Re: pci_intr_alloc() vs pci_intr_establish() - retry type?

[Jaromír Doleček <jaromir.dolecek%gmail.com@localhost>]

Le mer. 5 déc. 2018 à 08:39, Masanobu SAITOH <msaitoh%execsw.org@localhost> a écrit :
>   I suspect Serial ATA AHCI 1.2.1 specification page 111 has the hint.
>
>   Figure 23: Port/CCC and MSI Message Mapping, Example 1
>   Figure 24: Port and MSI Message Mapping, Example 2
>
> I suspect MSI-X also assume this layout. pci_msix_alloc_map() might
> be used.

Very likely we need to disable MSICAP.MME bit if MSI happens to have
more vectors to force single MSI operation.

There is no such bit for MSI-X. Let's work on this off-list, I have
some idea to test the interrupt mapping, and actually add support for
multiple vectors. Since in your case the thing only fails for
IDENTIFY, it looks like the registers mapped via regular BAR work and
just interrupt routing is bad. Performance wise it doesn't buy much
besides avoiding one register read per interrupt, because the whole
ATA subsystem is not MPSAFE and hence the interrupt handlers are
serialized via KERNEL_LOCK() anyway.

>   My SuperMicro A2SDi-H-TP4F's disk is connected to not channel 0
> but channel 6. I also verified that GHC.MRSM is 0.

GHC.MRSM is only set if hardware has less MSI/MSI-X vectors than AHCI
ports, and fallbacks to single MSI due to this. So it's actually
normal for this to be 0. It'd be quite strange for hardware to have
less MSI/MSI-X vectors than supported ports.

Jaromir