Re: [patch 04/32] genirq/msi: Provide a set of advanced MSI accessors and iterators

From: Jason Gunthorpe
Date: Mon Nov 29 2021 - 09:03:18 EST


On Mon, Nov 29, 2021 at 10:26:11AM +0100, Thomas Gleixner wrote:
> Jason,
>
> On Sun, Nov 28 2021 at 20:22, Thomas Gleixner wrote:
> > On Sat, Nov 27 2021 at 21:00, Jason Gunthorpe wrote:
> >> On Sat, Nov 27, 2021 at 02:22:33AM +0100, Thomas Gleixner wrote:
> >> I understand why that isn't desirable at this patch where the storage
> >> would have to be a list_head pointer, but still, seems like an odd
> >> place to end up at the end of the series.
> >>
> >> eg add index here unused and then the last patch uses it instead of
> >> __iter_idx.
> >
> > TBH, didn't think about doing just that. OTH, given the experience of
> > looking at the creative mess people create, this was probably also a
> > vain attempt to make it harder in the future.
> >
> >> Also, I don't understand why filter was stored in the dev and not
> >> passed into msi_next_desc() in the macro here?
> >
> > No real reason. I probably just stored it along with the rest. Lemme try
> > that index approach.
>
> After looking at all the call sites again, there is no real usage for
> this local index variable.
>
> If anything needs the index of a descriptor then it's available in the
> descriptor itself. That won't change because the low level message write
> code needs the index too and the only accessible storage there is
> msi_desc.

Oh, that makes it simpler, just use the current desc->index as the
input to the xa_for_each_start() and then there should be no need of
hidden state?

> What for? The usage sites should not have to care about the storage
> details of a facility they are using.

Generally for_each things shouldn't have hidden state that prevents
them from being nested. It is just an unexpected design pattern..

Jason