =", fidtype, ", fd =", fd, ", stuid =", stuid, ", stgid =", stgid, ", stsize =", stsize, ", ret =", ret, ")"
> +
> +	def cephfs_open2_return(self, path, flags, fcmode):
> +		print "RCEPHFSOPEN2 (path =", path, ", flags =", flags, "fcmode =", fcmode, ")"
> +
> +	def cephfs_symlink_return(self, oldpath, path, ret):
> +		print "RCEPHFSSYMLINK (oldpath =", oldpath, ", path =", path, ", ret =", ret, ")"
> +
> +	def cephfs_link_return(self, oldpath, path, ret):
> +		print "RCEPHFSLINK (oldpath =", oldpath, ", path =", path, ", ret =", ret, ")"
> +
> +	def cephfs_truncate_return(self, path, size, ret):
> +		print "RCEPHFSTRUNCATE (path =", path, ", size =", size, ", ret =", ret, ")"
> +
> +	def cephfs_rename_return(self, oldpath, newpath, ret):
> +		print "RCEPHFSRENAME (oldpath =", oldpath, ", newpath =", newpath, ", ret =", ret, ")"
> +
> +	def cephfs_chown_return(self, path, fcuid, fcgid, ret):
> +		print "RCEPHFSCHOWN (path =", path, ", fcuid =", fcuid, ", fcgid =", fcgid, ", ret =", ret, ")"
> +
> +	def cephfs_utimensat_return(self, path, ret):
> +		print "RCEPHFSUTIMENSAT (path =", path, ", ret =", ret, ")"
> +
> +	def cephfs_fsync_return(self, fd, datasync, ret):
> +		print "RCEPHFSFSYNC (fd =", fd, ", datasync =", datasync, ", ret =", ret, ")"
> +
> +	def cephfs_lgetxattr_return(self, path, name, ret):
> +		print "RCEPHFSLGETXATTR (path =", path, ", name =", name, ", ret =", ret, ")"
> +
> +	def cephfs_llistxattr_return(self, path, ret):
> +		print "RCEPHFSLLISTXATTR (path =", path, ", ret =", ret, ")"
> +
> +	def cephfs_lsetxattr_return(self, path, name, flags, ret):
> +		print "RCEPHFSLSETXATTR (path =", path, ", name =", name, ", flags =", flags, ", ret =", ret, ")"
> +
> +	def cephfs_lremovexattr_return(self, path, name, ret):
> +		print "RCEPHFSLREMOVEXATTR (path =", path, ", name =", name, ", ret =", ret, ")"
> +
> +	def cephfs_renameat_return(self, oldname, newname, ret):
> +		print "RCEPHFSRENAMEAT (oldname =", oldname, ", newname =", newname, ", ret =", ret, ")"
> +
> +	def cephfs_unlinkat_return(self, path, stmode, ret):
> +		print "RCEPHFSUNLINKAT (path =", path, ", stmode =", stmode, ", ret =", ret, ")"
> +
> +	def cephfs_init(self, path):
> +		print "RCEPHFSINIT (path =", path, ")"
> +
>  simpletrace.run(VirtFSRequestTracker())
> diff --git a/trace-events b/trace-events
> index 6fba6cc..11879d2 100644
> --- a/trace-events
> +++ b/trace-events
> @@ -1118,6 +1118,39 @@ v9fs_xattrcreate(uint16_t tag, uint8_t id, int32_t fid, char* name, int64_t size
>  v9fs_readlink(uint16_t tag, uint8_t id, int32_t fid) "tag %d id %d fid %d"
>  v9fs_readlink_return(uint16_t tag, uint8_t id, char* target) "tag %d id %d name %s"
> 
> +# hw/9pfs/9p-cephfs.c 
> +cephfs_lstat_return(char *path, int stmode, int stuid, int stgid, int stsize, int ret) "path %s stmode %d stuid %d stgid %d stsize %d ret %d"
> +cephfs_readlink_return(char *path, int ret) "path %s ret %d"
> +cephfs_open_return(char *path, int flags, int mode, int fd) "path %s flags %d mode %d fd %d"
> +cephfs_opendir_return(char *path, int ret) "path %s ret %d"
> +cephfs_rewinddir(void *dir) "dir %p"
> +cephfs_telldir(void *dir) "dir %p"
> +cephfs_readdir_r_return(void *tmpent, void *entry, int ret) "tmpent %p entry %p ret %d"
> +cephfs_seekdir(void *dir, int off) "dir %p off %d"
> +cephfs_preadv(int iovcnt, int len) "iovcnt %d len %d"
> +cephfs_preadv_return(int iovcnt, int len, long ret) "iovcnt %d len %d ret %l"

ret %ld

> +cephfs_pwritev(int iovcnt, int len, int offset) "iovcnt %d len %d offset %d"
> +cephfs_pwritev_return(int iovcnt, int len, int offset, long ret) "iovcnt %d len %d offset %d ret %l"cephfs_chmod(char *path, int fcmode) "path %s fcmode %d"

ret %ld for cephfs_pwritev_return() and missing newline before cephfs_chmod()

I'm wondering if this patch was build tested before being posted to the list.

> +cephfs_chmod_return(char *path, int fcmode, int ret) "path %s fcmode %d ret %d"
> +cephfs_mknod_return(char *path, int fcmode, uint32_t fcrdev, int ret) "path %s fcmode %d fcrdev %u ret %d"
> +cephfs_mkdir_return(char *path, int fcmode, int ret) " path %s fcmode %d ret %d"
> +cephfs_fstat_return(int fidtype, int fd, int stuid, int stgid, int stsize, int ret) "fidtype %d fd %d stuid %d stgid %d stsize %d ret %d"
> +cephfs_open2_return(char *path, int flags, int fcmode) "path %s flags %d fcmode %d"
> +cephfs_symlink_return(const char *oldpath, char *path, int ret) "oldpath %s path %s ret %d"
> +cephfs_link_return(char *oldpath, char *path, int ret) "oldpath %s path %s ret %d"
> +cephfs_truncate_return(char *path, int size, int ret) "path %s size %d ret %d"
> +cephfs_rename_return(const char *oldpath, const char *newpath, int ret) "oldpath %s newpath %s ret %d"
> +cephfs_chown_return(char *path, int fcuid, int fcgid, int ret) "path %s fcuid %d fcgid %d ret %d"
> +cephfs_utimensat_return(char *path, int ret) "path %s ret %d"
> +cephfs_fsync_return(int fd, int datasync, int ret) "fd %d datasync %d ret %d"
> +cephfs_lgetxattr_return(char *path, const char *name, int ret) "path %s name %s ret %d"
> +cephfs_llistxattr_return(char *path, int ret) "path %s ret %d"
> +cephfs_lsetxattr_return(char *path, const char *name, int flags, int ret) "path %s name %s flags %d ret %d"
> +cephfs_lremovexattr_return(char *path, const char *name, int ret) "path %s name %s ret %d"
> +cephfs_renameat_return(const char *oldname, const char *newname, int ret) "oldname %s newname %s ret %d"
> +cephfs_unlinkat_return(char *path, int stmode, int ret) "path %s stmode %d ret %d"
> +cephfs_init(char *path) "path %s"
> +
>  # target-sparc/mmu_helper.c
>  mmu_helper_dfault(uint64_t address, uint64_t context, int mmu_idx, uint32_t tl) "DFAULT at %"PRIx64" context %"PRIx64" mmu_idx=%d tl=%d"
>  mmu_helper_dprot(uint64_t address, uint64_t context, int mmu_idx, uint32_t tl) "DPROT at %"PRIx64" context %"PRIx64" mmu_idx=%d tl=%d"
Eric Blake March 9, 2016, 8:09 p.m. UTC | #4
On 03/09/2016 12:02 PM, Greg Kurz wrote:
> On Wed,  2 Mar 2016 23:41:43 +0800
> Jevon Qiao <scaleqiao@gmail.com> wrote:
> 

>> +}
>> +
>> +static int cephfs_readdir_r(FsContext *ctx, V9fsFidOpenState *fs,
>> +                            struct dirent *entry,
>> +                            struct dirent **result)
>> +{
>> +    int ret;

>> +    
>> +    return ret;
> 
> This function should behave like the original readdir_r() function from the
> C library, but it doesn't.
> 

readdir_r() is hopelessly broken.  POSIX is withdrawing it as such.
http://austingroupbugs.net/view.php?id=696

readdir() should be all the more any sane program needs, because it
should already be thread-safe.

> According to the the libcephfs.h header:
> 
>  * @returns 1 if the next entry was filled in, 0 if the end of the directory stream was reached,
>  *          and a negative error code on failure.
>  */
> int ceph_readdir_r(struct ceph_mount_info *cmount, struct ceph_dir_result *dirp, struct dirent *de);
> 
> and the readdir_r() manual page says:
> 
>    The  readdir_r() function returns 0 on success.  On error, it returns a
>    positive error number (listed under ERRORS).  If the end of the  direc?
>    tory  stream  is  reached,  readdir_r()  returns 0, and returns NULL in
>    *result.

readdir_r() can silently overflow buffers, with no recourse.  Its use
should not be encouraged.
Greg Kurz March 10, 2016, 9:08 a.m. UTC | #5
On Wed, 9 Mar 2016 13:09:58 -0700
Eric Blake <eblake@redhat.com> wrote:

> On 03/09/2016 12:02 PM, Greg Kurz wrote:
> > On Wed,  2 Mar 2016 23:41:43 +0800
> > Jevon Qiao <scaleqiao@gmail.com> wrote:
> >   
> 
> >> +}
> >> +
> >> +static int cephfs_readdir_r(FsContext *ctx, V9fsFidOpenState *fs,
> >> +                            struct dirent *entry,
> >> +                            struct dirent **result)
> >> +{
> >> +    int ret;  
> 
> >> +    
> >> +    return ret;  
> > 
> > This function should behave like the original readdir_r() function from the
> > C library, but it doesn't.
> >   
> 
> readdir_r() is hopelessly broken.  POSIX is withdrawing it as such.
> http://austingroupbugs.net/view.php?id=696
> 
> readdir() should be all the more any sane program needs, because it
> should already be thread-safe.
> 

I wasn't aware that readdir_r() was so badly broken. It is currently
used here:

hw/9pfs/9p-handle.c:    return readdir_r(fs->dir, entry, result);
hw/9pfs/9p-local.c:    ret = readdir_r(fs->dir, entry, result);
hw/9pfs/9p-proxy.c:    return readdir_r(fs->dir, entry, result);

I'll see how we can move to readdir().

> > According to the the libcephfs.h header:
> > 
> >  * @returns 1 if the next entry was filled in, 0 if the end of the directory stream was reached,
> >  *          and a negative error code on failure.
> >  */
> > int ceph_readdir_r(struct ceph_mount_info *cmount, struct ceph_dir_result *dirp, struct dirent *de);
> > 
> > and the readdir_r() manual page says:
> > 
> >    The  readdir_r() function returns 0 on success.  On error, it returns a
> >    positive error number (listed under ERRORS).  If the end of the  direc?
> >    tory  stream  is  reached,  readdir_r()  returns 0, and returns NULL in
> >    *result.  
> 
> readdir_r() can silently overflow buffers, with no recourse.  Its use
> should not be encouraged.
> 

Sure... this being said, fsdev currently exposes a readdir_r operation, not
a readdir one. Until we decide to change that, backends must follow the
readdir_r() API.

Cheers.

--
Greg

--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Jevon Qiao March 14, 2016, 2:02 a.m. UTC | #6
Hi Greg,

Thank you very much for reviewing the code, I've revised the code per 
your comments, will post it later. Meanwhile, please see my answers to 
your questions below.
>
>> +    }
>> +    fs->dir = (DIR *)result;
>> +    if (!fs->dir) {
>> +        fprintf(stderr, "ceph_opendir return NULL for ceph_dir_result\n");
> Hmm... is this a message for the QEMU user, so that she can fix something and
> retry ? I suspect it is more for debugging purposes, in which case I'd rather
> add a result argument to trace_cephfs_opendir_return() above.
Actually, this message is mainly used for debugging.
> And BTW, can ceph_opendir() return success without filling the structure ?
I double checked the Ceph code, it can not. I've removed this error message.
>> +
>> +# hw/9pfs/9p-cephfs.c
>> +cephfs_lstat_return(char *path, int stmode, int stuid, int stgid, int stsize, int ret) "path %s stmode %d stuid %d stgid %d stsize %d ret %d"
>> +cephfs_readlink_return(char *path, int ret) "path %s ret %d"
>> +cephfs_open_return(char *path, int flags, int mode, int fd) "path %s flags %d mode %d fd %d"
>> +cephfs_opendir_return(char *path, int ret) "path %s ret %d"
>> +cephfs_rewinddir(void *dir) "dir %p"
>> +cephfs_telldir(void *dir) "dir %p"
>> +cephfs_readdir_r_return(void *tmpent, void *entry, int ret) "tmpent %p entry %p ret %d"
>> +cephfs_seekdir(void *dir, int off) "dir %p off %d"
>> +cephfs_preadv(int iovcnt, int len) "iovcnt %d len %d"
>> +cephfs_preadv_return(int iovcnt, int len, long ret) "iovcnt %d len %d ret %l"
> ret %ld
>
>> +cephfs_pwritev(int iovcnt, int len, int offset) "iovcnt %d len %d offset %d"
>> +cephfs_pwritev_return(int iovcnt, int len, int offset, long ret) "iovcnt %d len %d offset %d ret %l"cephfs_chmod(char *path, int fcmode) "path %s fcmode %d"
> ret %ld for cephfs_pwritev_return() and missing newline before cephfs_chmod()
>
> I'm wondering if this patch was build tested before being posted to the list.
Good catch. Yes, the code had been compiled and tested. However, it 
seems the compiler will not report this error until the trace point is 
used. Since 'cephfs_chmod' is not used in 9p-cephfs.c, so I just forget 
to test it. That's why the error was not found. I've already removed 
this trace point in the revision.

