homed: don't insist on authentication against host-copy user record
homed maintains two or three copies of the user's identity record per
home directory: one on the host, one inside the LUKS header, and one
embedded in the home directory.
Previously we'd insist that if a user logs in they have to authenticate
against all three, as a safety feature. This broke logging into
unfixated records however, since in that case the host version is
synthetic and thus does not carry any authentication data.
Let's hence losen the strictness here: accept authentication against
host records that carry no auth data. This should be safe as we know
after all that the second/third record will catch invalid accesses.
Jeremy Cline [Fri, 15 May 2020 20:28:35 +0000 (16:28 -0400)]
shared: treat generator units as vendor units
Units created with "systemctl edit --full" from generator units aren't
removed with "systemctl revert" because units in the generator,
generator_early, and generator_late paths aren't considered vendor
units.
Alter path_is_vendor() to return true if the given path is a
generator unit, which in turn causes "systemctl revert" to remove units
created with "systemctl edit --full" as they now have vendor versions.
ATA devices should use the ATA ids like port number and (possibly)
master/slave or multiplier id, not the generic SCSI ID.
Currently only port number is included in the link. With this patch
the link would be able to support more cases, which are a) when the
device is behind a port multiplexer b) the distinction between master
and slave (distinguished by target id).
I tried to verify scenario a) with this patch, but I failed to find a
machine with PMP SATA devices attached. But the link below
https://github.com/systemd/systemd/issues/3943
could show what's the difference. Here is my test for scenario b)
Changelog:
v5: add another parameter compat_link in handle_scsi()
v4: comment for ID_PATH_ATA_COMPAT
get string length with pointer difference
(suggested by Franck Bui<fbui@suse.com>)
v3: creating compatible link from env
variables type change
v2: remove udev rules modification for compatible link
setup a test scenario of master/slave ATA devices
v1: initial patch
We generally return > 1 if any of the actions we are doing is instantly
complete and == 0 when we started doing it asynchronously (by forking
off homework), in our functions that execute operations on homes.
Fix a mix-up where the test for this was reversed in
home_dispatch_release() and home_dispatch_lock_all().
proc-cmdline: make checking of EFI options variable optional
And do not use it in the IMPORT{cmdline} udev code. Wherever we expose
direct interfaces to check the kernel cmdline, let's not consult our
systemd-specific EFI variable, but strictly use the actual kernel
variable, because that's what we claim we do. i.e. it's fine to use the
EFI variable for our own settings, but for the generic APIs to the
kernel cmdline we should not use it.
Specifically, this applies to IMPORT{cmdline} and
ConditionKernelCommandLine=. In the latter case we weren#t checking the
EFI variable anyway, hence let's do the same for the udev case, too.
bus-message: fix negative offset with ~empty message
In the linked reproducer, m->fields_size == 0, and we calculate ri == -1, which
of course doesn't end well. Skip the whole calculation if m->fields_size == 0,
and also check that we don't go negative even if it is non-zero.
Fixes https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=19446 and #15583.
udev: when random MACs are requested, generate them with genuine randomness
This is a security feature, and we thus shouldn't derive the random MACs
from a potentially guessable source. MAC addresses are after all facing
to the outside, and can be interacted with from untrusted environments.
Hence, let's generate them the same way as we generate UUIDs: from
getrandom() or /dev/urandom, and optionally with RDRAND if that's
supported.
RDRAND should be fine, since this is not cryptographic key material, but
ultimately public information. We just want to make sure conflicts are
not likely.
Previously we'd generate the MACs via rand(), which means given the
short seed they are a little bit too guessable, making collisions too
likely. See #14355 in particular.
Fixes: #14355
(Note that #14355 was already fixed by a0f11d1d11a546f791855ec9c47c2ff830e6a5aa, but I think we should do
better even, and not rely on rand() and uninitialized random pools)
sd-network: DHCPv6 - Add support to send vendor class data
```
21.16. Vendor Class Option
This option is used by a client to identify the vendor that
manufactured the hardware on which the client is running. The
information contained in the data area of this option is contained in
one or more opaque fields that identify details of the hardware
configuration. The format of the Vendor Class option is:
enterprise-number The vendor's registered Enterprise Number as
maintained by IANA [IANA-PEN]. A 4-octet
field containing an unsigned integer.
vendor-class-data The hardware configuration of the node on
which the client is running. A
variable-length field (4 octets less than the
value in the option-len field).
