Hello,

I have again seen reports about the hpet_addr panic.

Damien Zammit, le dim. 29 mars 2026 09:15:02 +0000, a ecrit:
> Well OK, but I think every Intel x86 machine since 2005 that has acpi tables 
> and an APIC also has a HPET.  They are independent things, sure, but maybe 
> virtual machines emulating x86 should enable it by default to be a compatible 
> machine?

You are trying to fight the rest of the world.

That won't work.

> It is on by default in qemu, for example.
> Alternatively, we need to do more work to calibrate another timer just for 
> accurate delays.

We don't really need another timer. The pit timer is fine enough for
most uses. It's only a few software which require sub-millisecond
timing. If some people really want to run such software, they can look
at adding high-precision clock in their VM. But we shouldn't make it
mandatory just for the sake of it.

Samuel

> -------- Original Message --------
> On 29/3/26 7:44 pm, Samuel Thibault <[email protected]> wrote:
> 
> >  Hello,
> >  
> >  As reported on #debian-hurd and https://bugs.debian.org/1132186, using a 
> > VM in virt-manager leads to
> >  
> >  Crash with "hpet_init: Assertion 'hpet_addr != 0' failed.panic"
> >  
> >  We really shouldn't assume that APIC implies HPET.
> >  
> >  Samuel

Reply via email to