Using sync init (nm_client_new()) has an overhead as it requires an internal
GMainContext to ensure preserving the order of D-Bus messages. Let's avoid
that by using the async init. Note that the difference here is that we will
iterate the caller's GMainContext while creating the instance. But that
is no problem for nmtui at that point.
Benefits:
- nmc_client_new_async*() allows to set properties on the NMClient
instance before calling g_async_initable_init_async().
It also allows to subscribe to any signals (like NM_CLIENT_DEVICE_ADDED)
before actually iterating the GMainContext. This is a sensible and
supported thing to do!
- nmc_client_new_waitsync() iterates the GMainContext until the (async)
initialization is complete. That is different from synchronous nm_client_new(),
which does not iterate the caller's GMainContext, and hence needs an
internal context to ensure the order of events is honored.
- nmc_client_new_waitsync() always returns the NMClient instance, even
if initialization fails.
That is useful if you need the nm_client_get_context_busy_watcher() instance
to ensure all pending messages are completed.
We have "shared/nm-libnm-core-aux", which is shared code that can be used
by anybody (including libnm-core, src, libnm and clients).
We have "clients/common", which are helper function for clients. But
that implies that the code is inside "clients". I think it would be
useful to have auxiliary code that extends libnm, but is not only
usable by code in "clients". In other words, "shared/nm-libnm-aux"
is a better place than "clients/common", and I think most of the
functionality form "clients/common" should move there.
"shared/nm-utils" got long renamed and split into separate parts. The remaining
tests are really to test nm-std-aux and nm-glib-aux (no libnm dependencies). Move
the tests to the appropriate place.
Currently, NMClient by default always fetches the permissions
("GetPermissions()") and refreshes them on "CheckPermissions" signal.
Fetching permissions is relatively expensive, while they are not used
most of the time. Allow the user to opt out of this.
For that, have a NMClientInstanceFlags to enable/disable automatic
fetching. Also add a "permissions-state" property that allows the user
to understand whether the cached permissions are up to date or not.
This is a bit an awkward API for handling this. E.g. you cannot
explicitly request permissions, you can just enable/disable fetching
permissions. And then you can watch the permission-state to know whether
you are ready. It's done this way because it fits the previous model
and extends the API with a (relative) small amount of new functions and
properties.
Add a flags property to control behavior of NMClient.
Possible future use cases:
- currently it would always automatically fetch permissions. Often that
is not used and the user could opt out of it.
- currently, using sync init creates an internal GMainContext. This
has an overhead and may be undesirable. We could implement another
"sync" initialization that would merely iterate the callers mainloop
until the initialization completes. A flag would allow to opt in.
- currently, NMClient always fetches all connection settings
automatically. Via a flag the user could opt out of that.
Instead NMClient could provide an API so the user can request
settings as they are needed.
On D-Bus, the permission names are just the PolicyKit action names, like
"org.freedesktop.NetworkManager.wifi.scan". But NMClient already
ignores all strings that it doesn't know at compile time and only
keeps track of well known permission.
And neither does the API nm_client_get_permissions_result() allow to
expose permissions unknown to libnm.
Maybe the API of NMClient should be more generic and allow exposing
any permissions announced by NetworkManager. As it is however, it's
not necessary to track the permissions in a hash table. An array with
fixed indices is sufficient.
In NetworkManager we have NMAuthCallResult, which is really a duplicate
of NMClientPermissionResult.
Maybe NMAuthCallResult should be entirely replaced by NMClientPermissionResult.
But the name NMClientPermissionResult is a bit awkward. But then the
duplication is even more awkward... fixing this is left for another day.
Reuse the list of all permissions and don't duplicate it.
Also, now the result of GetPermissions() on D-Bus contains the
permissions sorted by name. We get it almost for free, and it's
a nice property.
For example with
mount -o remount,rw,hidepid=1 /proc/
all permission checks will fail with an error. Internally, we map the
failure to NM_AUTH_CALL_RESULT_UNKNOWN.
<trace> [1575645672.5958] auth: call[1069]: CheckAuthorization(org.freedesktop.NetworkManager.enable-disable-connectivity-check), subject=unix-process[pid=468316, uid=1000, start=1912881]
<trace> [1575645672.6295] auth: call[1069]: completed: failed: GDBus.Error:org.gtk.GDBus.UnmappedGError.Quark._g_2dfile_2derror_2dquark.Code4: Failed to open file “/proc/468316/status”: No such file or directory
<debug> [1575645672.6296] manager: unknown auth chain result 0
First of all, we should not log a debug message about that (we already log the
result of permission checks separately).
