Re: [PATCH+DISCUSSION] irqchip: armada-370-xp: Remove redundant ops assignment

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

On Wed, 2019-06-12 at 15:16 +1000, Benjamin Herrenschmidt wrote:
> pci_msi_create_irq_domain -> pci_msi_domain_update_chip_ops will
> set those two already since the driver sets MSI_FLAG_USE_DEF_CHIP_OPS
> 
> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> ---
> 
> [UNTESTED]
> 
> Just something I noticed while browsing through those drivers in
> search of ways to factor some of the code.
> 
> That leads to a question here:
> 
> Some MSI drivers such as this one (or any using the defaults
> mask/unmask
> provided by drivers/pci/msi.c) only call the PCI MSI mask/unmask
> functions.
> 
> Some other drivers call those PCI function but *also* call the parent
> mask/unmask (giv-v2m for example) which generally is the inner domain
> which just itself forwards to its own parent.

  .../...

So I looked at x86 and it also only uses pci_msi_unmask_irq, it doesn't
mask at the parent level. And it also specifies those explicitly which
isn't necessary so the same trivial cleanup patch could be done (happy
to do it unless I missed something here).

Question: If that's indeed the rule we want to establish, should we
consider making all MSI controllers just use the PCI masking and remove
the forwarding to the parent ?

The ones that do the parent, at least in drivers/irqchip/* and
drivers/pci/controller/* (ther are more in arch code) are all the GIC
ones (v2m, v3-its, v3-mbi), alpine which was copied on GIC I think,
tango and dwc.

The other approach would be to make the generic ops setup by
pci_msi_domain_update_chip_ops call the parent as well .. if there is
one and it has corresponding mask/unmask callbacks. That means things
like armada_370 would be unaffected since their "middle" irqdomain chip
doesn't have them, at least until somebody decides that masking at the
parent level as well is a good thing. I *think* it would also work for
x86 since the parent in that case is x86_vector_domain which also
doesn't have mask and unmask callbacks, so it would be a nop change.

Let me know what you think.

Cheers,
Ben.