Re: [PATCH] x86: find nr_irqs_gsi with mp_ioapic_routing

[Len Brown <lenb@kernel.org>]

> >> Impact: find right nr_irqs_gsi on some systems.
> >>
> >> one system has gap between gsi.
> >> [    0.000000] ACPI: IOAPIC (id[0x04] address[0xfec00000] gsi_base[0])
> >> [    0.000000] IOAPIC[0]: apic_id 4, version 0, address 0xfec00000, GSI 0-23
> >> [    0.000000] ACPI: IOAPIC (id[0x05] address[0xfeafd000] gsi_base[48])
> >> [    0.000000] IOAPIC[1]: apic_id 5, version 0, address 0xfeafd000, GSI 48-54
> >> [    0.000000] ACPI: IOAPIC (id[0x06] address[0xfeafc000] gsi_base[56])
> >> [    0.000000] IOAPIC[2]: apic_id 6, version 0, address 0xfeafc000, GSI 56-62
> >> ...
> >> [    0.000000] nr_irqs_gsi: 38
> > 
> > I've never seen gaps in GSIs.
> > I've never seen IOAPICs with 7 redirection table entries.
> > 
> > This system looks quite broken even before the probe for nr_irqs_gsi.
> > Do you have more information on it?
> 
> it is one AMD based system from Sun.
> it has two hypertransport chains.
> one: 8132 + ck804
> one: io4
> 
> when 8 sockets are installed, will get
> [    0.000000] ACPI: IOAPIC (id[0x00] address[0xfec00000] gsi_base[0])
> [    0.000000] IOAPIC[0]: apic_id 0, version 0, address 0xfec00000, GSI 0-23
> [    0.000000] ACPI: IOAPIC (id[0x01] address[0xfdefd000] gsi_base[48])
> [    0.000000] IOAPIC[1]: apic_id 1, version 0, address 0xfdefd000, GSI 48-54
> [    0.000000] ACPI: IOAPIC (id[0x02] address[0xfdefc000] gsi_base[56])
> [    0.000000] IOAPIC[2]: apic_id 2, version 0, address 0xfdefc000, GSI 56-62
> [    0.000000] ACPI: IOAPIC (id[0x03] address[0xfeaff000] gsi_base[24])
> [    0.000000] IOAPIC[3]: apic_id 3, version 0, address 0xfeaff000, GSI 24-47
> 
> 
> when 2 sockets are installed, only HT chains is used. will get
> [    0.000000] ACPI: IOAPIC (id[0x04] address[0xfec00000] gsi_base[0])
> [    0.000000] IOAPIC[0]: apic_id 4, version 0, address 0xfec00000, GSI 0-23
> [    0.000000] ACPI: IOAPIC (id[0x05] address[0xfeafd000] gsi_base[48])
> [    0.000000] IOAPIC[1]: apic_id 5, version 0, address 0xfeafd000, GSI 48-54
> [    0.000000] ACPI: IOAPIC (id[0x06] address[0xfeafc000] gsi_base[56])
> [    0.000000] IOAPIC[2]: apic_id 6, version 0, address 0xfeafc000, GSI 56-62
> 
> acpi madt ioapic-entries does include gsi_base field.
> 
> amd 8131: has 4 gsi
> amd 8132: has 7 gsi, but only 3 can be used.
> 
> anyway that system is not broken, and it is quite solid.

I don't know if it violates the ACPI spec to have gaps in the GSI's,
but it does violate my expectations when I read the spec -- so i guess
that "unexpected" or "unconventional" may be a better word than "broken".

At the same time, "unexpected" and "unconventional" BIOS implementations
cause things like "broken" OS -- thus the need for your patch:-)

> this patch seems to be a 2.6.29 material.

I agree.

thanks,
Len Brown, Intel Open Source Technology Center