This handles the situation where there are no ethernet ports on the bridge
and we want our wifi STA interface to have the MAC address of the bridge
so we can link assoclists with IPs.
Approved-by: Andrew Dodd
Approved-by: Evan Benn
The cgi-upload script (part of cgi-io) doesn't send any responses
while the upload is happening, only when it's finished.
Ideally this should be re-architected to stream 'something', but
for now, the simple option is just to bump the timeout here.
At 1Mbps, this means we should just be able to upload a 30MiB firmware
image within the timeout.
Approved-by: Sophronia Koilpillai
Approved-by: Prabha Murugesan
iwinfo will now ask the driver (i.e. BCF/OTP) for the countrylist,
and use the 1.16 morse driver channel list.
Approved-by: Sophronia Koilpillai
Approved-by: Andrew Dodd
Previously, this was done in the frontend display only (e.g. model name).
We should have just properly done the change originally.
Approved-by: Prabha Murugesan
Approved-by: Lyall Beveridge
Without this change, the bcm bluetooth module wasn't built on non-bcm
boards but it was still included in the package so building
the ipk would fail. This change makes the bcm module be built
like the other modules.
I think the question re RAVE_SP is also related to the bluetooth
change, but I've seen it pop up way too many times to fix it again
in a specific config file (and putting a not set in is a safe operation).
Approved-by: Sophronia Koilpillai
Approved-by: Evan Benn
- add the ability to override the reset button handling
- add a package which handles the 'improved' reset button handling
- make the ekh-mt76x8 targets have a reset button
- allow diag_override.sh to use existing code in diag.sh
Approved-by: Sophronia Koilpillai
Approved-by: Milad Mohtashamirad
When 2000 stations are connected the iwinfo assoclist command returns
a large amount of data. Use the head for the buffer in the rpcd iwinfo plugin,
the buffer is increased in a separate iwinfo patch.
Also increase the ubus max message size. 2000 stations sends a ~1.7MB
json message. This patch does not work for 8000 stations, which will
likely require a more well thought out fix.
Approved-by: James Haggerty
Approved-by: Lyall Beveridge
Previously, if umdns crashed/terminated for some reason, when it respawned
it would not have any config (and listen on any interfaces).
This has been fixed in upstream by adding a new callback to
procd (service_running), so we pull in these patches.
cf
9a79cdc7ee
Approved-by: Andrew Dodd
Approved-by: Lyall Beveridge
Without this, changing the ip addresses (and possible other
properties?) on an interface would lead to that interface
no longer being responsive.
Raised with upstream:
https://github.com/openwrt/mdnsd/pull/21
Approved-by: Sophronia Koilpillai
Approved-by: Lyall Beveridge
This is taken from upstream, but is more significant than you might think.
This is because if you don't clear the cache it still has references
to the freed ifaces, so it ends up crashing at some point in the future.
What's worse, when it crashes, even though umdns has respawn
it doesn't do set_config again. This means that any umdns crash
means it never recovers unless it's reconfigured, since set_config
is necessary to make it start listening.
Approved-by: Sophronia Koilpillai
Approved-by: Lyall Beveridge
Took this patch from the main branch. It makes umdns
a bit less chatty, which is useful given that we can
lose multicast packets (particularly in mesh/powersave).
Approved-by: Sophronia Koilpillai
Approved-by: Lyall Beveridge
Initially, the `reload_service` function in the UMDNS init script was renamed to allow PROCD to automatically handle reloads and trigger management. This should have been the preferred approach, as PROCD can dynamically manage triggers and decide when a full restart of the UMDNS service is necessary.
However, in practice, this did not work as expected. Despite `ubus call service list '{"name":"umdns", "verbose":true}'` showing the triggers being correctly added, they were not applied. Only the initially registered triggers responded to interface events, while newly added interfaces (e.g., ahwlan being added to the UMDNS UCI after the wizard) did not cause UMDNS to reload when they became active.
To ensure predictable behaviour, the UMDNS UCI triggers now explicitly restart the service which will ensure that `start_service` adds triggers properly so that they can detect interface events on new networks (i.e. ahwlan).
Approved-by: Sophronia Koilpillai
Approved-by: James Haggerty
Rather than use morse_setup.sh to force iwinfo from morse_feed, we instead change the normal Makefile.
This has the disadvantage that we're modifying 'real' OpenWrt, and making our custom package less visible, but has two advantages:
- users can use the standard OpenWrt commands to interact with our repo (without relying on morse_setup.sh to make things work)
- if upstream changes iwinfo, we'll conflict (i.e. it will be a prompt to update it)
Approved-by: Sophronia Koilpillai
Approved-by: Milad Mohtashamirad
Approved-by: Lyall Beveridge
The reason for this change is to ensure the wifi module is inserted before morse.
This reintroduces a change which we 'lost' in the rebase as theoretically
not necessary, but found it introduced a regression.
Putting the module this early in the load ensures that phy0 is reliably allocated
to the mt76 module; we were finding that it was variable due to mt7603/morse being next to each other alphabetically, which meant that the LED (set to phy0-ap0)
would sometimes not come up.
Ideally, we would have either a generic way of determining module order for
a particular board, OR make the LED code smarter such that you could
specify a device path or similar instead of an ifname, but both of these
are large changes that do not belong in a point release.
Approved-by: Sophronia Koilpillai
Approved-by: Evan Benn
SW-12275 add support for releasing matter with openwrt
* Update bluetooth make components (other.mk) with correct config option
* Add matter defconfig for all ekh boards
Approved-by: Cedric Van den Bergh
Approved-by: James Haggerty
Approved-by: Julia Page
libiwinfo changes to support s1g need to be propagated through rpcd
so applications sharing information over ubus can receive the right
values.
Approved-by: James Haggerty
Previously, this was done in the theme. With the new Argon theme,
we didn't have this, and it was ugly to either copy the banner
over the top (since base-files 'owns' the banner) or mutate
the existing banner.
Doing it this way means:
- we see this banner regardless of the installed theme
- we include the version info
- there's no confusing banner over-writing/mutating
But has the disadvantage that we carry an extra commit
that we'll never upstream.
e.g.
-----------------------------------------------------
OpenWrt 23.05.2, Morse-2.6-dev
-----------------------------------------------------
Approved-by: Arien Judge
Approved-by: Milad Mohtashamirad
Modifies the OpenWrt device tree LED handler to allow
a third-party package to install custom LED handling.
Approved-by: Sophronia Koilpillai
Approved-by: Evan Benn
This causes DHCP renewals when you disconnect/reconnect cables or
re-establish wifi connections on a bridge.
https://github.com/openwrt/netifd/pull/35
I think we want this, but it's a bit of a scary change that I
might be missing some implications for, so I'd like to get it
in now to give us some time to experience it.
Approved-by: Sophronia Koilpillai
Approved-by: Evan Benn
