Age | Commit message (Collapse) | Author | Files | Lines |
|
Good drivers indicate when they're up by the IFF_UP flag.
This is the end of the (undocumented) UP_SLEEP variable and introduction of the UP_TIMEOUT (in seconds, default: 5) variable.
This has 'noticeable speed improvement'-potential.
|
|
We hardcode a timeout of 10 seconds. In case of a timeout, we use whatever results we have. It is unlikely that a timeout is hit.
For fast scans, we are done half a second earlier than before.
|
|
When we don't really care which file implements the needed command (i.e. we don't call it by its full path), we use type to establish availability of the command.
In wifi-menu we add a check for the dialog command and remove potential clearing of relevant error messages.
|
|
Needed for bonding interfaces.
|
|
pppd is a bit stupid with regard to configuration: It will always read /etc/ppp/options.
This overrides the default 'auth' setting.
|
|
|
|
In IPv6 you cannot bind to an address as long as it is not determined to be unique through Duplicate Address Detection. This can cause problems when routes in ROUTES6 reference a tentative source address or when something that starts after netcfg tries to.
This patch does a few things:
- Improve ROUTES(6) logic so that non-static connections can make use of them as intended according to the commit message of c8be1.
- Make additional arguments possible for ADDR6 addresses and ROUTES6 routes, such as ADDR6=('1234:5678:9abc:def::1/64 nodad').
- Introduce a DAD_TIMEOUT variable that governs how long to wait for DAD to finish.
- Move IPCFG processing to after IPv6 address set-up (it already was after IPv4 address set-up), so that custom commands can be sure to have working addresses.
WARNING: This changes make it impossible to use ROUTES6 and set the address through IPCFG instead of ADDR6. This was bad, unsupported practice, but sometimes needed as ADDR6 did not allow config flags to be used. Resolving this issue should be no problem: you should be able to properly use ADDR6 now. If not, then be a real hacker and set the routes through IPCFG too.
Many thanks to Steve Caligo for reporting this bug and helping out in finding a solution.
|
|
BOND_INTERFACES was documented, but SLAVE_INTERFACES was implemented.
|
|
Some changes after reading through a few files.
Only two modifications impact functionality.
- Not having dhclient is now fatal for IPv6 connections that use dhcp.
- Allow authentication/associaton time-out of wpa to be set through TIMEOUT for wired connections too.
dhclient no longer needs to be killed manually, so those lines are removed.
|
|
Installing Virtualbox guest additions should work again.
|
|
Previously the packagers username would be visible in the tarball and all files laying around in docs/ would be included.
|
|
|
|
iproute2 does not allow multiple default gateways in ipv6. this patch
uses 'ip route replace' which will replace an existing route, or add a
new one if none preexists.
|
|
I noticed most of your website links are http://; at least for
wiki/bbs/bugs these are all https so you might as well skip the
redirect and link directly to the secure site.
|
|
Don't try to bring down backgrounded profiles if they are not up, preventing "not connected" messages.
|
|
Give users a way to stumble upon this marvelous piece of literature that had completely been forgotten (it was introduced in the repository around netcfg 2.6, but never installed).
|
|
This was the result of commit b6964 which said to fix FS#17406 by reverting to the subshell-less for-loop that is once again removed in this commit.
The real fix to FS#17406 was commit c3276 to the initscripts repository.
|
|
There is support for IP='no' option in profiles to have the interface be
brought up but assign no addresses. This patch adds documentation of
that option to the man page.
|
|
50 is a better number for netcfg anyway.
|
|
We need to make sure no the systemd services are only considered
started when the network is actually up. Otherwise, things relying
on the network don't know when they can start.
This means that we need to use Type=forking or Type=oneshot, and in
the case of ifplugd we must daemonize and wait for the daemon to be
ready before we continue.
Also do some minor cleanup:
- don't include env files that are not used
- don't ignore return values
Signed-off-by: Tom Gundersen <teg@jklm.no>
|
|
It is not enough to order ourselves Before network target, we must also
make sure it is pulled in. This means that anything that needs the network
to be up before they start can order themselves After=network.target.
Reported-by: Tobias Powalowski <tpowa@archlinux.org>
Signed-off-by: Tom Gundersen <teg@jklm.no>
|
|
Moved from the pm-utils package.
|
|
|
|
- Initscripts users are now offered familiar output.
- Configuration through /etc/rc.conf was not gracefully deprecated.
|
|
It is once again possible to use NET= at the kernel command line when using net-profiles.
|
|
|
|
Configuration is done through /etc/conf.d/netcfg.
Compatibility with configuration through /etc/rc.conf is dropped.
|
|
The netcfg service provides support for the NETWORKS array in /etc/conf.d/netcfg through netcfg-daemon.
The netcfg@ service allows individual profiles to be treated as services. These services wait until a connection is established (the other service has support for backgrounding)!
|
|
|
|
For reuse in a systemd service
|
|
/etc/rc.d/functions is owned by initscripts on which netcfg does not explicitly depend.
The revised layout is inspired by systemd.
|
|
Provide a sample hook file for supporting firewall rule sets based on network 'locations'.
|
|
Having multiple versions of netcfg around is not supported.
|
|
This will allow us to get rid of /lib eventually.
Signed-off-by: Tom Gundersen <teg@jklm.no>
|
|
Add spaces where necessary.
|
|
Commit 6b43b missed the point.
As noted by Cláudio, Jesse Young, Maciej Sitarz and Harley Laue.
|
|
- wireless-dbus is not simple enough for netcfg
- some contributed code was not used and the rest of the code had caught up functionality wise
- we switched to iproute so we need to stop pretending it is an 'option'
- it is no longer fatal in the ifplugd scrict to have multiple preferred profiles
|
|
|
|
archlinux.org points to gudrun, but tarballs should be uploaded to
gerolde.
Signed-off-by: Florian Pritz <bluewind@xinu.at>
|
|
|
|
- Introduce $SYSCTL_INTERFACE.
- Don't disable routing advertisements for stateful DHCPv6, as routes aren't sent by the DHCP server, but rely on routing advertisements to be picked up by the client.
By: Steve Caligo, Fixes FS#26259
|
|
- Use "-K" to prevent unnecessary remove/re-acquire a lease.
- Respect /etc/network.d/interfaces/$interface and the dhcp options set there when connecting without a matching profile. Also use the same options used in $CONN_DIR/ethernet.
Fixes by Thomas Bächler
|
|
|
|
This shaves a few kilobytes.
|
|
No need to keep support the "-old" syntax when it doesn't really do anything.
|
|
It needs to be quoted when used as an argument. Putting the name in curly braces is unnecessary.
|
|
Bash completion does not deal properly with profile filenames that contain whitespace.
There are some minor tweaks in this commit as well, but those should be harmless.
|
|
Nothing special.
|
|
|
|
This commit does not change any installed code.
- The tarball now includes pre-built documentation, removing the build dependency on asciidoc from the PKGBUILD.
- The tarball is now xz compressed.
- News is updated.
|