On Mon, Jul 18, 2022 at 6:37 AM Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote:
>
> On Sun, Jul 17, 2022 at 12:34:53PM +0900, Stafford Horne wrote:
> > Two things to note are:
> >
> > - isa_dma_bridge_buggy, traditionally this is defined in asm/dma.h but
> > these architectures avoid creating that file and add the definition
> > to asm/pci.h.
>
> This doesn't have anyting to do with PCI support. I think adding a
> separate header just for this that always stubs it out unless a config
> option is set (which x86 then selects) is the besy idea here. I also
> think the isa_dma_bridge_buggy needs to move out of the PCI code as
> well.
Most architectures have it in asm/dma.h, which is probably the right place
(if we end up keeping it), since this is for the ISA DMA API.
I would copy this declaration from x86
#ifdef CONFIG_PCI
extern int isa_dma_bridge_buggy;
#else
#define isa_dma_bridge_buggy (0)
#endif
to asm-generic/dma.h and remove it from arch/sh to avoid the
one duplicate definition. The architectures that have the declaration
in asm/pci.h (arm64, csky, riscv) already get the asm-generic version
of asm/dma.h.
As mentioned before, it would be even better to just remove it
entirely from everything except x86, and enclose the four
references in an explicit "#ifdef X86_32". The variable declaration
only exists because drivers/pci/quirks.c is compiled on all
architecture, but the individual quirk is only active based on
the PCI device ID of certain early PCI-ISA bridges.
Arnd
