Skip to content

Flashing names what is in the way

Flashing now waits up to 15 seconds for the module to show up on USB, and when it does not, the error lists the other USB devices on the node and says a card in the mini-PCIe slot may be the cause. A failed or cancelled flash powers the node off and puts the USB route back.

Firmware v2.44.0, 15 October 2026 — the first image to carry bmcd 2.40.0. Every item below is the repository's own changelog entry, taken apart; the whole entry is a click away under each one.

  • Flashing a node waits up to 15 s for the module's USB boot device. Instead of looking once, a second after power-on, and giving up.

    The whole entry

    A module that enumerates slowly, or that fails high-speed enumeration and is retried by the kernel, was refused with "Compute module's USB interface not found or supported"; the daemon now looks every 250 ms. Found from a user report on Discord.

  • When the boot device does not appear, the error names what is in the way. It lists the other USB devices seen on the node's lines (root hubs excluded), with a hint: a USB device in the mini-PCIe slot of node 1 or 2 shares the USB lines used for flashing, so remove it, or flash the module in slot 3 or 4.

  • Flashing waits up to 15 s for the module's USB boot device, and names what is in the way when it does not appear. It used to look at the bus once, a second after power-on, and give up with "Compute module's USB interface not found or supported". bmcd 2.40.0

    The whole entry

    A module that enumerates slowly, or that fails high-speed enumeration and is retried by the kernel, was refused for that. It now looks every 250 ms. If nothing turns up the error says so for the node (numbered as users see it) and lists the other USB devices seen on the node's lines, root hubs excluded, with a hint that a device in the mini-PCIe slot of nodes 1 and 2 shares the USB pair used for flashing. On a board with a fanout hub a supported device on the wrong port is still reported as such. Found from a user report on Discord, where a Zigbee stick (1a86:7523) in node 2's slot kept an RK1 from enumerating.

  • A cancelled flash stops waiting at once, and a flash that never starts puts the node back. The wait for the boot device honours the cancel token, so a cancel (or a client that gave up) no longer leaves a worker polling for up to 15 s against a restarted flash. bmcd 2.40.0

    The whole entry

    When the device does not appear, the node is powered off and USB restored exactly as after any other failed flash, instead of staying powered in flashing mode while the error tells the user to retry; the original error is the one returned.

  • A failed or cancelled flash now powers the node off and restores the USB route,. And cancelling stops the wait for the boot device at once.

Every release of the firmware · the roadmap · follow by feed

Last updated: