Source code for pylorax.imgutils

# - utility functions/classes for building disk images
# Copyright (C) 2011  Red Hat, Inc.
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# GNU General Public License for more details.
# You should have received a copy of the GNU General Public License
# along with this program.  If not, see <>.
# Author(s):  Will Woods <>

import logging
logger = logging.getLogger("pylorax.imgutils")

import os, tempfile
from os.path import join, dirname
from subprocess import Popen, PIPE, CalledProcessError
import sys
import time
import traceback
import multiprocessing
from time import sleep

from pylorax.sysutils import cpfile
from pylorax.executils import execWithRedirect, execWithCapture
from pylorax.executils import runcmd, runcmd_output

######## Functions for making container images (cpio, tar, squashfs) ##########

[docs]def compress(command, rootdir, outfile, compression="xz", compressargs=None): '''Make a compressed archive of the given rootdir. command is a list of the archiver commands to run compression should be "xz", "gzip", "lzma", "bzip2", or None. compressargs will be used on the compression commandline.''' if compressargs is None: compressargs = ["-9"] if compression not in (None, "xz", "gzip", "lzma", "bzip2"): raise ValueError, "Unknown compression type %s" % compression if compression == "xz": compressargs.insert(0, "--check=crc32") if compression is None: compression = "cat" # this is a little silly compressargs = [] # make compression run with multiple threads if possible if compression in ("xz", "lzma"): compressargs.insert(0, "-T%d" % multiprocessing.cpu_count()) logger.debug("find %s -print0 |%s | %s %s > %s", rootdir, " ".join(command), compression, " ".join(compressargs), outfile) find, archive, comp = None, None, None try: find = Popen(["find", ".", "-print0"], stdout=PIPE, cwd=rootdir) archive = Popen(command, stdin=find.stdout, stdout=PIPE, cwd=rootdir) comp = Popen([compression] + compressargs, stdin=archive.stdout, stdout=open(outfile, "wb")) (_stdout, _stderr) = comp.communicate() return comp.returncode except OSError as e: logger.error(e) # Kill off any hanging processes _ = [p.kill() for p in (find, archive, comp) if p] return 1
[docs]def mkcpio(rootdir, outfile, compression="xz", compressargs=None): return compress(["cpio", "--null", "--quiet", "-H", "newc", "-o"], rootdir, outfile, compression, compressargs)
[docs]def mktar(rootdir, outfile, compression="xz", compressargs=None): return compress(["tar", "--no-recursion", "--selinux", "--acls", "--xattrs", "-cf-", "--null", "-T-"], rootdir, outfile, compression, compressargs)
[docs]def mksquashfs(rootdir, outfile, compression="default", compressargs=None): '''Make a squashfs image containing the given rootdir.''' if compressargs is None: compressargs = [] if compression != "default": compressargs = ["-comp", compression] + compressargs return execWithRedirect("mksquashfs", [rootdir, outfile] + compressargs)
[docs]def mkrootfsimg(rootdir, outfile, label, size=2, sysroot=""): """ Make rootfs image from a directory :param str rootdir: Root directory :param str outfile: Path of output image file :param str label: Filesystem label :param int size: Size of the image in GiB, if None computed automatically :param str sysroot: path to system (deployment) root relative to physical root """ if size: fssize = size * (1024*1024*1024) # 2GB sparse file compresses down to nothin' else: fssize = None # Let mkext4img figure out the needed size mkext4img(rootdir, outfile, label=label, size=fssize) # Reset selinux context on new rootfs with LoopDev(outfile) as loopdev: with Mount(loopdev) as mnt: cmd = [ "setfiles", "-e", "/proc", "-e", "/sys", "-e", "/dev", "-e", "/install", "/etc/selinux/targeted/contexts/files/file_contexts", "/"] root = join(mnt, sysroot.lstrip("/")) runcmd(cmd, root=root)
