More general purpose API for generic options of settings.
The predicate function is also nicely usable via bindings.
One question is about the form of the predicate. In this case,
it is convenient to pass nm_ethtool_optname_is_coalesce(). On the
other hand, it's not very flexible as it does not accept a user
data argument. Use NMUtilsPredicateStr here, which is not flexible
but convenient for where it's used.
NMSettingEthtool is implemented using "gendata", meaning a hash
of GVariant. This is different from most other settings that have
properties implemented as GObject properties. There are two reasons
for this approach:
- The setting is transferred via D-Bus as "a{sv}" dictionary.
By unpacking the dictionary into GObject properties, the setting
cannot handle unknown properties. To be forward compatible (and
due to sloppy programming), unknown dictionary keys are silently
ignored when parsing a NMSetting. That is error prone and also
prevents settings to be treated loss-less.
Instead, we should at first accept all values from the dictionary.
Only in a second step, nm_connection_verify() rejects invalid settings
with an error reason. This way, the user can create a NMSetting,
but in a separate step handle if the setting doesn't verify.
"gendata" solves this by tracking the full GVariant dictionary.
This is still not entirely lossless, because multiple keys are
combined.
This is for example interesting if an libnm client fetches a connection
from a newer NetworkManager version. Now the user can modify the
properties that she knows about, while leaving all unknown
properties (from newer versions) in place.
- the approach aims to reduce the necessary boiler plate to create
GObject properties. Adding a new property should require less new code.
This approach was always be intended to be suitable for all settings, not only
NMSettingEthtool. We should not once again try to add API like
nm_setting_ethtool_set_feature(), nm_setting_ethtool_set_coalesce(), etc.
Note that the option name already fully encodes whether it is a feature,
a coalesce option, or whatever. We should not have
"nm_setting_set_$SUB_GROUP (setting, $ONE_NAME_FROM_GROUP)" API, but
simply "nm_setting_option_set (setting, $OPTION)" accessors.
Also, when parsing a NMSettingEthtool from a GVariant, then a feature
option can be any kind of variant. Only nm_setting_verify() rejects
variants of the wrong type. As such, nm_setting_option_set*() also
doesn't validate whether the variant type matches the option. Of course,
if you set a value of wrong type, verify() will reject the setting.
Add new general purpose API for this and expose it for NMSetting.
We are going to expose some of this API in libnm.
The name "gendata" (for "generic data") is not very suited. Instead,
call the public API nm_setting_option_*(). This also brings no naming
conflict, because currently no API exists with such naming.
Rename the internal API, so that it matches the API that we are going
to expose next.
This was intended for when the gendata hash should be converted
to/from a GValue/GHashTable. This would have been used, if
we also would have added a GObject property that exposes
the hash. But that was never done (at least not for NMSettingEthtool
and not yet).
This code is not used. If you ever need it, revert the patch
or implement it anew.
This function is not used nor does it seem useful.
Either you only need the names (nm_setting_gendata_get_all_names())
or you need the names and values together (_nm_setting_gendata_get_all()).
Getting the values without knowing the corresponding name makes
little sense. If you need it, call _nm_setting_gendata_get_all()
instead.
The filter function in nm_setting_gendata_clear_all() is useful
for when you want to only clear values according to a predicate,
if no such function is supplied all values will be cleared.
https://bugzilla.redhat.com/show_bug.cgi?id=1614700
Add '_nm_setting_bond_get_option_or_default()' and move all the custom
policies applied by NM for bond options in there.
One such example of a custom policy is to set 'miimon' to 0 (instead of its
default value of 100) if 'arp_interval' is explicitly enabled
and 'miimon' is not.
This means removing every piece of logic from
nm_setting_bond_add_option() which used to clear out 'arp_interval' and
'arp_ip_target' if 'miimon' was set or clear out 'miimon' along with
'downdelay', 'updelay' and 'miimon' if 'arp_interval' was set.
This behaviour is a bug since the kernel allow setting any combination
of this options for bonds and NetworkManager should not limit the user
to do so.
Also use 'set_bond_attr_or_default()' instead of 'set_bond_attr()' as
the former calls '_nm_setting_bond_get_option_or_default()' to implement
the right logic to retrieve bond options according to current bond
configuration.
Doing 'verify()' with options such as 'miimon' and 'num_grat_arp' set to
arbitrary values it's not consistent with what NetworkManager does to
bond options when activating the bond through 'apply_bonding_config()'
(at a later stage) because the said values do not
correspond to what the default values for those options are.
This leads to an inconsistency with the 'miimon' parameter for example,
where 'verify()' is done while assuming it's 0 if not set but its
default value is actually 100.
Fixes: 8775c25c33 ('libnm: verify bond option in defined order')
G_PARAM_CONSTRUCT cause to explicitly initialize the property during
object construction. This is an unnecessary overhead that we can easily
avoid.