Thanks,
Jevon
>> +cephfs_chmod_return(char *path, int fcmode, int ret) "path %s fcmode %d ret %d"
>> +cephfs_mknod_return(char *path, int fcmode, uint32_t fcrdev, int ret) "path %s fcmode %d fcrdev %u ret %d"
>> +cephfs_mkdir_return(char *path, int fcmode, int ret) " path %s fcmode %d ret %d"
>> +cephfs_fstat_return(int fidtype, int fd, int stuid, int stgid, int stsize, int ret) "fidtype %d fd %d stuid %d stgid %d stsize %d ret %d"
>> +cephfs_open2_return(char *path, int flags, int fcmode) "path %s flags %d fcmode %d"



--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch
diff mbox

diff --git a/configure b/configure
index 0c0472a..a2627be 100755
--- a/configure
+++ b/configure
@@ -275,6 +275,7 @@  trace_backends="log"
 trace_file="trace"
 spice=""
 rbd=""
+cephfs=""
 smartcard=""
 libusb=""
 usb_redir=""
@@ -1019,6 +1020,10 @@  for opt do
   ;;
   --enable-rbd) rbd="yes"
   ;;
+  --disable-cephfs) cephfs="no"
+  ;;
+  --enable-cephfs) cephfs="yes"
+  ;;
   --disable-xfsctl) xfs="no"
   ;;
   --enable-xfsctl) xfs="yes"
@@ -1345,6 +1350,7 @@  disabled with --disable-FEATURE, default is enabled if available:
   vhost-net       vhost-net acceleration support
   spice           spice
   rbd             rados block device (rbd)
+  cephfs          Ceph File System 
   libiscsi        iscsi support
   libnfs          nfs support
   smartcard       smartcard support (libcacard)
@@ -3087,6 +3093,28 @@  EOF
 fi
 
 ##########################################
+# cephfs probe
+if test "$cephfs" != "no" ; then
+  cat > $TMPC <<EOF
+#include <stdio.h>
+#include <cephfs/libcephfs.h>
+int main(void) {
+    struct ceph_mount_info *cmount;
+    ceph_create(&cmount, NULL);
+    return 0;
+}
+EOF
+  cephfs_libs="-lcephfs -lrados"
+  if compile_prog "" "$cephfs_libs" ; then
+    cephfs=yes
+  else
+    if test "$cephfs" = "yes" ; then
+      feature_not_found "cephfs" "Install libcephfs/ceph devel"
+    fi
+    cephfs=no
+  fi
+fi
+##########################################
 # libssh2 probe
 min_libssh2_version=1.2.8
 if test "$libssh2" != "no" ; then
@@ -4760,6 +4788,7 @@  else
 echo "spice support     $spice"
 fi
 echo "rbd support       $rbd"
+echo "cephfs support    $cephfs"
 echo "xfsctl support    $xfs"
 echo "smartcard support $smartcard"
 echo "libusb            $libusb"
@@ -5224,6 +5253,10 @@  if test "$rbd" = "yes" ; then
   echo "RBD_CFLAGS=$rbd_cflags" >> $config_host_mak
   echo "RBD_LIBS=$rbd_libs" >> $config_host_mak
 fi
+if test "$cephfs" = "yes" ; then
+  echo "CONFIG_CEPHFS=m" >> $config_host_mak
+  echo "CEPHFS_LIBS=$cephfs_libs" >> $config_host_mak
+fi
 
 echo "CONFIG_COROUTINE_BACKEND=$coroutine" >> $config_host_mak
 if test "$coroutine_pool" = "yes" ; then
diff --git a/fsdev/qemu-fsdev.c b/fsdev/qemu-fsdev.c
index bf7f0b0..7f07a2a 100644
--- a/fsdev/qemu-fsdev.c
+++ b/fsdev/qemu-fsdev.c
@@ -27,6 +27,7 @@  static FsDriverTable FsDrivers[] = {
 #endif
     { .name = "synth", .ops = &synth_ops},
     { .name = "proxy", .ops = &proxy_ops},
+    { .name = "cephfs", .ops = &cephfs_ops},
 };
 
 int qemu_fsdev_add(QemuOpts *opts)
diff --git a/fsdev/qemu-fsdev.h b/fsdev/qemu-fsdev.h
index 9fa45bf..86a17b8 100644
--- a/fsdev/qemu-fsdev.h
+++ b/fsdev/qemu-fsdev.h
@@ -22,7 +22,7 @@ 
  * fstype | ops
  * -----------------
  *  local | local_ops
- *  .     |
+ *  cephfs| cephfs_ops
  *  .     |
  *  .     |
  *  .     |
@@ -45,4 +45,5 @@  extern FileOperations local_ops;
 extern FileOperations handle_ops;
 extern FileOperations synth_ops;
 extern FileOperations proxy_ops;
+extern FileOperations cephfs_ops;
 #endif
diff --git a/hw/9pfs/9p-cephfs.c b/hw/9pfs/9p-cephfs.c
new file mode 100644
index 0000000..f18ec89
--- /dev/null
+++ b/hw/9pfs/9p-cephfs.c
@@ -0,0 +1,739 @@ 
+/*
+ * Virtio 9p cephfs callback
+ *
+ * Copyright UnitedStack, Corp. 2016
+ *
+ * Authors:
+ *    Jevon Qiao <scaleqiao@gmail.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 "qemu/iov.h"
+#include "9p.h"
+#include "9p-xattr.h"
+#include "trace.h"
+#include <cephfs/libcephfs.h>
+#include "fsdev/qemu-fsdev.h"   /* cephfs_ops */
+#include <arpa/inet.h>
+#include <pwd.h>
+#include <grp.h>
+#include <sys/socket.h>
+#include <sys/un.h>
+#include "qemu/xattr.h"
+#include "qemu/error-report.h"
+#include <libgen.h>
+#include <unistd.h>
+#include <linux/fs.h>
+#ifdef CONFIG_LINUX_MAGIC_H
+#include <linux/magic.h>
+#endif
+#include <sys/ioctl.h>
+
+#define CEPH_VER_LEN        32
+#define MON_NAME_LEN        32
+#define MON_SECRET_LEN      64
+
+#ifndef LIBCEPHFS_VERSION
+#define LIBCEPHFS_VERSION(maj, min, extra) ((maj << 16) + (min << 8) + extra)
+#define LIBCEPHFS_VERSION_CODE LIBCEPHFS_VERSION(0, 0, 0)
+#endif
+
+struct cephfs_data {
+    int	major, minor, patch;
+    char ceph_version[CEPH_VER_LEN];
+    struct  ceph_mount_info *cmount;
+};
+
+/*
+ * Helper function for cephfs_preadv and cephfs_pwritev
+ */
+inline static ssize_t preadv_pwritev(struct ceph_mount_info *cmount, int fd,
+                              	    const struct iovec *iov, int iov_cnt,
+                              	    off_t offset, bool do_write)
+{
+    ssize_t ret = 0;
+    size_t i = 0;
+    size_t len = 0;
+    void *buf, *buftmp;
+    size_t bufoffset = 0;
+
+    len = iov_size(iov, iov_cnt);
+    buf = g_new0(uint8_t, len);
+    buftmp = buf;
+    if (do_write) {
+        for (i = 0; i < iov_cnt; i++) {
+            memcpy((buftmp + bufoffset), iov[i].iov_base, iov[i].iov_len);
+            bufoffset += iov[i].iov_len;
+        }
+        ret = ceph_write(cmount, fd, buf, len, offset);
+        if (ret <= 0) {
+           errno = -ret;
+           ret = -1;
+        }
+    } else {
+        ret = ceph_read(cmount, fd, buf, len, offset);
+        if (ret <= 0) {
+            errno = -ret;
+            ret = -1;
+        } else {
+            for (i = 0; i < iov_cnt; i++) {
+                memcpy(iov[i].iov_base, (buftmp + bufoffset), iov[i].iov_len);
+                bufoffset += iov[i].iov_len;
+            }
+        }
+    }
+
+    free(buf);
+    return ret;
+}
+
+static int cephfs_update_file_cred(struct ceph_mount_info *cmount,
+				   const char *name, FsCred *credp)
+{
+    int fd, ret;
+    fd = ceph_open(cmount, name, O_NONBLOCK | O_NOFOLLOW, credp->fc_mode);
+    if (fd < 0) {
+        return fd;
+    }
+    ret = ceph_fchown(cmount, fd, credp->fc_uid, credp->fc_gid);
+    if (ret < 0) {
+        goto err_out;
+    }
+    ret = ceph_fchmod(cmount, fd, credp->fc_mode & 07777);
+err_out:
+    close(fd);
+    return ret;
+}
+
+static int cephfs_lstat(FsContext *fs_ctx, V9fsPath *fs_path,
+                        struct stat *stbuf)
+{
+    int ret;
+    char *path = fs_path->data;
+    struct cephfs_data *cfsdata = fs_ctx->private;
+
+    ret = ceph_lstat(cfsdata->cmount, path, stbuf);
+    trace_cephfs_lstat_return(path, stbuf->st_mode, stbuf->st_uid, stbuf->st_gid, stbuf->st_size, ret);
+    if (ret){
+        errno = -ret; 
+        ret = -1;
+    }
+    return ret;
+}
+
+static ssize_t cephfs_readlink(FsContext *fs_ctx, V9fsPath *fs_path,
+                               char *buf, size_t bufsz)
+{
+    int ret;
+    char *path = fs_path->data;
+    struct cephfs_data *cfsdata = fs_ctx->private;
+
+    ret = ceph_readlink(cfsdata->cmount, path, buf, bufsz);
+    trace_cephfs_readlink_return(path, ret);
+    return ret;
+}
+
+static int cephfs_close(FsContext *ctx, V9fsFidOpenState *fs)
+{
+    struct cephfs_data *cfsdata = ctx->private;
+
+    return ceph_close(cfsdata->cmount, fs->fd);
+}
+
+static int cephfs_closedir(FsContext *ctx, V9fsFidOpenState *fs)
+{
+    struct cephfs_data *cfsdata = ctx->private;
+   
+    return ceph_closedir(cfsdata->cmount, (struct ceph_dir_result *)fs->dir);
+}
+
+static int cephfs_open(FsContext *ctx, V9fsPath *fs_path,
+                       int flags, V9fsFidOpenState *fs)
+{
+    struct cephfs_data *cfsdata = ctx->private;
+
+    fs->fd = ceph_open(cfsdata->cmount, fs_path->data, flags, 0777);
+    trace_cephfs_open_return(fs_path->data, flags, 0777, fs->fd);
+    return fs->fd;
+}
+
+static int cephfs_opendir(FsContext *ctx,
+                          V9fsPath *fs_path, V9fsFidOpenState *fs)
+{
+    int ret;
+    struct ceph_dir_result *result;
+    struct cephfs_data *cfsdata = ctx->private;
+    char *path = fs_path->data;
+  
+    ret = ceph_opendir(cfsdata->cmount, path, &result);
+    trace_cephfs_opendir_return(path, ret);
+    if (ret) {
+        fprintf(stderr, "ceph_opendir=%d\n", ret);
+        return ret;
+    }
+    fs->dir = (DIR *)result;
+    if (!fs->dir) {
+        fprintf(stderr, "ceph_opendir return NULL for ceph_dir_result\n");
+        return -1;
+    }
+    return 0;
+}
+ 
+static void cephfs_rewinddir(FsContext *ctx, V9fsFidOpenState *fs)
+{
+    struct cephfs_data *cfsdata = ctx->private;
+
+    trace_cephfs_rewinddir(fs->dir);
+    return ceph_rewinddir(cfsdata->cmount, (struct ceph_dir_result *)fs->dir);
+}
+
+static off_t cephfs_telldir(FsContext *ctx, V9fsFidOpenState *fs)
+{
+    int ret;
+    struct cephfs_data *cfsdata = ctx->private;
+
+    trace_cephfs_telldir(fs->dir);
+    ret = ceph_telldir(cfsdata->cmount, (struct ceph_dir_result *)fs->dir);
+    return ret;
+}
+
+static int cephfs_readdir_r(FsContext *ctx, V9fsFidOpenState *fs,
+                            struct dirent *entry,
+                            struct dirent **result)
+{
+    int ret;
+    struct dirent *tmpent;
+    struct cephfs_data *cfsdata = ctx->private;
+
+    tmpent = entry;
+    ret = ceph_readdir_r(cfsdata->cmount, (struct ceph_dir_result *)fs->dir,
+		    	entry);
+    trace_cephfs_readdir_r_return(tmpent, entry, ret);
+    if (ret > 0 && entry != NULL)
+    {
+        *result = entry;
+    } else if (!ret)
+    {
+        *result = NULL;
+        entry = tmpent;
+    }
+    
+    return ret;
+}
+
+static void cephfs_seekdir(FsContext *ctx, V9fsFidOpenState *fs, off_t off)
+{
+    struct cephfs_data *cfsdata = ctx->private;
+
+    trace_cephfs_seekdir(fs->dir, off);
+    return ceph_seekdir(cfsdata->cmount, (struct ceph_dir_result*)fs->dir, off);
+}
+
+static ssize_t cephfs_preadv(FsContext *ctx, V9fsFidOpenState *fs,
+                             const struct iovec *iov,
+                             int iovcnt, off_t offset)
+{
+    ssize_t ret = 0;
+    struct cephfs_data *cfsdata = ctx->private;
+
+    trace_cephfs_preadv(iovcnt, iov_size(iov, iovcnt));
