Re: [RFC 06/14] irq_domain/powerpc: Eliminate virq_is_host()

[Grant Likely <grant.likely@secretlab.ca>]

On Thu, Jan 12, 2012 at 3:17 AM, Milton Miller <miltonm@bga.com> wrote:
> On Wed Jan 11 2012 about 15:24:34 EST, Grant Likely wrote:
>> There is only one user, and it is trivial to open-code.
>
> I added virq_is_host because I had planned to change how we find the
> host (domain) from a virq.
>
> Instead of storing a pointer in irq_desc (a pointer for every irq),
> I planned to use a NR_IRQ(+extra) bitmap per domain.  This should be
> a win storage-wise when the number of irq controllers is less than the
> number of bits in a long.
>
> This would also convert scanning for a reverse map from walking every
> irqdesc to walking find_next_bit over the irqs assigned to the host.
>
> Thus my rule was "code outside kernel/irq must not touch domain";
> both the contents and how it was associated were abstracted.
>
> Other planned changes included splitting the reverse lookup into
> domain dependent pieces, creating the ida for sparse map at domain
> creation time (init irq is after radix_tree_init as its used by the
> current irq code) so we never fall back to linear search.  Linear
> populated the reverse map as the irq was assigned, and changed to
> a seperate subtype if it mapped an irq above the map size.
>
> I thought some of the domains would be split into seperate files
> selected by Kconfig, at least the sparse tree.
>
> There was also a nomap varient to handle iseries (and one of the cell
> varients) where the interrupt number to use for an event is controlled
> by the guest, that led to the discussion with tlgx about how to
> disallow the extra irqs above the limit set by the arch callback.
>
> Actually making virq-is-domain a domain callback could eliminate the
> need for the bitmap on legacy (range limited) domains.
>
>
> I have my work in progress patches from 2.6.39 if you would like to
> see them.  I was trying to clean up powerpc before pushing it over,
> and didn't get all the concepts written.  So I just sent Ben what
> was ready at the time and haven't had time to get back to it.

Go ahead and send me your patches.  I'll see what I can add into my series.

> Overall, I think most of the other concepts are ok, although I would
> have chosen to remove NO_IRQ before moving the code, and probably
> the rename from host to domain.

Haha.  Yeah, I debated reordering those a bit, but it would have
created a bunch of rebase work for little benefit.  Unless someone
really complains loudly, I'm going to leave it in the order as is.

>  I haven't studied the patches in
> detail since your tree is based on linux-next and my drive doesn't
> have space for that.  It took me a while to realize the code removed
> from the header file in 4/14 (powerpc use commmon host) was actually
> moved to irq-domain in 1/14 (a comment to that effect would be nice).

Okay, I'll add a comment to that effect.  If I rebase to Linus' tree
after 3.3-rc1 is released, can you help me test?

g.