Re: [PATCH 1/2] mm: Fix struct page layout on 32-bit systems

From: Matthew Wilcox
Date: Tue Apr 20 2021 - 07:33:03 EST


On Tue, Apr 20, 2021 at 09:39:54AM +0200, Geert Uytterhoeven wrote:
> > +++ b/include/linux/mm_types.h
> > @@ -97,10 +97,10 @@ struct page {
> > };
> > struct { /* page_pool used by netstack */
> > /**
> > - * @dma_addr: might require a 64-bit value even on
> > + * @dma_addr: might require a 64-bit value on
> > * 32-bit architectures.
> > */
> > - dma_addr_t dma_addr;
> > + unsigned long dma_addr[2];
>
> So we get two 64-bit words on 64-bit platforms, while only one is
> needed?

Not really. This is part of the 5-word union in struct page, so the space
ends up being reserved anyway, event if it's not "assigned" to dma_addr.

> > + dma_addr_t ret = page->dma_addr[0];
> > + if (sizeof(dma_addr_t) > sizeof(unsigned long))
> > + ret |= (dma_addr_t)page->dma_addr[1] << 16 << 16;
>
> We don't seem to have a handy macro for a 32-bit left shift yet...
>
> But you can also avoid the warning using
>
> ret |= (u64)page->dma_addr[1] << 32;

Sure. It doesn't really matter which way we eliminate the warning;
the code is unreachable.

> > +{
> > + page->dma_addr[0] = addr;
> > + if (sizeof(dma_addr_t) > sizeof(unsigned long))
> > + page->dma_addr[1] = addr >> 16 >> 16;
>
> ... but we do have upper_32_bits() for a 32-bit right shift.

Yep, that's what my current tree looks like.