NetworkManager.git
10 years agoipv6: fix incorrect address config signal emission
Dan Williams [Tue, 20 Oct 2009 18:23:10 +0000]
ipv6: fix incorrect address config signal emission

device->want_signal was never set to TRUE when addrconf was started,
causing random netlink events (say for link-local address addition
or removal) to trigger the config-changed signal from
nm_ip6_device_sync_from_netlink() at the wrong time.  This would
cause IPv6 address configuration to look like it succeeded, when
in fact the config timeout was still in-force.  Thus device
activation would proceed if IPv4 was enabled, but a few seconds later
the device would be deactivated due to the still active IPv6
timeout.

So fix that and clarify when the events from the IPv6 manager happen,
and what the want_signal variable is really for.

10 years agoip6: save the accept_ra value and re-set it when the device is deactivated
Dan Williams [Mon, 19 Oct 2009 22:38:21 +0000]
ip6: save the accept_ra value and re-set it when the device is deactivated

10 years agoip6: use device's IP interface, not its physical interface
Dan Williams [Mon, 19 Oct 2009 21:05:36 +0000]
ip6: use device's IP interface, not its physical interface

10 years agoifcfg-rh: fix writing LEAP connections
Dan Williams [Mon, 19 Oct 2009 17:20:25 +0000]
ifcfg-rh: fix writing LEAP connections

10 years agolibnm-util: fix checking for TLS and TTLS phase2 secrets
Dan Williams [Mon, 19 Oct 2009 06:36:47 +0000]
libnm-util: fix checking for TLS and TTLS phase2 secrets

Two errors here; first, need_secrets_tls() was not updated correctly
for the certificate paths changes that landed recently, and would
have incorrectly returned "no secrets required" for the PATH scheme.

Second, an incorrect strcmp() comparison in need_secrets_phase2()
meant that the wrong TTLS phase2 method would get asked if it
required secrets.

10 years agolibnm-util: ensure GSM setting default values
Dan Williams [Mon, 19 Oct 2009 05:51:43 +0000]
libnm-util: ensure GSM setting default values

Broken by 00f945e54efc112e60ed547c9c7796a7657e4117.

10 years agomanager: automatically pick a base connection for VPNs
Dan Williams [Fri, 16 Oct 2009 18:52:27 +0000]
manager: automatically pick a base connection for VPNs

If let callers pass "/" for the specific object, and NM will
automatically pick the default device.

10 years agosystem-settings: fix writing connections when an earlier plugin can't (bgo #581758)
Dan Williams [Fri, 16 Oct 2009 03:52:00 +0000]
system-settings: fix writing connections when an earlier plugin can't (bgo #581758)

The error object passed to the plugin's add-connection handler wasn't
getting properly cleared if an earlier plugin had failed to write
the connection and fell back to the current plugin.

10 years agoifcfg-rh: add testcase for (rh #528068)
Dan Williams [Tue, 13 Oct 2009 19:32:06 +0000]
ifcfg-rh: add testcase for (rh #528068)

10 years agovpn: clear secrets when the connection fails
Dan Williams [Fri, 9 Oct 2009 06:00:01 +0000]
vpn: clear secrets when the connection fails

NM previously only cleared secrets when the VPN service daemon quit,
and the service daemons are on a 10-second inactivity timer.  So if
the user tried to re-activate the failed VPN connection within 10
seconds the old secrets would get used, which clearly isn't what we
want.  Ensure that whenever the VPN connection fails or disconnects,
we ask the settings service for secrets again the next time.

10 years agoifcfg-rh: recognize 'static' BOOTPROTO (rh #528068)
Dan Williams [Thu, 8 Oct 2009 23:18:04 +0000]
ifcfg-rh: recognize 'static' BOOTPROTO (rh #528068)

10 years agodoc: update code style docs
Adam Langley [Wed, 7 Oct 2009 19:28:10 +0000]
doc: update code style docs

10 years agogsm: correct GSM band implementation and settings
Dan Williams [Wed, 7 Oct 2009 19:18:57 +0000]
gsm: correct GSM band implementation and settings

NM didn't pass it to MM anyway, so it was mainly unused, but the band settings
were still wrong.  Fix that (and still preserve ABI) by adding a new property
for allowed bands that can actually hold all the bands instead of limiting
to 16-bits.  Clean up some of the deprecation stuff at the same time to make
it clearer what's deprecated and what to do about it.

