Due to various ordering issues in ModemManager, when the Enable()
call fails because a PIN is required, MM will disable the modem
and *then* complete the dbus call. That means that NM gets the
Enable property change before the Enable() call completes, and NM
would then transition the device to DISCONNECTED. What we
really want to do is get the response from MM and then figure out
whether to fail activation or get secrets, not go to DISCONNECTED.
To fix this, ignore Enabled property changes to FALSE while
activating so we can handle it ourselves. There are still a few
fixes left in nm-applet to make the secrets request for PINs work
on-the-fly, but saved in the connection works fine.
If the user has already mapped the hostname to some other address,
assume the user knows what they are doing (since they probably do)
and leave that mapping alone, and create a minimal 127.0.0.1 entry.
If there was no hostname mapping, add it to the 127.0.0.1 entry
like NM did before.
Having the original hostname checks in the settings-service code
didn't allow the policy enough granularity to differentiate between
a plugin-provided hostname and the original hostname. We want to
fall back to the original hostname if there isn't a persistent
hostname (from a plugin) and if there isn't a DHCP-provided
hostname. Moving the original hostname checks to the policy
makes that possible. Clarify the precedence order at the same
time, and minimally validate the DHCP hostname as well.
Sometimes cause the cache not to refill with all interfaces,
meaning NM sometimes got the wrong carrier state from the
kernel which prevented NM from taking over existing connections.
NM shouldn't really be calling Enable(False) except in response to
direct user requests to turn off WWAN, much like rfkill, since
Enable(False) /is/ essentially rfkill for 3G. Instead, we should
be powering up the modem before trying to use it, and only
disconnecting after we're done. Let the user do enable/disable
when they want to.
This also fixes issues with other devices like GPS potentially
using the modem-manager at the same time as NM and ensures that NM
won't punch the modem in the face while GPS is using it.
And remove cargo-culted internal stuff which is no longer needed.
The ifcfg-rh sha1 stuff wasn't even used anymore after the move to
certificate paths.
The ipw2x00 drivers won't be converted over to the kernel's rfkill
subsystem until 2.6.33, and thus listening for udev rfkill change
events on these devices doesn't work. So until then, poll rfkill
state for ipw2x00 devices every few seconds in addition to listening
to other rfkill sources.
Shouldn't be allowing scan requests when associating or when the
supplicant is otherwise busy doing something else.
Older fullmac cards are much more likely to run into this problem
since they usually take longer to connect; since they take so
long, NM may sometimes request a scan during association or
during DHCP which can cause the card to miss DHCP replies. I've
never seen this happen with mac80211 drivers though.
Since rfkill state is saved but not acted upon during sleep
(since NM shouldn't be touching devices while sleeping) we have to
remember to act on the new state when waking up.
3rd patch in a series with:
0bbdc6b0fcb135fa3265
With NM 0.8 the system settings service was integrated into NM and
thus nm_connection_clear_secrets() acts directly on the system
settings plugins' NMConnection objects. So when NM cleared secrets
(for example after determining that they might be bad in a device's
stage2 handler), we completely lost the secrets forever.
With this commit, the secrets are now cached and updated whenever
the connection is updated, and thus are again available to send to
NetworkManager when needed.
nm_connection_replace_settings() replaces the connection's settings
but doesn't allow interception of the new settings. Plugins would then
send out the update signal, but secrets are scrubbed out of them to
ensure secrets aren't leaked out into D-Bus signals.
With NM 0.8 the system settings service was integrated into NM and
thus nm_connection_clear_secrets() acts directly on the system
settings plugins' NMConnection objects. So when NM cleared secrets
(for example after determining that they might be bad in a device's
stage2 handler), we completely lost the secrets forever.
Adding this function allows the system settings service to hook into
the connection updates when the plugin connection's backing storage
(like config files or whatever) changes and cache the secrets for
use in NMSettingsConnectionInterface get_secrets() requestes.
Turn DHCP and DNS debugging on with NM_DNSMASQ_DEBUG.
Without --strict-order, dnsmasq will round-robin queries which in
the case of VPN connections may result in the query going to the
non-VPN nameserver. Also, allow dnsmasq to poll resolv.conf for
nameserver updates so that when the default connection changes,
it knows about the new nameservers.
If the supplicant interface became ready but the device was still
UNAVAILABLE, requesting a scan before transitioning to DISCONNECTED
would just fail because can_scan() returns FALSE when the device's
state is not >= DISCONNECTED. Change device state first so that
doesn't happen.
Impact of this bug is likely limited to Ad-Hoc connections that don't
require a scan before activation since by the time the scan has finished,
the NMSupplicantInterface will be set up. However, this shows a bug where
Ad-Hoc connections can be immediately activated even if they don't have
the latest timestamp, because a scan hasn't completed yet and thus we don't
know if there are any usable APs around. Could be fixed by only letting
auto-activations happen after the first successful scan anyway. But whatever...
Log messages look like this:
NetworkManager: <info> Activation (wlan0/wireless): connection 'Wireless connection 1' requires no security. No secrets needed.
NetworkManager: <info> Config: added 'ssid' value 'foobar'
NetworkManager: <info> Config: added 'mode' value '1'
NetworkManager: <info> Config: added 'frequency' value '2412'
NetworkManager: <info> Config: added 'key_mgmt' value 'NONE'
(NetworkManager:28239): GLib-GObject-CRITICAL **: g_object_ref: assertion `G_IS_OBJECT (object)' failed
(NetworkManager:28239): GLib-GObject-CRITICAL **: g_type_instance_get_private: assertion `instance != NULL && instance->g_class != NULL' failed
NetworkManager: dbus_g_proxy_begin_call: assertion `DBUS_IS_G_PROXY (proxy)' failed
NetworkManager: <WARN> real_act_stage2_config(): Activation (wlan0/wireless): couldn't send wireless configuration to the supplicant.
NetworkManager: <info> (wlan0): device state change: 5 -> 9 (reason 9)
NetworkManager: <info> Activation (wlan0) failed for access point (foobar)
NetworkManager: <info> Marking connection 'Wireless connection 1' invalid.
This happened because the nm_device_wifi_set_enabled() only checked for
the existence of the NMSupplicantInterface, but not whether the supplicant
interface was ready to be used. The supplicant interface would be in the
middle of the getInterface or addInterface call and wouldn't have
initialized priv->iface_proxy yet, which is where that error message was
coming from.
So don't change device state from the wifi_enabled handler, just init
the supplicant interface (it should have been torn down already by
device_state_changed() when the device goes to UNAVAILABLE or UNMANAGED)
and wait for the supplicant interface state change to READY to change
the NMDeviceWifi state to DISCONNECTED in supplicant_iface_state_cb_handler().
Link against libm (using the LT_LIB_M macro and the $LIBM variable) as
we are using pow() and otherwise fail to build with stricter linkers like
binutils-gold.