[docs]def mkdiskfsimage(diskimage, fsimage, label="Anaconda"): """ Copy the / partition of a partitioned disk image to an un-partitioned disk image. diskimage is the full path to partitioned disk image with a / fsimage is the full path of the output fs image file label is the label to apply to the image. Defaults to "Anaconda" """ with PartitionMount(diskimage) as img_mount: if not img_mount or not img_mount.mount_dir: return None"Creating fsimage %s", fsimage) mkext4img(img_mount.mount_dir, fsimage, label=label) ######## Utility functions ###############################################
[docs]def mksparse(outfile, size): '''use os.ftruncate to create a sparse file of the given size.''' fobj = open(outfile, "w") os.ftruncate(fobj.fileno(), size)
[docs]def mkqcow2(outfile, size, options=None): '''use qemu-img to create a file of the given size. options is a list of options passed to qemu-img Default format is qcow2, override by passing "-f", fmt in options. ''' options = options or [] if "-f" not in options: options.extend(["-f", "qcow2"]) runcmd(["qemu-img", "create"] + options + [outfile, str(size)])
[docs]def loop_waitfor(loop_dev, outfile): """Make sure the loop device is attached to the outfile. It seems that on rare occasions losetup can return before the /dev/loopX is ready for use, causing problems with mkfs. This tries to make sure that the loop device really is associated with the backing file before continuing. Raise RuntimeError if it isn't setup after 5 tries. """ for _ in xrange(0,5): runcmd(["udevadm", "settle", "--timeout", "300"]) ## XXX Note that losetup --list output can be truncated to 64 bytes in some ## situations. Don't use it to lookup backing file, go the other way ## and lookup the loop for the backing file. See util-linux lib/loopdev.c ## loopcxt_get_backing_file() if get_loop_name(outfile) == os.path.basename(loop_dev): return # If this really is a race, give it some time to settle down time.sleep(1) raise RuntimeError("Unable to setup %s on %s" % (loop_dev, outfile))
[docs]def loop_attach(outfile): """Attach a loop device to the given file. Return the loop device name. On rare occasions it appears that the device never shows up, some experiments seem to indicate that it may be a race with another process using /dev/loop* devices. So we now try 3 times before actually failing. Raises CalledProcessError if losetup fails. """ retries = 0 while True: try: retries += 1 dev = runcmd_output(["losetup", "--find", "--show", outfile]).strip() # Sometimes the loop device isn't ready yet, make extra sure before returning loop_waitfor(dev, outfile) except CalledProcessError: # Problems running losetup are always errors, raise immediately raise except RuntimeError: # Try to setup the loop device 3 times if retries == 3: logger.error("loop_attach failed, retries exhausted.") raise logger.debug("Try %d failed, %s did not appear.", retries, dev) break return dev
[docs]def loop_detach(loopdev): '''Detach the given loop device. Return False on failure.''' return (execWithRedirect("losetup", ["--detach", loopdev]) == 0)
[docs]def get_loop_name(path): '''Return the loop device associated with the path. Raises RuntimeError if more than one loop is associated''' buf = runcmd_output(["losetup", "-j", path]) if len(buf.splitlines()) > 1: # there should never be more than one loop device listed raise RuntimeError("multiple loops associated with %s" % path) name = os.path.basename(buf.split(":")[0]) return name
[docs]def dm_attach(dev, size, name=None): '''Attach a devicemapper device to the given device, with the given size. If name is None, a random name will be chosen. Returns the device name. raises CalledProcessError if dmsetup fails.''' if name is None: name = tempfile.mktemp(prefix="lorax.imgutils.", dir="") runcmd(["dmsetup", "create", name, "--table", "0 %i linear %s 0" % (size/512, dev)]) return name
[docs]def dm_detach(dev): '''Detach the named devicemapper device. Returns False if dmsetup fails.''' dev = dev.replace("/dev/mapper/", "") # strip prefix, if it's there return execWithRedirect("dmsetup", ["remove", dev])
[docs]def mount(dev, opts="", mnt=None): '''Mount the given device at the given mountpoint, using the given opts. opts should be a comma-separated string of mount options. if mnt is none, a temporary directory will be created and its path will be returned. raises CalledProcessError if mount fails.''' if mnt is None: mnt = tempfile.mkdtemp(prefix="lorax.imgutils.") logger.debug("make tmp mountdir %s", mnt) cmd = ["mount"] if opts: cmd += ["-o", opts] cmd += [dev, mnt] runcmd(cmd) return mnt