The overhead is because G_PARAM_CONSTRUCT parameters are always set with
g_object_set() before calling constructed(). Even if they are not specified
during g_object_new(), in which case it calls set with the property's default
value. This also requires g_object_new() to iterate all properties to
find and sort the construct properties.
NMSetting are supposed to be simple classes. They don't need to have
their properties initialized before object construction completes.
Especially if the default values are NULL or zero, in which case there
is nothing to do. If the default value is not NULL or zero, we need
to initialize the field instead in the nm_setting*_init() function.
In total, we register 447 property informations. Out of these,
326 are plain, GObject property based without special implementations.
The NMSettInfoProperty had all function pointers directly embedded,
currently this amounts to 5 function pointers and the "dbus_type" field.
That means, at runtime we have 326 times trivial implementations with
waste 326*6*8 bytes of NULL pointers. We can compact these by moving
them to a separate structure.
Before:
447 * 5 function pointers
447 * "dbus_type" pointer
= 2682 pointers
After:
447 * 1 pointers (for NMSettInfoProperty.property_type)
89 * 6 pointers (for the distinct NMSettInfoPropertType data)
= 981 pointers
So, in total this saves 13608 byes of runtime memory (on 64 bit arch).
The 89 NMSettInfoPropertType instances are the remaining distinct instances.
Note that every NMSettInfoProperty has a "property_type" pointer, but most of them are
shared. That is because the underlying type and the operations are the same.
Also nice is that the NMSettInfoPropertType are actually constant,
static fields and initialized very early.
This change also makes sense form a design point of view. Previously,
NMSettInfoProperty contained both per-property data (the "name") but
also the behavior. Now, the "behavioral" part is moved to a separate
structure (where it is also shared). That means, the parts that are
concerned with the type of the property (the behavior) are separate
from the actual data of the property.
This is a complete refactoring of the bluetooth code.
Now that BlueZ 4 support was dropped, the separation of NMBluezManager
and NMBluez5Manager makes no sense. They should be merged.
At that point, notice that BlueZ 5's D-Bus API is fully centered around
D-Bus's ObjectManager interface. Using that interface, we basically only
call GetManagedObjects() once and register to InterfacesAdded,
InterfacesRemoved and PropertiesChanged signals. There is no need to
fetch individual properties ever.
Note how NMBluezDevice used to query the D-Bus properties itself by
creating a GDBusProxy. This is redundant, because when using the ObjectManager
interfaces, we have all information already.
Instead, let NMBluezManager basically become the client-side cache of
all of BlueZ's ObjectManager interface. NMBluezDevice was mostly concerned
about caching the D-Bus interface's state, tracking suitable profiles
(pan_connection), and moderate between bluez and NMDeviceBt.
These tasks don't get simpler by moving them to a seprate file. Let them
also be handled by NMBluezManager.
I mean, just look how it was previously: NMBluez5Manager registers to
ObjectManager interface and sees a device appearing. It creates a
NMBluezDevice object and registers to its "initialized" and
"notify:usable" signal. In the meantime, NMBluezDevice fetches the
relevant information from D-Bus (although it was already present in the
data provided by the ObjectManager) and eventually emits these usable
and initialized signals.
Then, NMBlue5Manager emits a "bdaddr-added" signal, for which NMBluezManager
creates the NMDeviceBt instance. NMBluezManager, NMBluez5Manager and
NMBluezDevice are strongly cooperating to the point that it is simpler
to merge them.
This is not mere refactoring. This patch aims to make everything
asynchronously and always cancellable. Also, it aims to fix races
and inconsistencies of the state.
- Registering to a NAP server now waits for the response and delays
activation of the NMDeviceBridge accordingly.
- For NAP connections we now watch the bnep0 interface in platform, and tear
down the device when it goes away. Bluez doesn't send us a notification
on D-Bus in that case.
- Rework establishing a DUN connection. It no longer uses blocking
connect() and does not block until rfcomm device appears. It's
all async now. It also watches the rfcomm file descriptor for
POLLERR/POLLHUP to notice disconnect.
- drop nm_device_factory_emit_component_added() and instead let
NMDeviceBt directly register to the WWan factory's "added" signal.
Add test for checking the meta data for expected consistency.
This is also useful if you want to check something about the meta data
programatically.
For example, if you have the question which (if any) properties
are GObject based but also implement a to_dbus_fcn() function. Then you
can extend this code with some simple printf debugging to get a list of
those.
Or, if you want to find how many NMSettInfoProperty instances are in
static data (e.g. to determine how much memory is used). You can easily
modify this code to count them (and find 447 properties). Out of these,
326 are plain GObject based properties. Meaning, we could refactor the
code to create smaller NMSettInfoProperty instances for those, saving
thus (326 * 4 * sizeof (gpointer)) bytes (10K).
Such questions are interesting when refactoring the code.