10 years agosharing: fix sharing after 715ddd2045cf786e6c80e6d3d8617fe70ad24552 (lp:444756)
Noel J. Bergman [Tue, 6 Oct 2009 23:45:39 +0000]
sharing: fix sharing after 715ddd2045cf786e6c80e6d3d8617fe70ad24552 (lp:444756)

Commit 715ddd2045cf786e6c80e6d3d8617fe70ad24552 broke sharing because
sometimes a multi-argument string gets passed in rule->rule.  Revert
most of that commit.  A bunch of cleanups by dcbw too.

10 years agosystem-settings: fix crash saving default wired connections to backing storage (lp...
Dan Williams [Tue, 6 Oct 2009 00:46:16 +0000]
system-settings: fix crash saving default wired connections to backing storage (lp:430781)

Fix up refcounting, plus it turns out that we already have the MAC address
lying around as a GObject data item, so we don't need to go grab it
from the connection itself.

10 years agointrospection: clarify wired device 'Speed' property units
Dan Williams [Mon, 5 Oct 2009 23:58:07 +0000]
introspection: clarify wired device 'Speed' property units

10 years agoppp: add minimal debugging output to the ppp plugin
Dan Williams [Mon, 5 Oct 2009 19:12:45 +0000]
ppp: add minimal debugging output to the ppp plugin

10 years agoifupdown: unmanage devices with "mapping" stanzas - lp:413622 bgo:591864
Alexander Sack [Mon, 5 Oct 2009 17:53:54 +0000]
ifupdown: unmanage devices with "mapping" stanzas - lp:413622 bgo:591864

10 years agowired: fix build_supplicant_config for wired - bgo:594038 lp:422597
Frej Soya [Mon, 5 Oct 2009 09:22:20 +0000]
wired: fix build_supplicant_config for wired - bgo:594038 lp:422597

10 years agomanager: fix use-after-free crash for cdc-ether modems
Dan Williams [Mon, 5 Oct 2009 06:36:06 +0000]
manager: fix use-after-free crash for cdc-ether modems

An ethernet device that's actually a component of a modem could get
added to the device list, but then destroyed without getting removed
from the device list.

10 years agolibnm-util: fix distcheck after 7a8611e0ad2966c71fd402d6d09f31f9a084f08f
Dan Williams [Sat, 3 Oct 2009 00:20:38 +0000]
libnm-util: fix distcheck after 7a8611e0ad2966c71fd402d6d09f31f9a084f08f

10 years agolibnm-util: fix NSS padding checking and add testcase
Dan Williams [Fri, 2 Oct 2009 20:26:40 +0000]
libnm-util: fix NSS padding checking and add testcase

10 years agoMerge branch 'iptables'
Tambet Ingo [Fri, 2 Oct 2009 06:20:48 +0000]
Merge branch 'iptables'

10 years agocore: Make iptables path configurable
Tambet Ingo [Thu, 1 Oct 2009 11:31:09 +0000]
core: Make iptables path configurable

On some systems, it's located in different path...

