Features
-
lager usb <net> cyclepower-cycles a port — off, wait, on — for every supported USB hub.--off-timesets how long the port stays unpowered (default 1s, range 0.5-10s), set above the slowest cold boot measured on real hardware. Too short an off time is the failure that matters: the DUT’s rails do not fully discharge and it warm-starts while appearing to have been reset. Prefer it over a scripted disable/sleep/enable — it holds the hub for the whole sequence so nothing else can switch the port while it is dark, restores power on every failure path, and reports whether the device came back. -
lager usb <net> recoverre-powers a port left unpowered by an interrupted command. On hubs where the whole physical device can be identified it re-asserts power on every port, since the reason to reach for it is usually that something is off and it is not obvious what. -
Plugable RTS5411 USB docks are supported as switchable-USB-power
instruments, a third option alongside Acroname and Yepkit hubs. Plugable
ships no SDK, so control is the standard USB hub class per-port power
switching rather than a vendor library. Only the four external Type-A sockets
switch VBUS; the dock’s internal tier (billboard, audio, ethernet) does not,
even though both tiers advertise per-port switching identically. Requires the
udev rule for vendor
2230, shipped in99-instrument.rules.
Bug Fixes
-
Disabling a USB hub port no longer reports a false failure and undoes
itself. A powered-off port raises no change notification, so the kernel never
processes the disconnect: the device keeps its
lsusbentry and its/devnodes for the whole off window. The driver had treated that as proof the hardware could not switch power, and then powered the port back on. It now checks the hub’s own power bit, which is the only thing observable while a port is off. Never test for a device’s absence to decide whether a port is off — usestate, or compare a device’s USB device number either side of acycle. -
lager diagnoseno longer reports a non-Acroname USB hub as wedged. Hub diagnostics are vendor-specific but every USB net reached them, so a hub from another vendor was opened with the wrong driver and the failure reported as an electrical fault. Such a hub now reportsNOT SUPPORTED, kept distinct from the transientBUSYstate, and its bus facts are still shown. -
PlugableUSBNetis importable from the Python API, andYKUSHUSBNetis the YKUSH driver again. A copy-pasted entry in the driver export table had given one driver the other’s name.