On 10/01/2024 10:26 am, Jan Beulich wrote: > On 10.01.2024 10:53, Roger Pau Monne wrote: >> The HVM pirq feature allows routing interrupts from both physical and >> emulated >> devices over event channels, this was done a performance improvement. >> However >> its usage is fully undocumented, and the only reference implementation is in >> Linux. It defeats the purpose of local APIC hardware virtualization, because >> when using it interrupts avoid the usage of the local APIC altogether. > So without sufficient APIC acceleration, isn't this arranging for degraded > performance then?
Maybe. Quite possibly. > IOW should the new default perhaps be dependent on the > degree of APIC acceleration? No. Getting things working is strictly more important than making things fast. HVM_PIRQ is an entirely undocumented pile of Xen-specific deviation from standards, with a single reference implementation and 0 people who understand it. It has also been shown to be the cause of real breakage, and disabling it has been shown to make guests function better. If and when we have two working schemes, we can choose a default based on performance. We are not in this position right now. ~Andrew