10 years agolibnm-glib: add class padding for future expansion
Dan Williams [Wed, 30 Sep 2009 16:20:44 +0000]
libnm-glib: add class padding for future expansion

10 years agolibnm-util: add class padding for future expansion
Dan Williams [Wed, 30 Sep 2009 16:11:54 +0000]
libnm-util: add class padding for future expansion

10 years agolibnm-util: fix build with gcc 4.4 - pass string format to g_set_error
Alexander Sack [Wed, 30 Sep 2009 14:01:16 +0000]
libnm-util: fix build with gcc 4.4 - pass string format to g_set_error

10 years agodoc: 802-1x doc fixes
Dan Williams [Mon, 28 Sep 2009 23:55:52 +0000]
doc: 802-1x doc fixes

10 years agodoc: rebuild sections file too
Dan Williams [Mon, 28 Sep 2009 23:55:44 +0000]
doc: rebuild sections file too

10 years agodoc: let gtk-doc rebuild libnm-util.types
Dan Williams [Mon, 28 Sep 2009 23:12:39 +0000]
doc: let gtk-doc rebuild libnm-util.types

10 years agodoc: regenerate docs when sources change
Dan Williams [Mon, 28 Sep 2009 23:01:21 +0000]
doc: regenerate docs when sources change

10 years agodoc: the sgml file is normally stored in RCS
Dan Williams [Mon, 28 Sep 2009 23:00:56 +0000]
doc: the sgml file is normally stored in RCS

10 years agolibnm-util: doc fixes
Dan Williams [Mon, 28 Sep 2009 23:00:30 +0000]
libnm-util: doc fixes

10 years agoifcfg-rh: really fix writing the WPA PSK and add testcase for it
Dan Williams [Mon, 28 Sep 2009 22:29:29 +0000]
ifcfg-rh: really fix writing the WPA PSK and add testcase for it

10 years agoifcfg-rh: treat BOOTPROTO=none as static IP
Dan Williams [Mon, 28 Sep 2009 22:08:07 +0000]
ifcfg-rh: treat BOOTPROTO=none as static IP

10 years agowired: defer carrier-off events while connected by a few seconds
Dan Williams [Mon, 28 Sep 2009 21:12:09 +0000]
wired: defer carrier-off events while connected by a few seconds

Don't immediately tear down an active wired connection when the carrier
flips to off, but wait a few seconds for it to come back before breaking
the user's network.

10 years agoifcfg-rh: quote WPA passphrases when writing connections
Dan Williams [Mon, 28 Sep 2009 20:34:42 +0000]
ifcfg-rh: quote WPA passphrases when writing connections

The reader requires passphrases to be quoted, but the writer wasn't
doing that.  Hex PSKs were fine though.

10 years agoifcfg-rh: handle BOOTPROTO errors better
Dan Williams [Mon, 28 Sep 2009 20:09:50 +0000]
ifcfg-rh: handle BOOTPROTO errors better

10 years agowired: link_timeout_id -> supplicant_timeout_id
Dan Williams [Mon, 28 Sep 2009 19:57:18 +0000]
wired: link_timeout_id -> supplicant_timeout_id

10 years agoethernet: make initial carrier check synchronous; require carrier for connection...
Dan Williams [Mon, 28 Sep 2009 18:09:31 +0000]
ethernet: make initial carrier check synchronous; require carrier for connection takeover (lp#417719)

10 years agonetlink: add blocking function to get carrier state
Dan Williams [Mon, 28 Sep 2009 17:50:22 +0000]
netlink: add blocking function to get carrier state

10 years agocore: spacing fix
Dan Williams [Mon, 28 Sep 2009 17:49:55 +0000]
core: spacing fix

10 years agolibnm-util: bump soname minor component (1.0.0 -> 1.1.0) for API additions
Dan Williams [Sat, 26 Sep 2009 22:45:45 +0000]
libnm-util: bump soname minor component (1.0.0 -> 1.1.0) for API additions

10 years agopo: remove dead translatable
Dan Williams [Sat, 26 Sep 2009 00:34:29 +0000]
po: remove dead translatable

10 years agoMerge commit 'origin/master' into cert-paths
Dan Williams [Fri, 25 Sep 2009 23:02:53 +0000]
Merge commit 'origin/master' into cert-paths

10 years agoRevert "libnm-util: bump soname for certificate changes"
Dan Williams [Fri, 25 Sep 2009 23:01:23 +0000]
Revert "libnm-util: bump soname for certificate changes"

This reverts commit 5b90df073707d88c4f97b70b71290f5a282bff66.

