diskimage-builder/diskimage_builder/block_device/level1/partitioning.py
Ian Wienand 5d5fa06e5c Sync after writing partition table
We introduced the "settle" in
I90103b59357edebbac7a641e8980cb282d37561b thinking that maybe kpartx
had not finished writing the partition.  This probably wasn't a bad
first assumption, since we used to have this -- but is seems
insufficient.

The other failiure here seems to be if kpartx hasn't actually seen the
updated partition table in the image, so it has correctly (in it's
mind) not mounted the partition.

Looking at strace of fdisk run manually on a loopback, it will do a
fsync on the raw device after writing and then a global sync as it
exits.

This replicates this; we flush and fsync in mbr.py in the exit handler
after writing the partition, before closing the file (i've updated one
of the unit tests to double-check the call).  In the partitioning.py
caller we execute a sync call too.

Since it does seem unlikely the "-s" option of kpartx is not working,
I've removed the udev settle work-around too.

Change-Id: Ia77a0ffe4c76854b326ed76490479d9c691b49aa
Partial-Bug: #1698337
2017-06-19 17:13:36 +10:00

171 lines
7 KiB
Python

# Copyright 2016 Andreas Florath (andreas@florath.net)
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.
import logging
import os
from diskimage_builder.block_device.exception import \
BlockDeviceSetupException
from diskimage_builder.block_device.level1.mbr import MBR
from diskimage_builder.block_device.level1.partition import PartitionNode
from diskimage_builder.block_device.plugin import PluginBase
from diskimage_builder.block_device.utils import exec_sudo
from diskimage_builder.block_device.utils import parse_abs_size_spec
from diskimage_builder.block_device.utils import parse_rel_size_spec
logger = logging.getLogger(__name__)
class Partitioning(PluginBase):
def __init__(self, config, default_config, state):
logger.debug("Creating Partitioning object; config [%s]", config)
super(Partitioning, self).__init__()
# Unlike other PluginBase we are somewhat persistent, as the
# partition nodes call back to us (see create() below). We
# need to keep this reference.
self.state = state
# Because using multiple partitions of one base is done
# within one object, there is the need to store a flag if the
# creation of the partitions was already done.
self.already_created = False
self.already_cleaned = False
# Parameter check
if 'base' not in config:
raise BlockDeviceSetupException("Partitioning config needs 'base'")
self.base = config['base']
if 'partitions' not in config:
raise BlockDeviceSetupException(
"Partitioning config needs 'partitions'")
if 'label' not in config:
raise BlockDeviceSetupException(
"Partitioning config needs 'label'")
self.label = config['label']
if self.label not in ("mbr", ):
raise BlockDeviceSetupException("Label must be 'mbr'")
# It is VERY important to get the alignment correct. If this
# is not correct, the disk performance might be very poor.
# Example: In some tests a 'off by one' leads to a write
# performance of 30% compared to a correctly aligned
# partition.
# The problem for DIB is, that it cannot assume that the host
# system uses the same IO sizes as the target system,
# therefore here a fixed approach (as used in all modern
# systems with large disks) is used. The partitions are
# aligned to 1MiB (which are about 2048 times 512 bytes
# blocks)
self.align = 1024 * 1024 # 1MiB as default
if 'align' in config:
self.align = parse_abs_size_spec(config['align'])
self.partitions = []
prev_partition = None
for part_cfg in config['partitions']:
np = PartitionNode(part_cfg, state, self, prev_partition)
self.partitions.append(np)
prev_partition = np
def get_nodes(self):
# return the list of partitions
return self.partitions
def _size_of_block_dev(self, dev):
with open(dev, "r") as fd:
fd.seek(0, 2)
return fd.tell()
# not this is NOT a node and this is not called directly! The
# create() calls in the partition nodes this plugin has
# created are calling back into this.
def create(self):
# This is a bit of a hack. Each of the partitions is actually
# in the graph, so for every partition we get a create() call
# as the walk happens. But we only need to create the
# partition table once...
if self.already_created:
logger.info("Not creating the partitions a second time.")
return
self.already_created = True
# the raw file on disk
image_path = self.state['blockdev'][self.base]['image']
# the /dev/loopX device of the parent
device_path = self.state['blockdev'][self.base]['device']
logger.info("Creating partition on [%s] [%s]", self.base, image_path)
assert self.label == 'mbr'
disk_size = self._size_of_block_dev(image_path)
with MBR(image_path, disk_size, self.align) as part_impl:
for part_cfg in self.partitions:
part_name = part_cfg.get_name()
part_bootflag = PartitionNode.flag_boot \
in part_cfg.get_flags()
part_primary = PartitionNode.flag_primary \
in part_cfg.get_flags()
part_size = part_cfg.get_size()
part_free = part_impl.free()
part_type = part_cfg.get_type()
logger.debug("Not partitioned space [%d]", part_free)
part_size = parse_rel_size_spec(part_size,
part_free)[1]
part_no \
= part_impl.add_partition(part_primary, part_bootflag,
part_size, part_type)
logger.debug("Create partition [%s] [%d]",
part_name, part_no)
# We're going to mount all partitions with kpartx
# below once we're done. So the device this partition
# will be seen at becomes "/dev/mapper/loop0pX"
assert device_path[:5] == "/dev/"
partition_device_name = "/dev/mapper/%sp%d" % \
(device_path[5:], part_no)
self.state['blockdev'][part_name] \
= {'device': partition_device_name}
# "saftey sync" to make sure the partitions are written
exec_sudo(["sync"])
# now all the partitions are created, get device-mapper to
# mount them
if not os.path.exists("/.dockerenv"):
exec_sudo(["kpartx", "-avs", device_path])
else:
# If running inside Docker, make our nodes manually,
# because udev will not be working. kpartx cannot run in
# sync mode in docker.
exec_sudo(["kpartx", "-av", device_path])
exec_sudo(["dmsetup", "--noudevsync", "mknodes"])
return
def cleanup(self):
# remove the partition mappings made for the parent
# block-device by create() above. this is called from the
# child PartitionNode umount/delete/cleanup. Thus every
# partition calls it, but we only want to do it once and our
# gate.
if not self.already_cleaned:
self.already_cleaned = True
exec_sudo(["kpartx", "-d",
self.state['blockdev'][self.base]['device']])