[docs]def umount(mnt, lazy=False, maxretry=3, retrysleep=1.0): '''Unmount the given mountpoint. If lazy is True, do a lazy umount (-l). If the mount was a temporary dir created by mount, it will be deleted. raises CalledProcessError if umount fails.''' cmd = ["umount"] if lazy: cmd += ["-l"] cmd += [mnt] count = 0 while maxretry > 0: try: rv = runcmd(cmd) except CalledProcessError: count += 1 if count == maxretry: raise logger.warn("failed to unmount %s. retrying (%d/%d)...", mnt, count, maxretry) if logger.getEffectiveLevel() <= logging.DEBUG: fuser = execWithCapture("fuser", ["-vm", mnt]) logger.debug("fuser -vm:\n%s\n", fuser) sleep(retrysleep) else: break if 'lorax.imgutils' in mnt: os.rmdir(mnt) logger.debug("remove tmp mountdir %s", mnt) return (rv == 0)
[docs]def copytree(src, dest, preserve=True): '''Copy a tree of files using cp -a, thus preserving modes, timestamps, links, acls, sparse files, xattrs, selinux contexts, etc. If preserve is False, uses cp -R (useful for modeless filesystems) raises CalledProcessError if copy fails.''' logger.debug("copytree %s %s", src, dest) cp = ["cp", "-a"] if preserve else ["cp", "-R", "-L"] cp += [".", os.path.abspath(dest)] runcmd(cp, cwd=src)
[docs]def do_grafts(grafts, dest, preserve=True): '''Copy each of the items listed in grafts into dest. If the key ends with '/' it's assumed to be a directory which should be created, otherwise just the leading directories will be created.''' for imgpath, filename in grafts.items(): if imgpath[-1] == '/': targetdir = join(dest, imgpath) imgpath = imgpath[:-1] else: targetdir = join(dest, dirname(imgpath)) if not os.path.isdir(targetdir): os.makedirs(targetdir) if os.path.isdir(filename): copytree(filename, join(dest, imgpath), preserve) else: cpfile(filename, join(dest, imgpath))
[docs]def round_to_blocks(size, blocksize): '''If size isn't a multiple of blocksize, round up to the next multiple''' diff = size % blocksize if diff or not size: size += blocksize - diff return size # TODO: move filesystem data outside this function
[docs]def estimate_size(rootdir, graft=None, fstype=None, blocksize=4096, overhead=128): if graft is None: graft = {} getsize = lambda f: os.lstat(f).st_size if fstype == "btrfs": overhead = 64*1024 # don't worry, it's all sparse if fstype == "hfsplus": overhead = 200 # hack to deal with two bootloader copies if fstype in ("vfat", "msdos"): blocksize = 2048 getsize = lambda f: os.stat(f).st_size # no symlinks, count as copies total = overhead*blocksize dirlist = graft.values() if rootdir: dirlist.append(rootdir) for root in dirlist: for top, dirs, files in os.walk(root): for f in files + dirs: total += round_to_blocks(getsize(join(top,f)), blocksize) if fstype == "btrfs": total = max(256*1024*1024, total) # btrfs minimum size: 256MB return total
[docs]def default_image_name(compression, basename): """ Return a default image name with the correct suffix for the compression type. :param str compression: Compression type :param str basename: Base filename :returns: basename with compression suffix If the compression is unknown it defaults to xz """ SUFFIXES = {"xz": ".xz", "gzip": ".gz", "bzip2": ".bz2", "lzma": ".lzma"} return basename + SUFFIXES.get(compression, ".xz") ######## Execution contexts - use with the 'with' statement ##############
[docs]class LoopDev(object): def __init__(self, filename, size=None): self.loopdev = None self.filename = filename if size: mksparse(self.filename, size) def __enter__(self): self.loopdev = loop_attach(self.filename) return self.loopdev def __exit__(self, exc_type, exc_value, exc_tb): loop_detach(self.loopdev)
[docs]class DMDev(object): def __init__(self, dev, size, name=None): (, self.size, = (dev, size, name) self.mapperdev = None def __enter__(self): self.mapperdev = dm_attach(, self.size, return self.mapperdev def __exit__(self, exc_type, exc_value, exc_tb): dm_detach(self.mapperdev)