Since commit 71219015ea3a52ecb8cd0519247a964e4cbf515a adds all the
0.7.x certificate and key functions back, we can revert the soname
change and preserve API/ABI.

10 years agolibnm-util: add 0.7 cert/key functions back
Dan Williams [Fri, 25 Sep 2009 23:00:19 +0000]
libnm-util: add 0.7 cert/key functions back

Since there's a more or less direct mapping between the 0.7.x and
the 0.8.x certificate and key operations, we might as well just
deprecate them instead of removing them entirely.

10 years agolibnm-glib: add 'connections-read' signal to settings interface
Witold Sowa [Wed, 23 Sep 2009 06:44:45 +0000]
libnm-glib: add 'connections-read' signal to settings interface

To let listeners know when all connections have been found.

10 years agolibnm-glib: add service-running property to NMRemoteSettings
Witold Sowa [Wed, 23 Sep 2009 06:29:02 +0000]
libnm-glib: add service-running property to NMRemoteSettings

So that users can actually tell if the remote service is running or not.
It doesn't have to be when the object is created.

11 years agoMerge commit 'origin/master' into cert-paths
Dan Williams [Mon, 21 Sep 2009 17:59:53 +0000]
Merge commit 'origin/master' into cert-paths

11 years agolibnm-util: silence some of the GSM PUK warnings
Dan Williams [Mon, 21 Sep 2009 17:59:19 +0000]
libnm-util: silence some of the GSM PUK warnings

11 years agolibnm-util: bump soname for certificate changes
Dan Williams [Mon, 21 Sep 2009 17:57:52 +0000]
libnm-util: bump soname for certificate changes

11 years agoMerge commit 'origin/master' into cert-paths
Dan Williams [Sat, 19 Sep 2009 14:01:49 +0000]
Merge commit 'origin/master' into cert-paths

11 years agolibnm-util: deprecate GSM PUK property
Dan Williams [Sat, 19 Sep 2009 00:08:30 +0000]
libnm-util: deprecate GSM PUK property

Unused anyway, and PUK is dangerous enough that it should be handled
by the client explicitly and not by NM.

11 years agoMerge commit 'origin/master' into cert-paths
Dan Williams [Fri, 18 Sep 2009 16:09:03 +0000]
Merge commit 'origin/master' into cert-paths

11 years agocore: device disconnection cleanups
Dan Williams [Fri, 18 Sep 2009 16:05:11 +0000]
core: device disconnection cleanups

Return an error when trying to disconnect an already-disconnected
or deactivated device.

11 years agolibnm-glib: make nm_device_disconnect() async
Dan Williams [Fri, 18 Sep 2009 15:38:29 +0000]
libnm-glib: make nm_device_disconnect() async

11 years agolibnm-glib: add nm_device_disconnect function to glib API
Alexander Sack [Thu, 17 Sep 2009 23:55:31 +0000]
libnm-glib: add nm_device_disconnect function to glib API

11 years agocore: implement manual disconnect feature
Alexander Sack [Wed, 16 Sep 2009 11:18:24 +0000]
core: implement manual disconnect feature

In the past networkmanager did not allow to manually disconnect devices.

Manually disconnected devices will not be automatically reconnected until one
of the following events occur:
 1. user activates a connection for the currently disconnected device
 2. network manager awakes from hibernate/suspend
 3. network manager is restarted (e.g. reboot)

Add a Disconnect method to generic NMDevice dbus interface; set a new private
autoconnect_inhibit flag if Disconnect method is called through dbus.
Based on this auto activation for devices gets inhibited until one
of the above events occur.

11 years agonm-tool: don't depend on private NM includes
Dan Williams [Thu, 17 Sep 2009 21:07:53 +0000]
nm-tool: don't depend on private NM includes

11 years agoMerge commit 'origin/master' into cert-paths
Dan Williams [Thu, 17 Sep 2009 18:24:56 +0000]
Merge commit 'origin/master' into cert-paths

11 years agovpn: fix crash caused by 8d205212dcf67fa0bf4e4730bf8353f65e7059ee
Dan Williams [Thu, 17 Sep 2009 18:23:44 +0000]
vpn: fix crash caused by 8d205212dcf67fa0bf4e4730bf8353f65e7059ee

Typo; should have been 'unref'.

