Hi:
We get the same issue (or pretty similar) in real hardware, in many
Thinkpad models with Core2Duo.
And the already known problem with SMP even in more modern Thinkpad models
(probably the problem keeps in some other computers, but I don't have
another PC to test it).
I think that the problem is close related to the Qemu issue. In my Thinkpad
T410, the rumpdisk works in "up" gnumach, but doesn't works in "smp"
version.
In previous model, like T60, as I previosly said, rumpdisk doesn't detect
the disk neither in AHCI or "compatibiliy" (IDE emulation), even with
gnumach-up.
We need a deep debugging to solve all this problems, with the objective to
get test gnumach-smp in real hardware.
El sáb, 21 jun 2025 a las 7:32, Damien Zammit (<dam...@zamaudio.com>)
escribió:
> Hi,
>
> As discussed, it would be good to fix piixide driver when i440fx chipset
> is used by default in qemu ("-M pc").
>
> Currently we get "lost interrupt" messages appearing from rumpdisk and
> the disk access is not working.
>
> I tried compiling gnumach **without** --enable-apic and bypassing the
> acpi translator startup. In this case, piixide driver works correctly.
>
> I think there is some kind of disconnect between I/O APIC in qemu and
> the PIIX3 IDE controller. (Maybe they cannot be used concurrently?)
>
> Alternatively, if you really want to use -M pc default machine, you can
> attach an ich9-ahci controller and connect the primary disk to this
> controller as a SATA disk:
>
> -drive id=disk,file=/path/to/disk.img,if=none \
> -device ich9-ahci,id=ahci \
> -device ide-hd,bus=ahci.0,drive=disk \
>
> This works without specifying a -M option with acpi and rumpdisk with an
> --enable-apic gnumach, by using an AHCI controller instead of a PIIX3
> IDE controller.
>
> In other news, we still have a bug with rumpdisk timing out on AHCI->wd0
> on native controllers. I am not sure what causes this but I suspect
> some kind of race condition between startup translators.
>
> Damien
>
>
>
