On 30/06/2026 00:18, Samuel Thibault wrote:
Michael Kelly, le lun. 29 juin 2026 06:27:30 +0100, a ecrit:
Using e1000e with netdde (0.0.20200330-13) does indeed fail on hurd-i386 and
hurd-amd64. The latter hangs after issuing the DHCPREQUEST whilst the former
like Jason reported. e1000e worked in the previous netdde. I hadn't been
aware of this Qemu NIC option but I see that Linux 'Virtual Machine Manager
5.0.0' now lists e1000e rather than e1000 which I think might be a recent
change on my host.
The workaround is to substitute e1000e for e1000 but fixing this properly is
now number 1 on my list.
Should version 12 be restored as the latest 'sid' for the time being ?
It's not that easy to do. The archive versions are monotonic, so it'd
mean reuploading libacpica with some patches reverted, then re-build
netdde against that newer version.
Honestly, if we can rather just upload a fix that'd be best.
I have a fix that allows e1000e to intialise and Hurd to boot. There are
changes to libacpica so that the PCI device/function parameters are
adjusted as specified by the _ADR of the subordinate device. This was
missing functionality which is required by PCIe devices.
I have tested on hurd-amd64 VM using rtl8139, e1000 and e1000e and my
hurd-amd64 hardware platform with Atheros. hurd-i386 VM is next on the
list.
I'd like to test as many network interfaces as possible. Are there other
ones offered by Qemu that Hurd supports using netdde ?
Thanks,
Mike.