Also, we should include the unknown result in the response. The permission was
checked, and omitting it from GetPermissions() result seems wrong (even if we
failed to get the result).
Note that "unknown" is now a new possible return value on D-Bus. But
see how nm_permission_result_to_client() would map such a value to
"unknown" as well. So, it's probably a fine extension of the D-Bus API.
Note that NMClient API is currently quite limited. The user won't know
whether permissions were received (and if they were received, they
could not distinguish between UNKNOWN and absent). Hence, returning
all permissions as unknown (or not at all) causes `nmcli general permissions`
to hang. The solution here is to improve NMClient API to allow the user
to know when the permissions are received. But this patch doesn't
fix the hanging of nmcli nor the limitation of NMClient's API.
If the activation of an assumed device fails, we first set the device
state to FAILED and then to ACTIVATED. In the FAILED state, the active
connection transitions to DEACTIVATED and clears its device pointer;
hence we end up with an inconsistent state which causes assertion
failures in other parts of the code (for example, get_best_ip_config()
assumes that the device of the best active connection is not NULL).
Don't first transition to FAILED and then to ACTIVATED, just set the
latter.
https://bugzilla.redhat.com/show_bug.cgi?id=1737774https://gitlab.freedesktop.org/NetworkManager/NetworkManager/merge_requests/351
(cherry picked from commit 93e9010b75)
If the activation of an assumed device fails, we first set the device
state to FAILED and then to ACTIVATED. In the FAILED state, the active
connection transitions to DEACTIVATED and clears its device pointer;
hence we end up with an inconsistent state which causes assertion
failures in other parts of the code (for example, get_best_ip_config()
assumes that the device of the best active connection is not NULL).
Don't first transition to FAILED and then to ACTIVATED, just set the
latter.
https://bugzilla.redhat.com/show_bug.cgi?id=1737774https://gitlab.freedesktop.org/NetworkManager/NetworkManager/merge_requests/351
inet_aton() is very accepting when parsing the address. For example,
it accepts addresses with fewer octets (interpreting the last octet
as a number in network byte order for multiple bytes). It also ignores
any trailing garbage after the first delimiting whitespace (at least,
the glibc implementation). It also accepts octets in hex and octal
notation.
For the initrd reader we want to be more forgiving than inet_pton()
and also accept addresses like 255.000.000.000 (octal notation). For
that we would want to use inet_aton(). But we should not accept all the
craziness that inet_aton() otherwise accepts.
Use nm_utils_parse_inaddr_bin_full() instead. This function implements
our way how we want to interpret IP addresses in string representation.
Under the hood, of course it also uses inet_pton() and even inet_aton(),
but it is stricter than inet_aton() and only accepts certain formats.
(cherry picked from commit d68373c305)
The @family argument is an input and output argument.
Initially, the family is set to AF_UNSPEC, in which case the family
gets detected based on the IP address. However, we call
dt_get_ipaddr_property() multiple times to parse the netmask, the
gateway and the IP address.
That means, after the first successfull call, the @family is set to
AF_INET or AF_INET6.
Note that the previous code (in the switch block) would only check that
the family is set to AF_UNSPEC, but it would not check that the @family
matches the expected binary address length @len. Later, we then might call
nm_ip_address_new_binary() with a family and a binary address of
unexpected length.
Also drop the error checking for nm_ip_address_new_binary().
nm_ip_address_new_binary() can only fail if the prefix length is larger
than 32/128. The function has no way to validate the input arguments
beyond that and can thus not fail (short of undefined behavior).
(cherry picked from commit 9618f1bb4b)
inet_aton() also supports IPv4 addresses in octal (with a leading '0')
or where not all 4 digits of the address are present.
Add nm_utils_parse_inaddr_bin_full() to optionally fallback to
parse the address with inet_aton().
Note taht inet_aton() also supports all crazy formats, including
ignoring trailing garbage after a whitespace. We don't want to accept
that in general.
Note that even in legacy format we:
- accept everything that inet_pton() would accept
- additionally, we also accept some forms which inet_aton() would
accept, but not all.
That means, the legacy format that we accept is a superset of
inet_pton() and a subset of inet_aton(). Which is desirable.
(cherry picked from commit 8fbf67d138)
