Re: [PATCH] RFC: kills consistent_dma_mask

[Krzysztof Halasa <khc@pm.waw.pl>]

Pete Zaitcev <zaitcev@redhat.com> writes:

> Are you talking about doing tripple calls, e.g.
> 
>        pci_set_dma_mask(pdev, 0xFFFFFFFF);
>        foo = pci_alloc_consistent(pdev, size, &handle);
>        // Restore for upcoming streaming allocations
>        pci_set_dma_mask(pdev, 0xFFFFFFFFFFFFFFFF);
> 
> Possibly Jes considered that alternative and decided that it
> did not allow for sufficient performance.

Possibly. Is that true?

I could imagine even something like that:

init_module()
{
        using_dac = 1;
        if (!pci_dma_supported(dev, 0xFFFFFFFFFFFFFFFF)) {
                if (!pci_dma_supported(dev, 0xFFFFFFFF))
                      	error;
                using_dac = 0;
        }
}

...

        foo = pci_alloc_consistent(pdev, size, &handle, 0xFFFFFFFF);
        bar = pci_map_single(...,
                using_dac ? 0xFFFFFFFFFFFFFFFF : 0xFFFFFFFF);

I don't think it would be slower. If inlined, if would be even faster.

However, the main problem is not that it isn't beautiful but rather that
it's broken.

> Before you go for that, I'd rather see you implementing the
> double/tripple calls in drivers, check for effects, THEN
> go for removal of the mask.

The problem is that the official kernel does NOT contain any driver which
needs different masks.

> > This patch doesn't actually change any current kernel behaviour.
> 
> Sure it does. It blows all non-mmu ia64 out of the water.

No. The kernel (2.6.0-test3 at least) doesn't count on that under any
circumstances.

> The consistent mask looks a little distasteful to me, and I think
> it should not buy us performance because consistent allocations
> are not supposed to be fast. They are bad, but what you are doing
> is worse: you are trying to ruin the day of legitimate users.

Of course this isn't what I'm trying to do.
-- 
Krzysztof Halasa
Network Administrator