Fixes a regression from 0c11cea where local composite actions
(.github/actions/) were referenced with relative paths. When spins
reuse our reusable workflows, GitHub resolves relative action paths
in the context of the calling repo, not infix — breaking the build.
Replace all uses of ./.github/actions/{cache-restore,podman-cleanup}
with kernelkit/actions/{cache-restore,podman-cleanup}@v1 (new repo).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Replace ERROR() with ERRNO() in all error paths following POSIX API
calls (fopen, rename, realloc, fmkpath, etc.), removing redundant
manual strerror(errno) formatting where present.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Before netd starts confd might want to create .conf file for it to run,
so we sort of depend on /etc/net.d existing at boot.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Set dhcp-authoritative so clients with unexpired leases are not NAKed
after a router crash or factory reset that clears the lease database.
Drop the untagged suppress directives for options 3 (router) and 6
(dns-server): they conflict with the global options confd writes to
dnsmasq.d/, causing spurious "Ignoring duplicate dhcp-option" warnings.
Per-subnet tagged options already suppress auto-detection, and without
an active dhcp-range there is nothing to suppress anyway.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Ensure we translate the hostname before we generate the udhcpc command
line options, otherwise we'll send '-x hostname:rpi-%m' to udhcpc.
Also update all board specific factory configs with the missing setting
'"value": "auto"' for the DHCP hostname option. This to match what we
already do when we infer options when enabling a DHCP client in the CLI.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Two races could prevent DHCP-learned default routes from being installed
at boot:
1. The signal from the DHCP client script could be lost leaving conf.d
updated but frr.conf stale.
2. Even when the signal was received, 'vtysh -b' could fail because the
FRR daemon chain (mgmtd→zebra→staticd) writes PID files before being
fully operational, causing netd to give up with no retry.
Fix both by refactoring netd to use libev:
- Use an inotify watch of CONF_DIR, so netd reacts directly to file
changes without depending on signal delivery.
- On backend_apply() failure, schedule a retry in 5s and call pidfile()
unconditionally so dependent services are not blocked waiting for the
Finit condition 'pid/netd' to be satisfied.
We also take this opportunity to rename /etc/netd/conf.d/ to /etc/net.d/
The extra /etc/netd/ directory level served no purpose — nothing else
lives there. Flatten to /etc/net.d/ which reads more naturally as the
drop-in directory for network configuration snippets.
Also, reduce logging a bit since each netd backend already logs success
or fail which is sufficient to know that a configuration change has been
applied or not.
Finally, with the new inotify processing in netd it's redundant to call
'initctl reload netd' from the udhcpc script, in fact it will only cause
unnecessary overhead, so we drop it.
Fixes#1438
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Add a must expression to ensure users do not set static neighbors when
the interface type is not point-to-multipoint or non-broadcast.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Add board support for the BananaPi BPi-R64 router board, based on the
MediaTek MT7622 SoC and MT7531 Gigabit Ethernet switch.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In a hybrid MBR the conventional layout (used by gdisk and others) is:
MBR[0] = 0xEE (GPT protective, covers whole disk)
MBR[1..3] = user-selected GPT partitions with legacy type codes
genimage currently appends the 0xEE entry after the data partitions,
producing the opposite order. This is fine for most platforms, but
breaks firmware that detects GPT by inspecting only MBR entry 0: seeing
a non-0xEE type there it falls into pure-MBR mode and cannot look up
GPT partitions by name.
The TF-A partition driver (drivers/partition/partition.c) exhibits
exactly this behaviour: load_mbr_header() copies entry[0] and checks
type == 0xEE to select between MBR and GPT mode. The MediaTek MT7622
platform (BananaPi BPI-R64) is a concrete example where this matters.
On the other hand, Raspberry Pi firmware scans all MBR slots for a
bootable FAT32 partition. If slot 0 holds a 0xEE protective entry the
firmware switches to GPT detection, looks for an EFI System Partition
UUID, and reports "no bootable partitions" when none is found.
