Re: question regarding cacheline size

[Russell King <rmk+lkml@arm.linux.org.uk>]

On Thu, Sep 07, 2006 at 06:23:11AM -0600, Matthew Wilcox wrote:
> On Thu, Sep 07, 2006 at 01:07:56PM +0100, Russell King wrote:
> > I've often wondered why we don't set the cache line size when we set the
> > bus master bit - ISTR when I read the PCI spec (2.1 or 2.2) it implied
> > that this should be set for bus master operations.
> 
> It's not required ... 3.2.1 of pci 2.3 says:
> 
> While Memory Write and Invalidate is the only command that requires
> implementation of the Cacheline Size register, it is strongly suggested
> the memory read commands use it as well. A bridge that prefetches is
> responsible for any latent data not consumed by the master.
> 
> (obviously this is talking about requirements placed on the device, not
> on the OS, but it'd behoove us to help the device out here).
> 
> It's also useful to implement it for slave devices.  PCI 2.3 has the
> concept of cacheline wrap transactions -- eg with a cacheline size of
> 0x10, it can transfer data to 0x108, then 0x10C, 0x100, 0x104, then
> 0x118, etc

As does v2.2 and v2.1.

> So I think we should redo the PCI subsystem to set cacheline size during
> the buswalk rather than waiting for drivers to ask for it to be set.

Agreed, and this is something I'm already doing on ARM in my
pcibios_fixup_bus().

-- 
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 Serial core