`openssl dgst` and `openssl enc` used previously poorly handle key
stretching - in case of `openssl enc` encryption key is derived using
single MD5 iteration, without even any salt. This hardly prevent
brute force or even rainbow tables attacks. To make things worse, the
same key is used for encryption and integrity protection which ease
brute force even further.
All this is still about brute force attacks, so when using long, high
entropy passphrase, it should be still relatively safe. But lets do
better.
According to discussion in QubesOS/qubes-issues#971, scrypt algorithm is
a good choice for key stretching (it isn't the best of all existing, but
a good one and widely adopted). At the same time, lets switch away from
`openssl` tool, as it is very limited and apparently not designed for
production use. Use `scrypt` tool, which is very simple and does exactly
what we need - encrypt the data and integrity protect it. Its archive
format have own (simple) header with data required by the `scrypt`
algorithm, including salt. Internally data is encrypted with AES256-CTR
and integrity protected with HMAC-SHA256. For details see:
https://github.com/tarsnap/scrypt/blob/master/FORMAT
This means change of backup format. Mainly:
1. HMAC is stored in scrypt header, so don't use separate file for it.
Instead have data in files with `.enc` extension.
2. For compatibility leave `backup-header` and `backup-header.hmac`. But
`backup-header.hmac` is really scrypt-encrypted version of `backup-header`.
3. For each file, prepend its identifier to the passphrase, to
authenticate filename itself too. Having this we can guard against
reordering archive files within a single backup and across backups. This
identifier is built as:
backup ID (from backup-header)!filename!
For backup-header itself, there is no backup ID (just 'backup-header!').
FixesQubesOS/qubes-issues#971
Have a generic function `handle_streams`, instead of
`wait_backup_feedback` with open coded process names and manual
iteration over them.
No functional change, besides minor logging change.
Use just introduced tar writer to archive content of LVM volumes (or
more generally: block devices). Place them as 'private.img' and
'root.img' files in the backup - just like in old format. This require
support for replacing file name in tar header - another thing trivially
supported with tar writer.
tar can't write archive with _contents_ of block device. We need this to
backup LVM-based disk images. To avoid dumping image to a file first,
create a simple tar archiver just for this purpose.
Python is not the fastest possible technology, it's 3 times slower than
equivalent written in C. But it's much easier to read, much less
error-prone, and still process 1GB image under 1s (CPU time, leaving
along actual disk reads). So, it's acceptable.
Old backup metadata (old qubes.xml) does not contain info about
individual volume sizes. So, extract it from tar header (using verbose
output during restore) and resize volume accordingly.
Without this, restoring volumes larger than default would be impossible.
To ease all this, rework restore workflow: first create QubesVM objects,
and all their files (as for fresh VM), then override them with data
from backup - possibly redirecting some files to new location. This
allows generic code to create LVM volumes and then only restore its
content.
1. Add a helper function on vm.storage. This is equivalent of:
vm.storage.get_pool(vm.volumes[name]).export(vm.volumes[name])
2. Make sure the path returned by `export` on LVM volume is accessible.
First part - handling firewall.xml and rules formatting.
Specification on https://qubes-os.org/doc/vm-interface/
TODO (for dom0):
- plug into QubesVM object
- expose rules in QubesDB (including reloading)
- drop old functions (vm.get_firewall_conf etc)
QubesOS/qubes-issues#1815
Instead of excerpt from /proc/meminfo, use just one integer. This make
qmemman handling much easier and ease implementation for non-Linux OSes
(where /proc/meminfo doesn't exist).
For now keep also support for old format.
FixesQubesOS/qubes-issues#1312
There is no point in changing *public API* for just a change without any
better reason. It turned out most of those settings will be the same in
Qubes 4.0, so keep names the same.
This reverts commit 2d6ad3b60c.
QubesOS/qubes-issues#1812
This is migration of core2 commits:
commit d0ba43f253
Author: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Date: Mon Jun 6 02:21:08 2016 +0200
core: start guid as normal user even when VM started by root
Another attempt to avoid permissions-related problems...