Add an opt-in hdimage option `gpt-protective-first = true` that places
the 0xEE entry at slot 0 and shifts data partitions to slots 1..3.
The default (false) preserves the existing genimage behaviour so that
platforms like Raspberry Pi continue to work without any changes.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The Raspberry Pi 400 shares hardware with the RPi 4B but differs in one
important way: it has an internal keyboard and mouse connected through a
VIA Labs USB hub on the VL805 xHCI controller. Break the symlink to
raspberrypi,4-model-b and maintain a dedicated product directory.
Factory config changes from the 4B default:
- USB1/USB2/USB3 all unlocked: USB2 carries the internal keyboard hub
and USB3 the superspeed companion; all are unlocked as this is a
desktop device
- DHCP vendor-class corrected to "Raspberry Pi 400"
Also add interface-quirks.json for the eth0 PHY detach behaviour and
the smsc95xx broken flow-control quirk (inherited from the 4B config
but now explicit), and remove the spurious bootable flag from the
primary and secondary rootfs partitions in genimage.cfg.in.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Remove console=map:0 from the [pi4] section which was suppressing HDMI
output and disabling the display getty. Making it really difficult to
use it as a regular console.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Two related problems prevented devices behind an intermediate hub (e.g.
the VIA Labs hub on the RPi 400's VL805 xHCI) from being authorized
when confd unlocks a USB bus:
1. Cascade ordering: usb_authorize() ran nftw() with FTW_DEPTH, which
visits children before the root-hub entry. The hub was authorized
while authorized_default was still 2, so when the kernel probed the
hub's children it immediately denied them. Fix: write
authorized_default=1 on the bus before entering nftw.
2. Slow/async probe: hub port enumeration is asynchronous — devices
behind a hub may appear in sysfs after nftw has already finished.
Fix: add a udev rule that catches usb_device add events on buses
where authorized_default=1 is already set on any ancestor root hub
and authorizes them immediately.
The ATTRS{} matcher in udev walks the full sysfs parent chain, so
authorized_default=1 on the root hub is visible even for devices several
hubs deep. Buses still locked at authorized_default=0 are left alone.
Also expand the docstring in generic_usb_ports() to document the
design: each root hub gets its own uniquely-named entry (USB, or USB1/
USB2/... when multiple) so buses can be individually controlled, and
boards needing finer control should use DT usb-ports annotations.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Replace the costly read-only filter loop (awk+tr per line) with a
stderr redirect, use /sys/class/net/ instead of ip+jq, and batch
all sysctl writes into a single call.
Reduces boot time by ~4s on 32-bit Arm systems.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Calling tune2fs for ext4 partitions at boot costs more than one second
boot time on 32-bit Arm systems, with very little gain. Dropping this
restores the default periodic fsck *and* saves boot time.
Instead of calling sgdisk four times, call it only when resizing and
instead use sysfs to find named paritions. Saves seconds on boards
with weaker CPU and slow media.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The MVEBU SafeXcel Crypto Engine firmware (linux-firmware minifw) was
originally added in 70c12c3 to the generic aarch64 defconfigs to silence
kernel probe failures on Marvell Armada SoCs (37xx, 7k, 8k, CN913x):
crypto-safexcel f2800000.crypto: Firmware load failed.
crypto-safexcel f2800000.crypto: HW init failed (-2)
It was then accidentally dropped in 0e2d12e (kernel upgrade to 6.18),
which rebased on a tree that predated the firmware addition.
Rather than restoring it to the generic defconfigs, move it to the two
Marvell board Config.in files where it actually belongs — consistent
with how RTL8169 firmware was handled for the Raspberry Pi CM4 IoT
Router Board Mini in 68313773.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The awk insertion path used getline to peek at the line right after the
"### Changes", but only printed it when NF == 0 (blank line). If the
section already had a non-blank entry (e.g. a Buildroot upgrade line),
getline consumed it silently and the kernel line was written in its
place.
Fix by adding the missing else branch so the consumed line is always
re-emitted — blank lines before the new entry, non-blank lines after it.
