When a new wiphy is added, the kernel usually adds a default STA
interface as well. This interface is currently not signaled over
nl80211 in any way.
This implements a selective dump of the wiphy interfaces in order to
obtain the newly added netdev. Selective dump is currently not
supported by the kernel, so all netdevs will be returned. A patch on
linux-wireless is pending that implements the selective dump
functionality.
==24934== 16 bytes in 1 blocks are definitely lost in loss record 1 of 1
==24934== at 0x4C2C970: malloc (vg_replace_malloc.c:296)
==24934== by 0x41675D: l_malloc (util.c:62)
==24934== by 0x4033B3: netdev_set_linkmode_and_operstate
(netdev.c:149)
==24934== by 0x4042B9: netdev_free (netdev.c:221)
==24934== by 0x41735D: l_queue_clear (queue.c:107)
==24934== by 0x4173A8: l_queue_destroy (queue.c:82)
==24934== by 0x40543D: netdev_exit (netdev.c:1459)
==24934== by 0x402D6F: nl80211_vanished (main.c:126)
==24934== by 0x41E607: l_genl_family_unref (genl.c:1057)
==24934== by 0x402B50: main (main.c:237)
Instead of calling the device added or removed callback when the
interface is detected, call it when interface goes up or down. This
only affects the addition and removal of the WSC interface now.
During the network_info refactoring the adding of the connected BSS to
device->bss_list in case it is not in the scan results has moved to
after the l_hashmap_foreach_remove call meaning that the network could
be removed even though it is still pointed at by
device->connected networks. Reverse the order to what it was before.
Alternatively network_process network could take not of the fact the
network is connected and not call network_remove on it leaving it with
an empty bss_list.
It is probably rare that a disconnect should fail but if it happens the
device->state is not returned to CONNECTED and I'm not sure if it should
be, so the ConnectedNetwork property and other bits should probably be
reset at the start of the disconnection instead of at the end.
Also check if state is CONNECTED before calling network_disconnected
because network_connected may have not been called yet.
--interfaces (-i) tells iwd which interfaces to manage. If the option
is ommitted, all interfaces will be managed.
--nointerfaces (-I) tells iwd which interfaces to blacklist. If the
option is ommitted, no interfaces will be blacklisted.
When setting operstate to dormant or down, give it a callback for debug
purposes. It looks like that operstate down message does not have a
chance to go out currently.
knownnetworks.c/.h implements the KnownNetworks interface and loads the
known networks from storage on startup. The list of all the networks
including information on whether a network is known is managed in
network.c to avoid having two separate lists of network_info structures
and keeping them in sync. That turns out to be difficult because the
network.c list is sorted by connected_time and connected_time changes
can be triggered in both network.c or knownnetworks.c. Both can also
trigger a network_info to be removed completely.
network_info gets a is_known flag that is used for the
GetOrderedNetworks tracking and to implement the KnownNetworks
interface - loading of the list of known networks on startup and
forgetting networks.
For simplicity and future use (possibly performance), every struct network
gets a pointer to a network_info structure, there's one network_info for
every network being by any interface, not only known networks. The SSID
and security type information is removed from struct network because the
network_info holds that information.
network_info also gets a seen_count field to count how many references
from network.info fields it has, so as to fix the removal of
network_info structures. Previously, once they were added to the
networks list, they'd stay there forever possibly skewing the network
ranking results.
This also fixed the network ranking used by GetOrderedNetwork which
wasn't working due to a missing assignment of *index in
network_find_info also triggering valgrind alerts.