11 years agosystem-settings: handle byte-array secrets too
Dan Williams [Thu, 17 Sep 2009 17:38:54 +0000]
system-settings: handle byte-array secrets too

11 years agolibnm-glib: fix getting remote secrets
Dan Williams [Wed, 16 Sep 2009 20:50:35 +0000]
libnm-glib: fix getting remote secrets

11 years agolibnm-util: verify private keys with passwords; pkcs#12 crash fix
Dan Williams [Wed, 16 Sep 2009 19:42:52 +0000]
libnm-util: verify private keys with passwords; pkcs#12 crash fix

For private keys, don't just verify that the key can be read; try to
decrypt it to ensure that the password is actually valid.

Also fix a stupid {} mistake that caused crashes when setting pkcs#12
certificates because the check for PATH scheme wouldn't happen and
the private key would be NULL when trying to set the client cert to
the same data as the private key.

11 years agolibnm-util: clarify certificate/key path argument usage
Dan Williams [Wed, 16 Sep 2009 17:30:31 +0000]
libnm-util: clarify certificate/key path argument usage

In the future we'll allow cert/key IDs too, so don't lock ourselves
into filenames, but use the 'scheme' argument to specify what 'value'
means.

11 years agolibnm-util: add nm_utils_rsa_key_encrypt() and fix crypto padding mixups
Dan Williams [Tue, 15 Sep 2009 23:01:50 +0000]
libnm-util: add nm_utils_rsa_key_encrypt() and fix crypto padding mixups

To be backwards compatible clients need to handle both paths to private
keys and the decrypted private key data, which is what used to get passed
in the private-key and phase2-private-key attributes of the 802.1x setting.
When moving a connection around between system-settings and user-settings,
if the private key is decrypted data, the settings service needs to store
that decrypted data somewhere so that the key can be sent to NM during
the connection process.

But we don't want to store the decrypted private key data, so we have to
re-encrypt it (possibly generating a private key password if one wasn't
sent with the decrypted data) and save it to disk, then send NM a path
to that private key during connection.

To help clients do this, and so that they don't have to carry around
multiple crypto implementations depending on whether they want to use
NSS or gnutls/gcrypt, add a helper to libnm-util.

Furthermore, I misunderstood a bunch of stuff with crypto padding when
writing the encrypt/decrypt functions long ago, so fix that up.  Don't
return padding as part of the decrypted data, and make sure to verify
the padding's expected lengths and values when decrypting.  Many thanks
to Nalin Dahyabhai for pointing me in the right direction.

11 years agoifcfg-rh: clarify comment
Dan Williams [Tue, 15 Sep 2009 04:04:19 +0000]
ifcfg-rh: clarify comment

11 years agoMerge commit 'origin/master' into cert-paths
Dan Williams [Tue, 15 Sep 2009 02:34:25 +0000]
Merge commit 'origin/master' into cert-paths

11 years agopo: update British English (en_GB) translation (bgo #595072)
Philip Withnall [Mon, 14 Sep 2009 21:28:20 +0000]
po: update British English (en_GB) translation (bgo #595072)

11 years agoMerge commit 'origin/master' into cert-paths
Dan Williams [Mon, 14 Sep 2009 20:38:19 +0000]
Merge commit 'origin/master' into cert-paths

11 years agocore: rename nm_device_can_activate -> nm_device_is_available
Dan Williams [Mon, 14 Sep 2009 20:33:06 +0000]
core: rename nm_device_can_activate -> nm_device_is_available

Becuase when the device is in >= DISCONNECTED state, that already
means it can activate if a connection is available.