Also demote the missing-UNRELEASED guard from exit 1 to a warning with
return 0, so the workflow doesn't abort when a new release cycle hasn't
had its ChangeLog section opened yet.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Add three SSH-related commands to the operational CLI:
ssh [user <name>] [port <num>] <host>
Connect to a remote device over SSH, running as the
CLI user (not root) by dropping privileges before exec.
set ssh known-hosts <host> <keytype> <pubkey>
Pre-enroll a host public key received out-of-band (e.g.
via email after a factory reset) into ~/.ssh/known_hosts,
avoiding a TOFU prompt on first connect.
no ssh known-hosts <host>
Remove a stale host key entry using ssh-keygen -R, e.g.
after a device factory reset causes a key mismatch.
Tab completion is provided for key types (ssh-ed25519,
ecdsa-sha2-nistp256, etc.) and for known host names/IPs.
A new run_as_user() helper is introduced alongside the existing
run(), factoring out the fork+setuid+execvp pattern used by
infix_shell() so it can be shared across the SSH functions.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A legacy name length limit in firewalld triggered problems with longer
policy names. This patch to firewalld lifts that limit by checking the
backend in use, no limit for nftables.
Fixes#1389
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Always emit the "1..N" plan line before exiting with error so test
harnesses don't report "test error, no plan" for failed or aborted
tests.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Increase Zebra NETLINK buffer to fix reported issues with zebra being
out of sync with kernel routes and interface changes.
Also, try '-a' option to allow other processes to delete zebra routes.
Other changes in this commit:
- Minor fixes to Finit service files for consistency
- Empty daemon.conf stub files to silence bogus mgmtd errors at startup
- Relocate all skeleton files from board/common to separate from files
that are actual Infix additions and what's package system integration
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
network.local is a discovery/browse page, not a management interface.
Serving it over HTTP eliminates the self-signed certificate warning
browsers show for .local mDNS names without changing the security
posture of the actual device management pages (still HTTPS-only).
The new port-80 server_name block takes precedence over the catch-all
HTTP → HTTPS redirect in default.conf.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
avahi-browse returns 127.0.0.1 (or ::1) when resolving services on the
same machine netbrowse is running on. These addresses are meaningless
for display and misleading as click targets. Skip the entire 127.x/::1
range and link-local (fe80:) when choosing the preferred address for a
host card; the card addr field is simply omitted if no routable address
is seen.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- IP address shown in soft parens on the card header line, e.g.
"infix.local (192.168.1.10)"; IPv4 preferred over IPv6, link-local
skipped
- Product name and OS version (from mDNS TXT records "product=" and
"ov=") shown as a secondary line below the header when available
- Theme toggle icons changed from ◐/●/○ to ◐/☽/☀ (system/dark/light)
- browse.go: introduce Host struct (addr, product, version, other, svcs)
replacing the flat []Service map; scan() now returns map[string]Host
- Search also matches against IP, product, and version fields
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Apple devices (e.g. Apple TV) publish vv=1 in their AirPlay/RAOP TXT
records, causing a false positive with our Infix platform marker.
Tighten the filter to require vv=1 together with at least one management
service type (ssh, https, http, netconf, restconf), which Apple devices
never advertise. Keep on=Infix as a fallback for older firmware that
predates vv=1.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
avahi's parseable output uses \DDD decimal escape sequences, not octal.
The previous decoder treated them as octal, causing digits 8 and 9 to
fall through undecoded (e.g. \058 → ':' and \091 → '[' were shown
literally). Also handle \X non-digit escapes (e.g. \. → '.') used for
dots in service instance names.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Add a `hostname` leaf to the `mdns` YANG container with default `"%h"`,
allowing operators to override the avahi host-name used for mDNS A/AAAA
records without reflashing. The default expands to DEFAULT_HOSTNAME from
os-release, preserving existing behaviour for unconfigured deployments.
Format specifiers %h/%i/%m are supported via the existing hostnamefmt()
infrastructure, which is also fixed to copy the const fmt argument to a
local buffer before modification (UB when called with a libyang-owned
string).
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
