We make an effort to get a better fallback case with
_bad_random_bytes().
Also make an effort to get good randomness in the first place. Even if
we compile against libc headers that don't provide getrandom(). Also,
this isn't really ugly, because for a long time glibc was reluctant to
add getrandom() wrapper and using syscall() was the way to go.
(cherry picked from commit 05a6936bef)
nm_utils_random_bytes() tries to get good randomness. If it fails, we still
try our own approach, but also signal that the returned numbers are bad.
In practice, none of the callers cares about the return value, because they
wouldn't know what to do in case of bad randomness (abort() is not an
option and retry is not expected to help and sending an email to the
admin isn't gonna help either). So the fallback case really should try
its best.
The fallback case depends on a good random seed and a good pseudorandom
number generator.
Getting a good seed is in reality impossible, after kernel let us down.
That is part of the problem, but we try our best.
The other part is to use a cryptographic pseudorandom number generator.
GRand uses a Mersenne Twister, so that is not good enough. In this low
level code we also cannot call gnutls/nss, because usually we don't have
that dependency. Maybe we could copy&paste the chacha20 implementation,
it's simple enough and a compatible license. That might be good, but
instead cock our own by adding some sha256 into the mix. This is
fallback code after all, and we want to try hard, but not *that* hard to
add chacha20 to NetworkManager.
So, what we do is to use a well seeded GRand instance, and XOR that
output with a sha256 digest of the state. It's probably slow, but
performance is not the issue in this code path.
(cherry picked from commit c22c3ce9f9)
If a prefix delegation is needed, currently NM restarts DHCPv6 on the
device with default route, but only if DHCPv6 was already running.
Allow the device to start DHCPv6 for a PD even if it was running
without DHCPv6.
See also: https://github.com/coreos/fedora-coreos-tracker/issues/888
(cherry picked from commit 62869621bd)
Previously we sent announcements immediately for non-controllers, or
after the first port was attached for controllers.
This has two problems:
- announcements can be sent when there is no carrier and they would
be lost;
- if a controller has a port, the port could be itself a controller;
in that case we start sending ARPs with the fake address of the
port. Later, when a leaf port is added to the second-level
controller, the correct port MAC will be propagated by kernel up to
both controllers.
To solve both problems, send ARP announcements only when the interface
has carrier. This also solves the second issue because controllers
created by NM have carrier only when there is a port with carrier.
Fixes: de1022285a ('device: do ARP announcements only after masters have a slave')
https://bugzilla.redhat.com/show_bug.cgi?id=1956793
(cherry picked from commit 1377f160ed)
I missed that we already have a gettid() wrapper. Drop the duplicated
again and use nm_utils_gettid().
Fixes: e874c5bf6b ('random: Provide missing gettid() declaration')
(cherry picked from commit 0c4a65929c)
Coverity warns about this:
2. NetworkManager-1.32.2/tools/generate-docs-nm-property-infos.py:117: identical_branches:
The same code is executed regardless of whether "temp.index(subelement) < len(temp) - 1" is
true, because the then and else branches are identical. Should one of the branches be modified,
or the entire 'if' statement replaced?
Fixes: c3504f7e62 ('Rewrite `./tools/generate-docs-nm-property-infos.py` with XML library')
(cherry picked from commit dd595c6c1d)
If the kernel command-line doesn't contain an explict ip=$method,
currently the generator creates connections with both IPv4 and IPv6
set to 'auto', and both allowed to fail.
Since NM is run in configure-and-quit mode in the initrd, NM can get
an IPv4 address or an IPv6 one (or both) depending on which address
family is quicker to complete. This unpredictable behavior is not
present in the legacy module, which always does IPv4 only by default.
Set a required-timeout of 20 seconds for IPv4, so that NM will
preferably get an IPv4, or will fall back to IPv6.
See also: https://gitlab.freedesktop.org/NetworkManager/NetworkManager/-/issues/729
(cherry picked from commit 0a18e97345)
Change the logic in check_ip_state() to delay the connection ACTIVATED
state if an address family is pending and its required-timeout has not
expired.
(cherry picked from commit 35cccc41cb)
Add a new property to specify the minimum time interval in
milliseconds for which dynamic IP configuration should be tried before
the connection succeeds.
This property is useful for example if both IPv4 and IPv6 are enabled
and are allowed to fail. Normally the connection succeeds as soon as
one of the two address families completes; by setting a required
timeout for e.g. IPv4, one can ensure that even if IP6 succeeds
earlier than IPv4, NetworkManager waits some time for IPv4 before the
connection becomes active.
