#!/bin/bash

dir=$(dirname "$0")
. "$dir/block-common.sh"

expand_dev() {
  local dev
  case $1 in
  /*)
    dev=$1
    ;;
  *)
    dev=/dev/$1
    ;;
  esac
  echo -n $dev
}


##
# canonicalise_mode mode
#
# Takes the given mode, which may be r, w, ro, rw, w!, or rw!, or variations
# thereof, and canonicalises them to one of
#
#   'r': perform checks for a new read-only mount;
#   'w': perform checks for a read-write mount; or
#   '!': perform no checks at all.
#
canonicalise_mode()
{
  local mode="$1"

  if ! expr index "$mode" 'w' >/dev/null
  then
    echo 'r'
  elif ! expr index "$mode" '!' >/dev/null
  then
    echo 'w'
  else
    echo '!'
  fi
}


##
# check_sharing device mode
#
# Check whether the device requested is already in use.  To use the device in
# read-only mode, it may be in use in read-only mode, but may not be in use in
# read-write anywhere at all.  To use the device in read-write mode, it must
# not be in use anywhere at all.
#
# Prints one of
#
#    'local': the device may not be used because it is mounted in the current
#             (i.e. the privileged domain) in a way incompatible with the
#             requested mode;
#    'guest': the device may not be used because it already mounted by a guest
#             in a way incompatible with the requested mode; or
#    'ok':    the device may be used.
#
check_sharing()
{
  local dev="$1"
  local mode="$2"

  local devmm=$(device_major_minor "$dev")
  local file

  if [ "$mode" = 'w' ]
  then
    toskip="^$"
  else
    toskip="^[^ ]* [^ ]* [^ ]* ro[, ]"
  fi

  for file in $(cat /proc/mounts | grep -v "$toskip" | cut -f 1 -d ' ')
  do
    if [ -e "$file" ]
    then
      local d=$(device_major_minor "$file")

      if [ "$d" = "$devmm" ]
      then
        echo 'local'
        return
      fi
    fi
  done

  local base_path="$XENBUS_BASE_PATH/$XENBUS_TYPE"
  for dom in $(xenstore-list "$base_path")
  do
    for dev in $(xenstore-list "$base_path/$dom")
    do
      d=$(xenstore_read_default "$base_path/$dom/$dev/physical-device" "")

      if [ "$d" = "$devmm" ]
      then
        if [ "$mode" = 'w' ]
        then
          if ! same_vm $dom
          then
            echo 'guest'
            return
          fi
        else
          local m=$(xenstore_read "$base_path/$dom/$dev/mode")
          m=$(canonicalise_mode "$m")

          if [ "$m" = 'w' ]
          then
            if ! same_vm $dom
            then
              echo 'guest'
              return
            fi
          fi
        fi
      fi
    done
  done

  echo 'ok'
}


same_vm()
{
  local otherdom="$1"
  # Note that othervm can be MISSING here, because Xend will be racing with
  # the hotplug scripts -- the entries in /local/domain can be removed by
  # Xend before the hotplug scripts have removed the entry in
  # /local/domain/0/backend/.  In this case, we want to pretend that the
  # VM is the same as FRONTEND_UUID, because that way the 'sharing' will be
  # allowed.
  local othervm=$(xenstore_read_default "/local/domain/$otherdom/vm"         \
                  "$FRONTEND_UUID")

  [ "$FRONTEND_UUID" = "$othervm" ]
}


##
# check_device_sharing dev mode
#
# Perform the sharing check for the given physical device and mode.
#
check_device_sharing()
{
  local dev="$1"
  local mode=$(canonicalise_mode "$2")
  local result

  if [ "x$mode" = 'x!' ]
  then
    return 0
  fi

  result=$(check_sharing "$dev" "$mode")

  if [ "$result" != 'ok' ]
  then
    do_ebusy "Device $dev is mounted " "$mode" "$result"
  fi
}


##
# check_device_sharing file dev mode
#
# Perform the sharing check for the given file mounted through the given
# loopback interface, in the given mode.
#
check_file_sharing()
{
  local file="$1"
  local dev="$2"
  local mode="$3"

  result=$(check_sharing "$dev" "$mode")

  if [ "$result" != 'ok' ]
  then
    do_ebusy "File $file is loopback-mounted through $dev,
which is mounted " "$mode" "$result"
  fi
}