11 years agocore: generalize unavailable -> disconnected delayed transition
Dan Williams [Mon, 14 Sep 2009 20:24:29 +0000]
core: generalize unavailable -> disconnected delayed transition

Instead of doing this in every device subclass, do it in the NMDevice
superclass.  nm_device_can_activate() already did the same logic that
each of the subclass device_state_changed() handlers were doing to
figure out whether they could do the transition from unavailable
to disconnected, so just use that in NMDevice and kill lots of code.

11 years agobuild: remove nm-gsm-device.c
Dan Williams [Mon, 14 Sep 2009 20:22:46 +0000]
build: remove nm-gsm-device.c

No idea why this is still around; it's a left-over from 0.7.x and
replaced by src/modem-manager/nm-modem-gsm.c.

11 years agowired: get rid of nm_device_ethernet_get_carrier
Dan Williams [Mon, 14 Sep 2009 19:54:09 +0000]
wired: get rid of nm_device_ethernet_get_carrier

Only used internally; just use priv->carrier instead

11 years agoethernet: clean up netlink monitoring code
Dan Williams [Mon, 14 Sep 2009 19:12:33 +0000]
ethernet: clean up netlink monitoring code

Keep a reference to the netlink monitor for as long as we have
signals attached to it.  Also don't bother looking up the ifindex
in the carrier state signal handlers since we already have it
lying around in priv.

11 years agolibnm-glib-vpn: bump soname to match soname bump for libnm-glib
Dan Williams [Mon, 14 Sep 2009 14:55:09 +0000]
libnm-glib-vpn: bump soname to match soname bump for libnm-glib

So that Debian can bump package name too.

11 years agocore: always tear down devices when their hardware goes away
Dan Williams [Fri, 11 Sep 2009 00:49:11 +0000]
core: always tear down devices when their hardware goes away

Previous connection-assumption code that made sure not to touch
ethernet devices when NM quit also didn't clean up after devices
that were physically removed, which was wrong.

11 years agovpn: don't reset VPN gateway route when parent device is going down
Dan Williams [Fri, 11 Sep 2009 00:38:47 +0000]
vpn: don't reset VPN gateway route when parent device is going down

Or really, any time the parent IP4 config is cleared which is pretty
much only when the device is going down.  Avoids a warning when
nm_system_add_ip4_vpn_gateway_route() can't get the parent IP4Config
because it doesn't exist.

11 years agopo: rename sr@Latn -> sr@latin (bgo #555863)
Dan Williams [Fri, 11 Sep 2009 13:02:20 +0000]
po: rename sr@Latn -> sr@latin (bgo #555863)

11 years agopo: add Kannada (kn) translation (bgo #585826)
Shankar Prasad [Fri, 11 Sep 2009 12:56:33 +0000]
po: add Kannada (kn) translation (bgo #585826)

11 years agopo: add Telugu (te) translation (bgo #585553)
Krishna Babu K [Fri, 11 Sep 2009 12:54:21 +0000]
po: add Telugu (te) translation (bgo #585553)

11 years agopo: update Assamese (as) translation (bgo #585537)
Amitakhya Phukan [Fri, 11 Sep 2009 12:51:42 +0000]
po: update Assamese (as) translation (bgo #585537)

11 years agopo: updated Gujarati (gu) translation (bgo #585111)
Sweta Kothari [Fri, 11 Sep 2009 12:49:50 +0000]
po: updated Gujarati (gu) translation (bgo #585111)

11 years agopo: add Bengali (bn_IN) translation (bgo #585067)
Runa Bhattacharjee [Fri, 11 Sep 2009 12:46:20 +0000]
po: add Bengali (bn_IN) translation (bgo #585067)

11 years agopo: update Italian translation (bgo #575041)
Francesco Marletta [Fri, 11 Sep 2009 12:43:10 +0000]
po: update Italian translation (bgo #575041)

11 years agopo: update Marathi translations (bgo #573729)
Sandeep Shedmake [Fri, 11 Sep 2009 02:15:54 +0000]
po: update Marathi translations (bgo #573729)