(cherry picked from commit cb5960cef7)
g_rand_new() reads /dev/urandom and falls back to timestamp and pid.
At this point, we already unsuccessfully tried getrandom()/urandom,
so that doesn't seem promising to try.
Try harder to get good random seeds for our GRand instance.
Have one global instance, that gets seeded with various things that come
to mind. The random sequence of that instance is then used to initialize
the thread-local GRand instances.
Maybe this is all snake oil. If we fail to get good randomness by using
kernel API, what can we do? But really, callers also don't know how they
should handle a failure to get random data (short of abort() or
logging), so there is value in nm_utils_random_bytes() trying really
the best it can, and callers pretending that it doesn't fail.
This aims to improve the fallback case.
(cherry picked from commit 3649efe2b5)
The previous implementation was just wrong.
Fixes: e1ca3bf7ed ('shared: add nm_strerror_native() to replace strerror() and g_strerror()')
(cherry picked from commit 5bc39d9783)
_nm_thread_local is very neat, but when we allocate resources
we need to make sure that they are destroyed when the thread
exits.
We can use pthread_setspecific() for that, but using it is cumbersome.
Add a helper function to make that simpler.
Also, the number of possible pthread_key_t keys is limited. With this
way, we only need one key in total.
(cherry picked from commit bec4a40437)
We require these, otherwise we can get a linker error about
_nm_utils_monotonic_timestamp_initialized symbol being undefined.
(cherry picked from commit 09fb7877a9)
We have two GKeyfile files (timestamps and seen-bssids).
When a profile was deleted while NetworkManager was running, then
entries were removed from these keyfiles. But if a profile disappeared
while NetworkManger was stopped, then those UUIDs piled up.
This also happens if you have temporary connections in /run and reboot.
We need a way to garbage collect entries that are no longer relevant.
As the keyfile databases only get loaded once from disk, we will prune
all UUIDs for which we have no more connection loaded, on the first time
we write out the files again.
Note what this means: if you "temporarily" remove a connection profile
(without NetworkManager noticing) and restore it later, then the additional
information might have been pruned. There is no way how NetworkManager
could know that this UUID is coming back. The alternative is what we did
before: pile them up indefinitely. That seems more problematic.
(cherry picked from commit 2e720a1dc8)
Previously, there was no limit how many seen-bssids are tracked.
That seems problematic, also because there is no API how to get
rid of an excessive list of entries.
We should limit the number of entries. Add an (arbitrary) limit
of 30.
But this means that we drop the surplus of entries, and for that it
seems important to keep the newest, most recently seen entries.
Previously, entries were merely sorted ASCIIbetically. Now, honor
their order (with most recently seen first).
Also, normalize the BSSIDs. From internal code, we should only get
normalize strings, but when we load them from disk, they might be bogus.
As we might cut of the list, we don't want that invalid entries
cut of valid ones. And of course, invalid entries make no sense at
all.
(cherry picked from commit 8278719840)
ifcfg-rh plugin never stored the seen bssid list to file, and
keyfile no longer does, and it's no longer parsed from GVariant.
So there is actually no way how anything could be set here.
The seen-bssids should only be populate from
"/var/lib/NetworkManager/seen-bssids". Nowhere else.
(cherry picked from commit 15a0271781)
"wifi.seen-bssid" is an unusual property, therefore very ugly due to the
inconsistency.
It is not a regular user configuration that makes sense to store to
disk or modify by the user. It gets populated by the daemon, and
stored in "/var/lib/NetworkManager/seen-bssids" file.
As such, how to convert this to/from D-Bus needs special handling.
This means, that the to/from D-Bus functions will only serialize the
property when the seen-bssids are specified via
NMConnectionSerializationOptions, which is what the daemon does.
Also, the daemon ignores seen-bssids when parsing the variant.
This has the odd effect that when the client converts a setting to
GVariant, the seen-bssids gets lost. That means, a conversion to GVariant
and back looses information. I think that is OK in this case, because the
main point of to/from D-Bus is not to have a lossless GVariant representation
of a setting, but to transfer the setting via D-Bus between client and
daemon. And transferring seen-bssids via D-Bus makes only sense from the daemon
to the client.
(cherry picked from commit 4a4f214722)
"seen-bssids" primarily gets stored to "/var/lib/NetworkManager/seen-bssids",
it's not a regular property.
We want this property to be serialized/deserialized to/from GVariant,
because we expose these settings on the API like a property of the
profile. But it cannot be modified via nmcli, it cannot be stored
to ifcfg files, and it makes not sense to store it to keyfile either.
Stop doing that.
(cherry picked from commit d9ebcc8646)