[docs]class Mount(object): def __init__(self, dev, opts="", mnt=None): (, self.opts, self.mnt) = (dev, opts, mnt) def __enter__(self): self.mnt = mount(, self.opts, self.mnt) return self.mnt def __exit__(self, exc_type, exc_value, exc_tb): umount(self.mnt)
[docs]class PartitionMount(object): """ Mount a partitioned image file using kpartx """ def __init__(self, disk_img, mount_ok=None): """ disk_img is the full path to a partitioned disk image mount_ok is a function that is passed the mount point and returns True if it should be mounted. """ self.mount_dir = None self.mount_dev = None self.mount_size = None self.disk_img = disk_img self.mount_ok = mount_ok # Default is to mount partition with /etc/passwd if not self.mount_ok: self.mount_ok = lambda mount_dir: os.path.isfile(mount_dir+"/etc/passwd") # Example kpartx output # kpartx -p p -v -a /tmp/ # add map loop2p1 (253:2): 0 3481600 linear /dev/loop2 2048 # add map loop2p2 (253:3): 0 614400 linear /dev/loop2 3483648 kpartx_output = runcmd_output(["kpartx", "-v", "-a", "-s", self.disk_img]) logger.debug(kpartx_output) # list of (deviceName, sizeInBytes) self.loop_devices = [] for line in kpartx_output.splitlines(): # add map loop2p3 (253:4): 0 7139328 linear /dev/loop2 528384 # 3rd element is size in 512 byte blocks if line.startswith("add map "): fields = line[8:].split() self.loop_devices.append( (fields[0], int(fields[3])*512) ) def __enter__(self): # Mount the device selected by mount_ok, if possible mount_dir = tempfile.mkdtemp() for dev, size in self.loop_devices: try: mount( "/dev/mapper/"+dev, mnt=mount_dir ) if self.mount_ok(mount_dir): self.mount_dir = mount_dir self.mount_dev = dev self.mount_size = size break umount( mount_dir ) except CalledProcessError: logger.debug(traceback.format_exc()) if self.mount_dir:"Partition mounted on %s size=%d", self.mount_dir, self.mount_size) else: logger.debug("Unable to mount anything from %s", self.disk_img) os.rmdir(mount_dir) return self def __exit__(self, exc_type, exc_value, exc_tb): if self.mount_dir: umount( self.mount_dir ) os.rmdir(self.mount_dir) self.mount_dir = None execWithRedirect("kpartx", ["-d", "-s", self.disk_img]) ######## Functions for making filesystem images ##########################
[docs]def mkfsimage(fstype, rootdir, outfile, size=None, mkfsargs=None, mountargs="", graft=None): '''Generic filesystem image creation function. fstype should be a filesystem type - "mkfs.${fstype}" must exist. graft should be a dict: {"some/path/in/image": "local/file/or/dir"}; if the path ends with a '/' it's assumed to be a directory. Will raise CalledProcessError if something goes wrong.''' if mkfsargs is None: mkfsargs = [] if graft is None: graft = {} preserve = (fstype not in ("msdos", "vfat")) if not size: size = estimate_size(rootdir, graft, fstype) with LoopDev(outfile, size) as loopdev: try: runcmd(["mkfs.%s" % fstype] + mkfsargs + [loopdev]) except CalledProcessError as e: logger.error("mkfs exited with a non-zero return code: %d", e.returncode) logger.error(e.output) sys.exit(e.returncode) with Mount(loopdev, mountargs) as mnt: if rootdir: copytree(rootdir, mnt, preserve) do_grafts(graft, mnt, preserve) # Make absolutely sure that the data has been written runcmd(["sync"]) # convenience functions with useful defaults
[docs]def mkdosimg(rootdir, outfile, size=None, label="", mountargs="shortname=winnt,umask=0077", graft=None): mkfsimage("msdos", rootdir, outfile, size, mountargs=mountargs, mkfsargs=["-n", label], graft=graft)
[docs]def mkext4img(rootdir, outfile, size=None, label="", mountargs="", graft=None): mkfsimage("ext4", rootdir, outfile, size, mountargs=mountargs, mkfsargs=["-L", label, "-b", "1024", "-m", "0"], graft=graft)
[docs]def mkbtrfsimg(rootdir, outfile, size=None, label="", mountargs="", graft=None): mkfsimage("btrfs", rootdir, outfile, size, mountargs=mountargs, mkfsargs=["-L", label], graft=graft)
[docs]def mkhfsimg(rootdir, outfile, size=None, label="", mountargs="", graft=None): mkfsimage("hfsplus", rootdir, outfile, size, mountargs=mountargs, mkfsargs=["-v", label], graft=graft)