luci/modules/luci-base/root/usr/share, branch master

luci-base: drop init_action() and the luci.setInitAction ubus method

2026-08-03T09:58:22Z

Both duplicate rpcd's rc ubus namespace, which exposes the same enable/disable/start/stop/restart/reload set through rc.init and is what downstream code wishing to control services should be using. The duplicate was also the weaker of the two. init_action() interpolated both of its arguments into a shell command run as root: system(`env -i /etc/init.d/${basename(name)} ${action} >/dev/null`); basename() kept name inside /etc/init.d, but it strips directory components only - `;`, backticks, `$(`, `|` and newlines survive it - and action reached the shell with nothing checking it at all. Nothing untrusted got there: setInitAction switched on a fixed action list, and for name the stat() gate required the interpolated string to exist as a file under /etc/init.d, an existence check rather than a validation. But init_action() was exported from luci-base, so any future caller forwarding an RPC argument straight through inherited root command execution. rc.init has no shell to escape from - it execs the script - and it additionally refuses scripts that are not root-owned or that are world-writable. The in-tree users are ported in the preceding commits. Assisted-by: Claude:claude-opus-5 Signed-off-by: Julius Bairaktaris

luci-base: set password also for rpcd user

2026-06-03T10:49:07Z

This change allows the root user to change also rpcd logins via administration tab in system. This changes `setPassword` in a way that existing calls to it fall back to change unix passwords. The goal of this commit is to prepare the environment to add password policy (password length, valid property, special characters, etc) to the administration tab. Signed-off-by: Christian Korber

luci-base: replace ufpd access check with service run check

2026-05-18T07:48:13Z

It is not enough simply to check whether the tool is installed. It is essential to check whether the service is actually running. After all, we need the functionality provided by the service. Signed-off-by: Florian Eckert

luci-mod-network: check kernel for mptcp to display MPTCP options

2026-04-04T17:12:54Z

https://www.kernel.org/doc/html/v6.12/networking/mptcp-sysctl.html There may be user instances where mptcp was not compiled into the kernel or is an unloaded module, so gate the options behind the feature check. Closes #8522 Signed-off-by: Paul Donald

luci-base: fix ubus getFeatures() crash

2026-04-01T11:11:14Z

OpenWrt versions that provide odhcpd versions reporting a compile time features line in their help output trigger a crash in the getFeatures() procedure (attempt to set property on boolean value) - fix this issue by not predeclaring the odhcpd flag as `false`. Fixes: #8489 Fixes: 0529dca7a2 ("luci-base: fix feature detection patterns") Signed-off-by: Jo-Philipp Wich

luci-base: fix feature detection patterns

2026-03-30T13:08:45Z

Upstream ucode commit 8a8b4d1 ("types: fix regexp `/.../s` flag inversion") changed the semantics of /.../s regexp patterns to match the ECMAScript behavior, meaning `/s` enables single line matching and omitting it performs line-wise matching. Besides changing the behavior of `.` to match newlines, the `/s` flag also changes `^` and `$` to only match the begin and end of the entire string input, respectively instead of matching the start and end of lines. This broke feature pattern matching in the luci rpcd plugin, causing LuCI to incorrectly disable certain features. This commit changes the affected places to read the command outputs linewise and to perform the feature flag matching against each read line individually, including an explicit terminating newline to make the pattern match forward- and backward compatible with fixed and unfixed ucode versions. Fixes: https://github.com/jow-/ucode/issues/389 Signed-off-by: Jo-Philipp Wich

luci-base: resolve symlinks for PCI hardware

2026-03-16T14:54:07Z

getBuiltinEthernetPorts crashed with "Unknown error" on hardware where NICs have PCI subsystem symlinks at different sysfs depths (e.g. onboard NIC vs PCIe card NIC). The getBuiltinEthernetPorts x86_64 workaround block needs realpath() to resolve PCI subsystem symlinks. readlink() returns different strings ("../../../bus/pci" vs "../../../../bus/pci") even though both resolve to /sys/bus/pci, causing eth1 and eth2 to be incorrectly excluded from the result. E.g.: eth0: /sys/class/net/eth0/device/subsystem -> ../../../bus/pci eth1: /sys/class/net/eth1/device/subsystem -> ../../../../bus/pci eth2: /sys/class/net/eth2/device/subsystem -> ../../../../bus/pci 1. Add realpath to the 'fs' module import statement 2. Use realpath() to normalize paths Closes #8425 Reported-by: @MI2-2 Signed-off-by: Paul Donald

luci-base: load luci preferences early

2026-02-03T19:30:37Z

Follow-up to 315dbfc7498e2f43afb0119b992915e8f311bc37 In order to look at any saved uci value, the luci config must be loaded before we render the page, so we can determine stored preferences, for things like table filter, without encumbering all consumers of configurable elements to load the luci uci configuration. All (Table) elements extend CBIAbstractElement, so load there. Signed-off-by: Paul Donald

luci-base: poe and PSE details & configuration

2026-01-26T02:40:38Z

Adds PoE/PSE configuration support for modern linux (PSE-PD). This change is based on the PSE-PD backport (from 6.17) and netifd|ubus changes. * Add getPSE() [receive all status information] and hasPSE() [has the device PSE hardware?] * Changes ACL permissions to query network.device status information * Add two new PoE icons (PoE active with link up + link down) for the port status page * Changes port status to show PoE information, next to link information and data transfer. * Add a new tab for PoE/PSE to the device configuration, which will only be displayed if the device supports PSE Signed-off-by: Carlo Szelinsky

luci-base: rpc: fix luci/getMountPoints crash

2026-01-09T20:52:18Z

getMountPoints crash on path containing '\xxx'. `char` should be `chr` Fixes https://github.com/openwrt/openwrt/issues/21459 Signed-off-by: Liangbin Lian