The vendor-class-data field is composed of a series of separate
items, each of which describes some characteristic of the client's
hardware configuration. Examples of vendor-class-data instances
might include the version of the operating system the client is
running or the amount of memory installed on the client.
Each instance of vendor-class-data is formatted as follows:
The vendor-class-len field is 2 octets long and specifies the length
of the opaque vendor-class-data in network byte order.
Servers and clients MUST NOT include more than one instance of
OPTION_VENDOR_CLASS with the same Enterprise Number. Each instance
of OPTION_VENDOR_CLASS can carry multiple vendor-class-data
instances.
```
Let's make sure we always apply the process properties from the user
record, in all our three successful paths:
1. when we register a regular session
2. when we run for the systemd --user session
3. when we have no logind (but might still have complex user records
from elsewhere) and thus exit early
cryptsetup: optionally, see if empty password works for unlocking the file system
This adds a new switch try-empty-password. If set and none of PKCS#11 or
key files work, it is attempted to unlock the volume with an empty
password, before the user is asked for a password.
Usecase: an installer generates an OS image on one system, which is the
booted up for the first time in a possibly different system. The image
is encrypted using a random volume key, but an empty password. A tool
that runs on first boot then queries the user for a password to set or
enrols the volume in the TPM, removing the empty password. (Of course, in
such a scenario it is important to never reuse the installer image on
multiple systems as they all will have the same volume key, but that's a
different question.)
Let's make loading of keys a bit more automatic and define a common
place where key files can be placed. Specifically, whenever a volume of
name "foo" is attempted, search for a key file in
/etc/cryptsetup-keys.d/foo.key and /run/cryptsetup-keys.d/foo.key,
unless a key file is declared explicitly.
With this scheme we have a simple discovery in place that should make it
more straightfoward wher to place keys, and requires no explicit
configuration to be used.
That's reduce the number of functions dealing with configuration
parsing/loading and should make the code simpler especially since this function
was used only once.
Most complexity of this patch is due to the fact that some manager settings
(basically the watchdog properties) can be set at runtime and in this case the
runtime values must be retained over daemon-reload or daemon-reexec.
For consistency sake, all watchdog properties behaves now the same way, that
is:
- Values defined by config files can be overridden by writing the new value
through their respective D-BUS properties. In this case, these values are
preserved over reload/reexec until the special value '0' or USEC_INFINITY
is written, which will then restore the last values loaded from the config
files. If the restored value is '0' or 'USEC_INFINITY', the watchdogs will
be disabled and the corresponding device will be closed.
- Reading the properties from a user instance will return the USEC_INFINITY
value as these properties are only meaningful for PID1.
- Writing to one of the watchdog properties of a user instance's will be a
NOP.
journalctl,elsewhere: make sure --file=foo fails with sane error msg if foo is not readable
It annoyed me for quite a while that running "journalctl --file=…" on a
file that is not readable failed with a "File not found" error instead
of a permission error. Let's fix that.
We make this work by using the GLOB_NOCHECK flag for glob() which means
that files are not accessible will be returned in the array as they are
instead of being filtered away. This then means that our later attemps
to open the files will fail cleanly with a good error message.
user-class-data The user classes carried by the client. The
length, in octets, is specified by
option-len.
The information contained in the data area of this option is
contained in one or more opaque fields that represent the user class
or classes of which the client is a member. A server selects
configuration information for the client based on the classes
identified in this option. For example, the User Class option can be
used to configure all clients of people in the accounting department
with a different printer than clients of people in the marketing
department. The user class information carried in this option MUST
be configurable on the client.
The data area of the User Class option MUST contain one or more
instances of user-class-data information. Each instance of
user-class-data is formatted as follows: