Re: PCI-X support

[Richard B. Johnson]

On Fri, 1 Jul 2005, Vojtech Pavlik wrote:

> On Fri, Jul 01, 2005 at 03:02:11PM -0400, Richard B. Johnson wrote:
>
> > The driver (whatever that is), if it was written for a 64-bit
> > platform, can write a 64-bit word in one operation and it's
> > transparent. If the driver was written for a 32-bit environment,
> > it will still work because there is compatibility with PCI 2.x
> >
> > FYI, this machine has a PCI-X bus. I have some 32-bit cards
> > plugged into it (SCSI controller, etc.). They work. I also
> > have a 64-bit card plugged into it (fiber-optic data link).
> > It also works, but at 133 MHz. Software never talks to it
> > in 'long longs' so the increased data-width isn't being used.
>
> Are you sure about that? I'd assume when doing busmastering, it'll use
> 64-bit transfers, if the driver sets the correct DMA mask.

Yep. The board uses a PLX PCI 9656BA chip configured for local-bus
transfers during DMA The local-bus is 32-bits from a FIFO. There
is no way to collapse these writes from the 32-bit source/dest.
The transfers are 32 quadwords for a write and 16 quadwords for
a read, but that represents a burst, not the data-width. Without
a 64-bit data-path to memory on one side, and a 64-bit data-path
to the device (a FIFO) on the other side, you will not get 64-bit
transfers.


> --
> Vojtech Pavlik
> SuSE Labs, SuSE CR
> -

Cheers,
Dick Johnson
Penguin : Linux version 2.6.12 on an i686 machine (5537.79 BogoMips).
 Notice : All mail here is now cached for review by Dictator Bush.
                 98.36% of all statistics are fiction.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/