11 years agopo: add missing translatable
Dan Williams [Fri, 11 Sep 2009 02:15:22 +0000]
po: add missing translatable

11 years agopo: updated Polish translation (bgo #594577)
Piotr Drąg [Fri, 11 Sep 2009 01:23:48 +0000]
po: updated Polish translation (bgo #594577)

11 years agopo: updated Swedish translation (bgo #594203)
Daniel Nylander [Fri, 11 Sep 2009 01:22:03 +0000]
po: updated Swedish translation (bgo #594203)

11 years agopo: Updated Simplified Chinese translation (bgo #594310)
Aron Xu [Fri, 11 Sep 2009 01:16:50 +0000]
po: Updated Simplified Chinese translation (bgo #594310)

11 years agopo: update Simplified Chinese translation (bgo #593774)
Aron Xu [Fri, 11 Sep 2009 01:14:13 +0000]
po: update Simplified Chinese translation (bgo #593774)

11 years agodoc: fix FSF address (bgo #575625)
Dan Williams [Fri, 11 Sep 2009 01:03:30 +0000]
doc: fix FSF address (bgo #575625)

11 years agovpn: fix crash when canceling VPN connection secrets request
Dan Williams [Thu, 10 Sep 2009 19:18:18 +0000]
vpn: fix crash when canceling VPN connection secrets request

The secrets provider might be unrefed while processing the response,
causing a double-free when the unref triggers the weak ref notifier
and clears out the GetSecretsInfo, but then later the dbus handler
that got the response in the first place tries to free the info
explicitly.  So keep the provider around while processing the result.

11 years agopolicy: reschedule failed devices for activation after a short delay
Noel J. Bergman [Thu, 10 Sep 2009 00:19:39 +0000]
policy: reschedule failed devices for activation after a short delay

Instead of immediately...  devices handled externally from NM (by
ModemManager or Bluez or whatever) have a slight delay in reporting
to NM that they device has been removed (due to internal housekeeping
and then pushing the event into D-Bus signals or something) and thus
when NM notices PPP failures and tries to reschedule activation
of autoconnect=true connections, it fails the reactivation because
the device actually is no longer present, but it hasn't gotten the
device-removed event from MM or Bluez yet.  That sets the invalid
tag on the connection, and when the device gets replugged, the
invalid tag suppresses autoconnect.

11 years agodevice: only invalidate connections if they failed during activation
Dan Williams [Wed, 9 Sep 2009 19:21:55 +0000]
device: only invalidate connections if they failed during activation

Otherwise, connections used to get marked invalid on hot-unplug, which
isn't really correct.

11 years agoppp: ignore certain pppd status changes
Dan Williams [Wed, 9 Sep 2009 14:44:21 +0000]
ppp: ignore certain pppd status changes

We don't actually care when pppd goes into the 'network' phase or the
'authenticate' phase, because we're looking for the IP4 configuration
to come back, and the device is already in the IP_CONFIG state.
Handling those phases would cause the device's state to jump around
between NEED_AUTH and IP_CONFIG when we were already past that phase.

Specifically, when the PPP link went down, the device would jump from
FAILED to IP_CONFIG because pppd entered the 'network' phase when
cleaning up the link.  The device would also jump from IP_CONFIG to
NEED_AUTH during the connection process when we already had secrets.

11 years agoinclude: number device activation stages in NetworkManager.h
Dan Williams [Wed, 9 Sep 2009 14:11:29 +0000]
include: number device activation stages in NetworkManager.h

11 years agomodem: disable modem instead of just disconnecting
Dan Williams [Fri, 4 Sep 2009 22:42:09 +0000]
modem: disable modem instead of just disconnecting

Until later when we'll actually do stuff like probing signal
strength and such while disconnected.

11 years agoifupdown: do not export any connection in managed=false mode - lp:291902
Alexander Sack [Fri, 4 Sep 2009 16:02:24 +0000]
ifupdown: do not export any connection in managed=false mode - lp:291902