+#if defined(LIBCEPHFS_VERSION) && LIBCEPHFS_VERSION_CODE >= LIBCEPHFS_VERSION(9, 0, 3) 
+    ret = ceph_preadv(cfsdata->cmount, fs->fd, iov, iovcnt, offset); 
+#else
+    if (iovcnt > 1) {
+	ret = preadv_pwritev(cfsdata->cmount, fs->fd, iov, iovcnt, offset, 0);
+    } else if (iovcnt > 0) {
+	ret = ceph_read(cfsdata->cmount, fs->fd, iov[0].iov_base,
+			iov[0].iov_len, offset);
+    }
+#endif
+    trace_cephfs_preadv_return(iovcnt, iov_size(iov, iovcnt), ret);
+
+    return ret;
+}
+
+static ssize_t cephfs_pwritev(FsContext *ctx, V9fsFidOpenState *fs,
+                              const struct iovec *iov,
+                              int iovcnt, off_t offset)
+{
+    ssize_t ret = 0;
+    struct cephfs_data *cfsdata = ctx->private;
+
+    trace_cephfs_pwritev(iovcnt, iov_size(iov, iovcnt), offset);
+#if defined(LIBCEPHFS_VERSION) && LIBCEPHFS_VERSION_CODE >= LIBCEPHFS_VERSION(9, 0, 3) 
+    ret = ceph_pwritev(cfsdata->cmount, fs->fd, iov, iovcnt, offset);
+#else
+    if (iovcnt > 1) {
+	ret = preadv_pwritev(cfsdata->cmount, fs->fd, iov, iovcnt, offset, 1);
+    } else if (iovcnt > 0) {
+	ret = ceph_write(cfsdata->cmount, fs->fd, iov[0].iov_base,
+			iov[0].iov_len, offset);
+    }
+#endif
+    trace_cephfs_pwritev_return(iovcnt, iov_size(iov, iovcnt), offset, ret);
+
+#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 cephfs_chmod(FsContext *fs_ctx, V9fsPath *fs_path, FsCred *credp)
+{
+    int  ret = -1;
+    struct cephfs_data *cfsdata = fs_ctx->private;
+
+    ret = ceph_chmod(cfsdata->cmount, fs_path->data, credp->fc_mode);
+    trace_cephfs_chmod_return(fs_path->data, credp->fc_mode, ret);
+    return ret;
+}
+
+static int cephfs_mknod(FsContext *fs_ctx, V9fsPath *dir_path,
+                       const char *name, FsCred *credp)
+{
+    int ret;
+    V9fsString fullname;
+    struct cephfs_data *cfsdata = fs_ctx->private;
+
+    v9fs_string_init(&fullname);
+    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
+    ret = ceph_mknod(cfsdata->cmount, fullname.data, credp->fc_mode,
+		    credp->fc_rdev);
+    trace_cephfs_mknod_return(fullname.data, credp->fc_mode, credp->fc_rdev, ret);
+
+    v9fs_string_free(&fullname);
+    return ret;
+}
+
+static int cephfs_mkdir(FsContext *fs_ctx, V9fsPath *dir_path,
+                       const char *name, FsCred *credp)
+{
+    int ret;
+    V9fsString fullname;
+    struct cephfs_data *cfsdata = fs_ctx->private;
+
+    v9fs_string_init(&fullname);
+    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
+    ret = ceph_mkdir(cfsdata->cmount, fullname.data, credp->fc_mode);
+    trace_cephfs_mkdir_return(fullname.data, credp->fc_mode, ret);
+
+    v9fs_string_free(&fullname);
+    return ret;
+}
+
+static int cephfs_fstat(FsContext *fs_ctx, int fid_type,
+                        V9fsFidOpenState *fs, struct stat *stbuf)
+{
+    int fd = -1;
+    int ret;
+    struct cephfs_data *cfsdata = fs_ctx->private;
+
+    if (fid_type == P9_FID_DIR) {
+        fd = dirfd(fs->dir);
+    } else {
+        fd = fs->fd;
+    }
+    ret = ceph_fstat(cfsdata->cmount, fd, stbuf);
+    trace_cephfs_fstat_return(fid_type, fd, stbuf->st_uid, stbuf->st_gid, stbuf->st_size, ret);
+
+    return ret;
+}
+
+static int cephfs_open2(FsContext *fs_ctx, V9fsPath *dir_path, const char *name,
+                        int flags, FsCred *credp, V9fsFidOpenState *fs)
+{
+    int fd = -1, ret = -1;
+    V9fsString fullname;
+    struct cephfs_data *cfsdata = fs_ctx->private;
+
+    v9fs_string_init(&fullname);
+    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
+    fd = ceph_open(cfsdata->cmount, fullname.data, flags, credp->fc_mode);
+    trace_cephfs_open2_return(fullname.data, flags, credp->fc_mode);
+    if (fd >= 0) {
+        /* After creating the file, need to set the cred */
+        ret = cephfs_update_file_cred(cfsdata->cmount, name, credp);
+        if (ret < 0) {
+            ceph_close(cfsdata->cmount, fd);
+            errno = -ret;
+            fd = ret;
+        } else {
+            fs->fd = fd;
+        }
+    } else {
+       errno = -fd;
+    }
+
+    v9fs_string_free(&fullname);
+    return fd;
+}
+
+static int cephfs_symlink(FsContext *fs_ctx, const char *oldpath,
+                          V9fsPath *dir_path, const char *name, FsCred *credp)
+{
+    int ret = -1;
+    V9fsString fullname;
+    struct cephfs_data *cfsdata = fs_ctx->private;
+
+    v9fs_string_init(&fullname);
+    v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
+    ret = ceph_symlink(cfsdata->cmount, oldpath, fullname.data);
+    trace_cephfs_symlink_return(oldpath, fullname.data, ret);
+
+    v9fs_string_free(&fullname);
+    return ret;
+}
+
+static int cephfs_link(FsContext *ctx, V9fsPath *oldpath,
+                       V9fsPath *dirpath, const char *name)
+{
+    int ret = -1;
+    V9fsString newpath;
+    struct cephfs_data *cfsdata = ctx->private;
+    
+    v9fs_string_init(&newpath);
+    v9fs_string_sprintf(&newpath, "%s/%s", dirpath->data, name);
+    ret = ceph_link(cfsdata->cmount, oldpath->data, newpath.data);
+    trace_cephfs_link_return(oldpath->data, newpath.data, ret);
+
+    v9fs_string_free(&newpath); 
+    return ret;
+}
+
+static int cephfs_truncate(FsContext *ctx, V9fsPath *fs_path, off_t size)
+{
+    int ret = -1;
+    struct cephfs_data *cfsdata = ctx->private;
+
+    ret = ceph_truncate(cfsdata->cmount, fs_path->data, size);
+    trace_cephfs_truncate_return(fs_path->data, size, ret);
+
+    return ret;
+}
+
+static int cephfs_rename(FsContext *ctx, const char *oldpath,
+                         const char *newpath)
+{
+    int ret = -1;
+    struct cephfs_data *cfsdata = ctx->private;
+
+    ret = ceph_rename(cfsdata->cmount, oldpath, newpath);
+    trace_cephfs_rename_return(oldpath, newpath, ret);
+
+    return ret;
+}
+
+static int cephfs_chown(FsContext *fs_ctx, V9fsPath *fs_path, FsCred *credp)
+{
+    int ret = -1;
+    struct cephfs_data *cfsdata = fs_ctx->private;
+
+    ret = ceph_chown(cfsdata->cmount, fs_path->data, credp->fc_uid,
+		    credp->fc_gid);
+    trace_cephfs_chown_return(fs_path->data, credp->fc_uid, credp->fc_gid, ret);
+
+    return ret;
+}
+
+static int cephfs_utimensat(FsContext *ctx, V9fsPath *fs_path,
+                            const struct timespec *buf)
+{
+    int ret = -1;
+
+#ifdef CONFIG_UTIMENSAT
+    struct cephfs_data *cfsdata = ctx->private;
+
+    ret = ceph_utime(cfsdata->cmount, fs_path->data, (struct utimbuf *)buf);
+    trace_cephfs_utimensat_return(fs_path->data, ret);
+#else
+    ret = -1;
+    errno = ENOSYS;
+#endif
+
+    return ret;
+}
+
+static int cephfs_remove(FsContext *ctx, const char *path)
+{
+    errno = EOPNOTSUPP;
+    return -1;
+}
+
+static int cephfs_fsync(FsContext *ctx, int fid_type,
+                        V9fsFidOpenState *fs, int datasync)
+{
+    int ret = -1, fd = -1;
+    struct cephfs_data *cfsdata = ctx->private;
+
+    if (fid_type == P9_FID_DIR) {
+        fd = dirfd(fs->dir);
+    } else {
+        fd = fs->fd;
+    }
+    ret = ceph_fsync(cfsdata->cmount, fd, datasync);
+    trace_cephfs_fsync_return(fd, datasync, ret);
+
+    return ret;
+}
+
+static int cephfs_statfs(FsContext *ctx, V9fsPath *fs_path,
+                         struct statfs *stbuf)
+{
+    int ret;
+    char *path = fs_path->data;