QubesOS/qubes-issues#1768
commit 89d002a031
Author: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Date: Mon Jun 6 02:19:51 2016 +0200
core: use runuser instead of sudo for switching root->user
There are problems with using sudo in early system startup
(systemd-logind not running yet, pam_systemd timeouts). Since we don't
need full session here, runuser is good enough (even better: faster).
commit 2265fd3d52
Author: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Date: Sat Jun 4 17:42:24 2016 +0200
core: start qubesdb as normal user, even when VM is started by root
On VM start, old qubesdb-daemon is terminated (if still running). In
practice it happen only at VM startart (shutdown and quickly start
again). But in that case, if the VM was started by root, such operation
would fail.
So when VM is started by root, make sure that qubesdb-daemon will be
running as normal user (the first user in group 'qubes' - there should
be only one).
FixesQubesOS/qubes-issues#1745
Commit from core2:
commit 94d52a13e7
core: adjust guid parameters when running on KDE5
On KDE5 native decoration plugin is used and requires special properties
set (instead of `_QUBES_VMNAME` etc).
Special care needs to be taken when detecting environment, because
environment variables aren't good enough - this script may be running
with cleared environment (through sudo, or from systemd). So check
properties of X11 root window.
QubesOS/qubes-issues#1784
* core3-devices:
Fix core2migration and tests for new devices API
tests: more qubes.devices tests
qubes/ext/pci: implement pci-no-strict-reset/BDF feature
qubes/tools: allow calling qvm-device as qvm-devclass (like qvm-pci)
qubes: make pylint happy
qubes/tools: add qvm-device tool (and tests)
tests: load qubes.tests.tools.qvm_ls
tests: PCI devices tests
tests: add context manager to catch stdout
qubes/ext/pci: move PCI devices handling to an extension
qubes/devices: use more detailed exceptions than just KeyError
qubes/devices: allow non-persistent attach
qubes/storage: misc fixes for VM-exposed block devices handling
qubes: new devices API
FixesQubesOS/qubes-issues#2257
Instead of old per-VM flag 'pci_strictreset', now implement this as
per-device flag using features. To not fail on particular device
assignment set 'pci-no-strict-reset/DEVICE-BDF' to True. For
example 'pci-no-strict-reset/00:1b.0'.
QubesOS/qubes-issues#2257
Implement required event handlers according to documentation in
qubes.devices.
A modification of qubes.devices.DeviceInfo is needed to allow dynamic,
read-only properties.
QubesOS/qubes-issues#2257
Add 'backenddomain' element when source (not target) domain is not dom0.
Fix XML elemenet name. Actually set volume.domain when listing
VM-exposed devices.
QubesOS/qubes-issues#2256
Allow device plugin to list attached and available devices. Enforce
at API level every device being exposed by some domain.
This commit only changes devices API, but not update existing users
(pci) yet.
QubesOS/qubes-issues#2257
This script should run as fast as possible, so avoid importing large
module. In fact the only used thing was argparse wrapper, so switch to
the standard one and drop aliases.
QubesOS/qubes-issues#2256
Some tests do not apply, as there is no savefile and attributes
propagation is much simpler. Dropped tests:
- test_000_firewall_propagation
- test_001_firewall_propagation
- test_000_prepare_dvm
QubesOS/qubes-issues#2253
- fix assigning 'template' property - do not do it if VM already have it
set
- cap default maxmem at 4000, as we clamp it to 10*memory anyway (and
default memory is 400)
- DispVM is no longer a special case for storage
- Add missing 'rw=True' for volatile volume
- Handle storage initialization (copy&paste from AppVM)
- Clone properties from DispVM template
QubesOS/qubes-issues#2253
It is useful on some cases to prevent talking to hypervisor.
Warning - it may have sense only when action do not access any runtime
VM status. For example running the domain will fail, but changing its
properties should work.
Do not access vm.libvirt_domain after it being already removed - this
will redefine it again in libvirt, just to undefine it in a moment.
On the other hand, few lines below there is fallback libvirt cleanup, in
case of proper one not working.