##
# do_ebusy prefix mode result
#
# Helper function for check_device_sharing check_file_sharing, calling ebusy
# with an error message constructed from the given prefix, mode, and result
# from a call to check_sharing.
#
do_ebusy()
{
  local prefix="$1"
  local mode="$2"
  local result="$3"

  if [ "$result" = 'guest' ]
  then
    dom='a guest '
    when='now'
  else
    dom='the privileged '
    when='by a guest'
  fi

  if [ "$mode" = 'w' ]
  then
    m1=''
    m2=''
  else
    m1='read-write '
    m2='read-only '
  fi

  release_lock "block"
  ebusy \
"${prefix}${m1}in ${dom}domain,
and so cannot be mounted ${m2}${when}."
}


t=$(xenstore_read_default "$XENBUS_PATH/type" 'MISSING')

case "$command" in
  add)
    phys=$(xenstore_read_default "$XENBUS_PATH/physical-device" 'MISSING')
    if [ "$phys" != 'MISSING' ]
    then
      # Depending upon the hotplug configuration, it is possible for this
      # script to be called twice, so just bail.
      exit 0
    fi

    if [ -n "$t" ]
    then
      p=$(xenstore_read "$XENBUS_PATH/params")
      mode=$(xenstore_read "$XENBUS_PATH/mode")
    fi

    case $t in 
      phy)
        dev=$(expand_dev $p)
        FRONTEND_ID=$(xenstore_read "$XENBUS_PATH/frontend-id")
        FRONTEND_UUID=$(xenstore_read_default \
            "/local/domain/$FRONTEND_ID/vm" 'unknown')
        claim_lock "block"
        check_device_sharing "$dev" "$mode"
	write_dev "$dev"
        release_lock "block"
	exit 0
	;;

      file)
        # Canonicalise the file, for sharing check comparison, and the mode
        # for ease of use here.
        file=$(readlink -f "$p") || fatal "$p does not exist."
        mode=$(canonicalise_mode "$mode")

        claim_lock "block"

        if [ "$mode" = 'w' ] && ! stat "$file" -c %A | grep -q w
        then
          release_lock "block"
          ebusy \
"File $file is read-only, and so I will not
mount it read-write in a guest domain."
        fi

        loopdev=''
        for dev in /dev/loop*
        do
          if [ ! -b "$dev" ]
          then
            continue
          fi

          f=$(losetup "$dev" 2>/dev/null) || f=''

          if [ "$f" ]
          then
            # $dev is in use.  Check sharing.
            if [ "x$mode" = 'x!' ]
            then
              continue
            fi

            f=$(echo "$f" | sed -e 's/.*(\(.*\)).*/\1/g')

            # $f is the filename, as read from losetup, but the loopback
            # driver truncates filenames at 64 characters, so we need to go
            # trawling through the store if it's longer than that.  Truncation
            # is indicated by an asterisk at the end of the filename.
            if expr index "$f" '*' >/dev/null
            then
              found=""
              for dom in $(xenstore-list "$XENBUS_BASE_PATH")
              do
                for domdev in $(xenstore-list "$XENBUS_BASE_PATH/$dom")
                do
                  d=$(xenstore_read_default \
                        "$XENBUS_BASE_PATH/$dom/$domdev/node" "")
                  if [ "$d" = "$dev" ]
                  then
                    f=$(xenstore_read "$XENBUS_BASE_PATH/$dom/$domdev/params")
                    found=1
                    break 2
                  fi
                done
              done

              if [ ! "$found" ]
              then
                # This loopback device is in use by someone else, so skip it.
                log debug "Loopback sharing check skips device $dev."
                continue
              fi
            fi

            # Canonicalise the filename for the comparison.

            # I have seen this readlink fails because the filename given by
            # losetup is only the basename.  This cannot happen when the loop
            # device is set up through this script, because file is
            # canonicalised above, but it may happen when loop devices are set
            # up some other way.  This readlink may also conceivably fail if
            # the file backing this loop device has been removed.

            # For maximum safety, in the case that $f does not resolve, we
            # assume that $file and $f are in the same directory.

            # If you create a loopback filesystem, remove it and continue to
            # run on it, and then create another file with the same name, then
            # this check will block that -- don't do that.

            # If you create loop devices through some other mechanism, use
            # relative filenames, and then use the same filename through this
            # script, then this check will block that -- don't do that either.

            f=$(readlink -f "$f" || echo $(dirname "$file")/$(basename "$f"))


            if [ "$f" = "$file" ]
            then
              check_file_sharing "$file" "$dev" "$mode"
            fi
          else
            # $dev is not in use, so we'll remember it for use later; we want
            # to finish the sharing check first.

            if [ "$loopdev" = '' ]
            then
              loopdev="$dev"
            fi
          fi
        done

        if [ "$loopdev" = '' ]
        then
          fatal 'Failed to find an unused loop device'
        fi

        do_or_die losetup "$loopdev" "$file"
        xenstore_write "$XENBUS_PATH/node" "$loopdev"
        write_dev "$loopdev"
        release_lock "block"
        exit 0
	;;

      "")
        claim_lock "block"
        success
        release_lock "block"
	;;
    esac
    ;;

  remove)
    case $t in 
      phy)
	exit 0
	;;

      file)
        node=$(xenstore_read "$XENBUS_PATH/node")
	losetup -d "$node"
	exit 0
	;;

      "")
        exit 0
	;;
    esac
    ;;

esac

# If we've reached here, $t is neither phy nor file, so fire a helper script.
[ -x /etc/xen/scripts/block-"$t" ] && \
  /etc/xen/scripts/block-"$t" "$command" $node