Re: [RFC PATCH 13/18] aarch64: accept linux,initrd-* DTB nodes as bootstrap modules
Sergey Bugaev <[email protected]>
| Newsgroups | gmane.os.hurd.bugs |
|---|---|
| Message-ID | <CAN9u=HfkVrnd_MT7GuKwnd2EtTDpRXessBviennSxpbB=PfZRQ@mail.gmail.com> |
On Sun, May 24, 2026 at 4:44 AM Paulo Duarte <[email protected]> wrote: > > The existing load_boot_modules_from_dtb() reader walks > /chosen/multiboot,module nodes — a hand-rolled convention from > bugaevc/wip-aarch64 designed for bare-metal multi-module Hurd > boot via u-boot's `fdt mknod`. No off-the-shelf bootloader > emits it, which is why our test harness has been using QEMU's > -device guest-loader to synthesise it. > > Add a fallback for the standard arm64 boot protocol: when no > multiboot,module nodes are found, look for /chosen/linux,initrd- > start + linux,initrd-end + /chosen/bootargs and treat the pair > as a single bootstrap module. Single-module by design — the > standard protocol has only one initrd slot. Production > multi-module boots stay on the multiboot,module path above. But how are you going to use it, with a single module and no args? How would Mach even know what to do with the module, whether to task-create & task-resume on it, or what? Can you show the intended use case?