+    struct cephfs_data *cfsdata = ctx->private;
+
+    ret = ceph_statfs(cfsdata->cmount, path, (struct statvfs*)stbuf);
+    if (ret) {
+        fprintf(stderr, "ceph_statfs=%d\n", ret); 
+    }
+
+    return ret;
+}
+
+/*
+ * Get the extended attribute of normal file, if the path refer to a symbolic
+ * link, just return the extended attributes of the syslink rather than the
+ * attributes of the link itself.
+ */
+static ssize_t cephfs_lgetxattr(FsContext *ctx, V9fsPath *fs_path,
+                                const char *name, void *value, size_t size)
+{
+    int ret;
+    char *path = fs_path->data;
+    struct cephfs_data *cfsdata = ctx->private;
+
+    ret = ceph_lgetxattr(cfsdata->cmount, path, name, value, size);
+    trace_cephfs_lgetxattr_return(path, name, ret);
+
+    return ret;
+}
+
+static ssize_t cephfs_llistxattr(FsContext *ctx, V9fsPath *fs_path,
+                                 void *value, size_t size)
+{
+    int ret = -1;
+    struct cephfs_data *cfsdata = ctx->private;
+
+    ret = ceph_llistxattr(cfsdata->cmount, fs_path->data, value, size);
+    trace_cephfs_llistxattr_return(fs_path->data, ret);
+
+    return ret;
+}
+
+static int cephfs_lsetxattr(FsContext *ctx, V9fsPath *fs_path, const char *name,
+                            void *value, size_t size, int flags)
+{
+    int ret = -1;
+    struct cephfs_data *cfsdata = ctx->private;
+
+    ret = ceph_lsetxattr(cfsdata->cmount, fs_path->data, name, value, size,
+	flags);
+    trace_cephfs_lsetxattr_return(fs_path->data, name, flags, ret);
+
+    return ret;
+}
+
+static int cephfs_lremovexattr(FsContext *ctx, V9fsPath *fs_path,
+                               const char *name)
+{
+    int ret = -1;
+    struct cephfs_data *cfsdata = ctx->private;
+
+    ret = ceph_lremovexattr(cfsdata->cmount, fs_path->data, name);
+    trace_cephfs_lremovexattr_return(fs_path->data, name, ret);
+
+    return ret;
+}
+
+static int cephfs_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 {
+        /* if the path does not start from '/' */
+        v9fs_string_sprintf((V9fsString *)target, "%s", name);
+    }
+
+    /* Bump the size for including terminating NULL */ 
+    target->size++;
+    return 0;
+}
+
+static int cephfs_renameat(FsContext *ctx, V9fsPath *olddir,
+                           const char *old_name, V9fsPath *newdir,
+                           const char *new_name)
+{
+    int ret = -1;
+    struct cephfs_data *cfsdata = ctx->private;
+    
+    ret = ceph_rename(cfsdata->cmount, old_name, new_name);
+    trace_cephfs_renameat_return(old_name, new_name, ret);
+
+    return ret;
+}
+
+static int cephfs_unlinkat(FsContext *ctx, V9fsPath *dir,
+                           const char *name, int flags)
+{
+    int ret = 0;
+    char *path = dir->data;
+    struct stat fstat;
+    V9fsString fullname;
+    struct cephfs_data *cfsdata = ctx->private;
+
+    v9fs_string_init(&fullname);
+    v9fs_string_sprintf(&fullname, "%s/%s", dir->data, name);
+    path = fullname.data;
+    /* determine which kind of file is being destroyed */ 
+    ret = ceph_lstat(cfsdata->cmount, path, &fstat);
+    if (!ret) {
+        switch (fstat.st_mode & S_IFMT) {
+        case S_IFDIR:
+            ret = ceph_rmdir(cfsdata->cmount, path);
+            break;
+
+        case S_IFBLK:
+        case S_IFCHR:
+        case S_IFIFO:
+        case S_IFLNK:
+        case S_IFREG:
+        case S_IFSOCK:
+            ret = ceph_unlink(cfsdata->cmount, path);
+            break;
+
+        default:
+            fprintf(stderr, "ceph_lstat unknown stmode\n");
+            break;
+        }
+    } else {
+        errno = -ret;
+        ret = -1;
+    }
+    trace_cephfs_unlinkat_return(path, fstat.st_mode, ret);
+
+    v9fs_string_free(&fullname);
+    return ret;
+}
+
+/*
+ * Do two things in the init function:
+ * 1) Create a mount handle used by all cephfs interfaces.
+ * 2) Invoke ceph_mount() to initialize a link between the client and 
+ *    ceph monitor
+ */
+static int cephfs_init(FsContext *ctx)
+{
+    int ret;
+    const char *ver = NULL;
+    struct cephfs_data *data = g_malloc(sizeof(struct cephfs_data));
+
+    if (data == NULL) {
+	errno = ENOMEM;
+	return -1;
+    }
+    trace_cephfs_init(ctx->fs_root);
+    memset(data, 0, sizeof(struct cephfs_data));
+    ret = ceph_create(&data->cmount, NULL);
+    if (ret) {
+        fprintf(stderr, "ceph_create=%d\n", ret);
+        goto err_out;
+    }
+
+    ret = ceph_conf_read_file(data->cmount, NULL);
+    if (ret) {
+        fprintf(stderr, "ceph_conf_read_file=%d\n", ret);
+        goto err_out;
+    }
+
+    ret = ceph_mount(data->cmount, ctx->fs_root);
+    if (ret) {
+        fprintf(stderr, "ceph_mount=%d\n", ret);
+        goto err_out;
+    } else {
+        ctx->private = data;
+	/* CephFS does not support FS_IOC_GETVERSIO */ 
+	ctx->exops.get_st_gen = NULL;
+        goto out;
+    }
+
+    ver = ceph_version(&data->major, &data->minor, &data->patch);
+    memcpy(data->ceph_version, ver, strlen(ver) + 1);
+    
+err_out:
+    g_free(data);
+out:
+    return ret;
+}
+
+static int cephfs_parse_opts(QemuOpts *opts, struct FsDriverEntry *fse)
+{
+    const char *sec_model = qemu_opt_get(opts, "security_model");
+    const char *path = qemu_opt_get(opts, "path");
+
+    if (!sec_model) {
+        fprintf(stderr, "Invalid argument security_model specified with "
+		"cephfs fsdriver\n");
+        return -1;
+    }
+
+    if (!path) {
+        fprintf(stderr, "fsdev: No path specified.\n");
+        return -1;
+    }
+
+    fse->path = g_strdup(path);
+    return 0;
+}
+
+FileOperations cephfs_ops = {
+    .parse_opts   = cephfs_parse_opts,
+    .init         = cephfs_init,
+    .lstat        = cephfs_lstat,
+    .readlink     = cephfs_readlink,
+    .close        = cephfs_close,
+    .closedir     = cephfs_closedir,
+    .open         = cephfs_open,
+    .opendir      = cephfs_opendir,
+    .rewinddir    = cephfs_rewinddir,
+    .telldir      = cephfs_telldir,
+    .readdir_r    = cephfs_readdir_r,
+    .seekdir      = cephfs_seekdir,
+    .preadv       = cephfs_preadv,
+    .pwritev      = cephfs_pwritev,
+    .chmod        = cephfs_chmod,
+    .mknod        = cephfs_mknod,
+    .mkdir        = cephfs_mkdir,
+    .fstat        = cephfs_fstat,
+    .open2        = cephfs_open2,
+    .symlink      = cephfs_symlink,
+    .link         = cephfs_link,
+    .truncate     = cephfs_truncate,
+    .rename       = cephfs_rename,
+    .chown        = cephfs_chown,
+    .utimensat    = cephfs_utimensat,
+    .remove       = cephfs_remove,
+    .fsync        = cephfs_fsync,
+    .statfs       = cephfs_statfs,
+    .lgetxattr    = cephfs_lgetxattr,
+    .llistxattr   = cephfs_llistxattr,
+    .lsetxattr    = cephfs_lsetxattr,
+    .lremovexattr = cephfs_lremovexattr,
+    .name_to_path = cephfs_name_to_path,
+    .renameat     = cephfs_renameat,
+    .unlinkat     = cephfs_unlinkat,
+};
diff --git a/hw/9pfs/Makefile.objs b/hw/9pfs/Makefile.objs
index da0ae0c..a77a6f4 100644
--- a/hw/9pfs/Makefile.objs
+++ b/hw/9pfs/Makefile.objs
@@ -5,5 +5,8 @@  common-obj-y += coth.o cofs.o codir.o cofile.o
 common-obj-y += coxattr.o 9p-synth.o
 common-obj-$(CONFIG_OPEN_BY_HANDLE) +=  9p-handle.o
 common-obj-y += 9p-proxy.o
+common-obj-y += 9p-cephfs.o
 
 obj-y += virtio-9p-device.o
+
+9p-cephfs.o-libs := $(CEPHFS_LIBS)
diff --git a/scripts/analyse-9p-simpletrace.py b/scripts/analyse-9p-simpletrace.py
index 3c3dee4..fe0a496 100755
--- a/scripts/analyse-9p-simpletrace.py
+++ b/scripts/analyse-9p-simpletrace.py
@@ -210,4 +210,100 @@  class VirtFSRequestTracker(simpletrace.Analyzer):
         def v9fs_readlink_return(self, tag, id, target):
                 print "RREADLINK (tag =", tag, ", target =", target, ")"
 
+	def cephfs_lstat_return(self, path, stmode, stuid, stgid, stsize, ret):
+		print "RCEPHFSLSTAT (path =", path, ", stmode =", stmode, ", stuid =", stuid, ", stgid =", stgid, ", stsize =", stsize, ", ret =", ret, ")"
+
+	def cephfs_readlink_return(self, path, ret):
+		print "RCEPHFSREADLINK (path =", path, ", ret =", ret, ")"
+
+	def cephfs_open_return(self, path, flags, mode, fd):
+		print "RCEPHFSOPEN (path =", path, ", flags =", flags, ", mode =", mode, ", fd =", fd, ")"
+
+	def cephfs_opendir_return(self, path, ret):
+		print "RCEPHFSOPENDIR (path =", path, ", ret =", ret, ")"
+
+	def cephfs_rewinddir(self, dir):
+		print "TCEPHFSREWINDDIR (dir =", dir, ")"
+
+	def cephfs_telldir(self, dir):
+		print "TCEPHFSTELLDIR (dir =", dir, ")"
+
+	def cephfs_readdir_r_return(self, tmpent, entry, ret):
+		print "RCEPHFSREADDIRR (tmpent =", tmpent, ", entry =", entry, ", ret =", ret, ")"
+
+	def cephfs_seekdir(self, dir, off):
+		print "TCEPHFSSEEKDIR (dir =", dir, ", off =", off, ")"
+
+	def cephfs_preadv(self, iovcnt, len):
+		print "TCEPHFSPREADV (iovcnt=", iovcnt, ", len =", len, ")"
+
+	def cephfs_preadv_return(self, iovcnt, len, ret):
+		print "RCEPHFSPREADV (iovcnt=", iovcnt, ", len =", len, ", ret = ", ret, ")"
+
+	def cephfs_pwritev(self, iovcnt, len, offset):
+		print "TCEPHFSPWRITEV (iovcnt=", iovcnt, ", len =", len, ", offset =", offset, ")"
+
+	def cephfs_pwritev_return(self, iovcnt, len, offset, ret):
+		print "RCEPHFSPWRITEV (iovcnt=", iovcnt, ", len =", len, ", offset =", offset, ", ret = ", ret, ")"
+
+	def cephfs_chmod(self, path, fcmode):
+		print "TCEPHFSCHMOD (path =", path, ", fcmode =", fcmode, ")"
+
+	def cephfs_chmod_return(self, path, fcmode, ret):
+		print "RCEPHFSCHMOD (path =", path, ", fcmode =", fcmode, ", ret =", ret, ")"
+
+	def cephfs_mknod_return(self, path, fcmode, fcrdev, ret):
+		print "RCEPHFSMKNOD (path =", path, ", fcmode =", fcmode, ", fcrdev =", fcrdev, ", ret =", ret, ")"
+
+	def cephfs_mkdir_return(self, path, fcmode, ret):
+		print "RCEPHFSMKDIR (path =", path, ", fcmode =", fcmode, ", ret =", ret, ")"
+
+	def cephfs_fstat_return(self, fidtype, fd, stuid, stgid, stsize, ret):
+		print "RCEPHFSFSTAT (fidtype =", fidtype, ", fd =", fd, ", stuid =", stuid, ", stgid =", stgid, ", stsize =", stsize, ", ret =", ret, ")"
+
+	def cephfs_open2_return(self, path, flags, fcmode):
+		print "RCEPHFSOPEN2 (path =", path, ", flags =", flags, "fcmode =", fcmode, ")"
+
+	def cephfs_symlink_return(self, oldpath, path, ret):
+		print "RCEPHFSSYMLINK (oldpath =", oldpath, ", path =", path, ", ret =", ret, ")"
+
+	def cephfs_link_return(self, oldpath, path, ret):
+		print "RCEPHFSLINK (oldpath =", oldpath, ", path =", path, ", ret =", ret, ")"
+
+	def cephfs_truncate_return(self, path, size, ret):
+		print "RCEPHFSTRUNCATE (path =", path, ", size =", size, ", ret =", ret, ")"
+
+	def cephfs_rename_return(self, oldpath, newpath, ret):
+		print "RCEPHFSRENAME (oldpath =", oldpath, ", newpath =", newpath, ", ret =", ret, ")"
+
+	def cephfs_chown_return(self, path, fcuid, fcgid, ret):
+		print "RCEPHFSCHOWN (path =", path, ", fcuid =", fcuid, ", fcgid =", fcgid, ", ret =", ret, ")"
+
+	def cephfs_utimensat_return(self, path, ret):
+		print "RCEPHFSUTIMENSAT (path =", path, ", ret =", ret, ")"
+
+	def cephfs_fsync_return(self, fd, datasync, ret):
+		print "RCEPHFSFSYNC (fd =", fd, ", datasync =", datasync, ", ret =", ret, ")"
+
+	def cephfs_lgetxattr_return(self, path, name, ret):
+		print "RCEPHFSLGETXATTR (path =", path, ", name =", name, ", ret =", ret, ")"
+
+	def cephfs_llistxattr_return(self, path, ret):
+		print "RCEPHFSLLISTXATTR (path =", path, ", ret =", ret, ")"
+
+	def cephfs_lsetxattr_return(self, path, name, flags, ret):
+		print "RCEPHFSLSETXATTR (path =", path, ", name =", name, ", flags =", flags, ", ret =", ret, ")"
+
+	def cephfs_lremovexattr_return(self, path, name, ret):
+		print "RCEPHFSLREMOVEXATTR (path =", path, ", name =", name, ", ret =", ret, ")"
+
+	def cephfs_renameat_return(self, oldname, newname, ret):
+		print "RCEPHFSRENAMEAT (oldname =", oldname, ", newname =", newname, ", ret =", ret, ")"
+
+	def cephfs_unlinkat_return(self, path, stmode, ret):
+		print "RCEPHFSUNLINKAT (path =", path, ", stmode =", stmode, ", ret =", ret, ")"
+
+	def cephfs_init(self, path):
+		print "RCEPHFSINIT (path =", path, ")"
+
 simpletrace.run(VirtFSRequestTracker())
diff --git a/trace-events b/trace-events
index 6fba6cc..11879d2 100644
--- a/trace-events
+++ b/trace-events
@@ -1118,6 +1118,39 @@  v9fs_xattrcreate(uint16_t tag, uint8_t id, int32_t fid, char* name, int64_t size
 v9fs_readlink(uint16_t tag, uint8_t id, int32_t fid) "tag %d id %d fid %d"
 v9fs_readlink_return(uint16_t tag, uint8_t id, char* target) "tag %d id %d name %s"
 
+# hw/9pfs/9p-cephfs.c 
+cephfs_lstat_return(char *path, int stmode, int stuid, int stgid, int stsize, int ret) "path %s stmode %d stuid %d stgid %d stsize %d ret %d"
+cephfs_readlink_return(char *path, int ret) "path %s ret %d"
+cephfs_open_return(char *path, int flags, int mode, int fd) "path %s flags %d mode %d fd %d"
+cephfs_opendir_return(char *path, int ret) "path %s ret %d"
+cephfs_rewinddir(void *dir) "dir %p"
+cephfs_telldir(void *dir) "dir %p"
+cephfs_readdir_r_return(void *tmpent, void *entry, int ret) "tmpent %p entry %p ret %d"
+cephfs_seekdir(void *dir, int off) "dir %p off %d"
+cephfs_preadv(int iovcnt, int len) "iovcnt %d len %d"
+cephfs_preadv_return(int iovcnt, int len, long ret) "iovcnt %d len %d ret %l"
+cephfs_pwritev(int iovcnt, int len, int offset) "iovcnt %d len %d offset %d"
+cephfs_pwritev_return(int iovcnt, int len, int offset, long ret) "iovcnt %d len %d offset %d ret %l"cephfs_chmod(char *path, int fcmode) "path %s fcmode %d"
+cephfs_chmod_return(char *path, int fcmode, int ret) "path %s fcmode %d ret %d"
+cephfs_mknod_return(char *path, int fcmode, uint32_t fcrdev, int ret) "path %s fcmode %d fcrdev %u ret %d"
+cephfs_mkdir_return(char *path, int fcmode, int ret) " path %s fcmode %d ret %d"
+cephfs_fstat_return(int fidtype, int fd, int stuid, int stgid, int stsize, int ret) "fidtype %d fd %d stuid %d stgid %d stsize %d ret %d"
+cephfs_open2_return(char *path, int flags, int fcmode) "path %s flags %d fcmode %d"
+cephfs_symlink_return(const char *oldpath, char *path, int ret) "oldpath %s path %s ret %d"
+cephfs_link_return(char *oldpath, char *path, int ret) "oldpath %s path %s ret %d"
+cephfs_truncate_return(char *path, int size, int ret) "path %s size %d ret %d"
+cephfs_rename_return(const char *oldpath, const char *newpath, int ret) "oldpath %s newpath %s ret %d"
+cephfs_chown_return(char *path, int fcuid, int fcgid, int ret) "path %s fcuid %d fcgid %d ret %d"
+cephfs_utimensat_return(char *path, int ret) "path %s ret %d"
+cephfs_fsync_return(int fd, int datasync, int ret) "fd %d datasync %d ret %d"
+cephfs_lgetxattr_return(char *path, const char *name, int ret) "path %s name %s ret %d"
+cephfs_llistxattr_return(char *path, int ret) "path %s ret %d"
+cephfs_lsetxattr_return(char *path, const char *name, int flags, int ret) "path %s name %s flags %d ret %d"
+cephfs_lremovexattr_return(char *path, const char *name, int ret) "path %s name %s ret %d"
+cephfs_renameat_return(const char *oldname, const char *newname, int ret) "oldname %s newname %s ret %d"
+cephfs_unlinkat_return(char *path, int stmode, int ret) "path %s stmode %d ret %d"
+cephfs_init(char *path) "path %s"
+
 # target-sparc/mmu_helper.c
 mmu_helper_dfault(uint64_t address, uint64_t context, int mmu_idx, uint32_t tl) "DFAULT at %"PRIx64" context %"PRIx64" mmu_idx=%d tl=%d"
 mmu_helper_dprot(uint64_t address, uint64_t context, int mmu_idx, uint32_t tl) "DPROT at %"PRIx64" context %"PRIx64" mmu_idx=%d tl=%d"