Re: [PATCH] Lower e100 latency

[Jesse Brandeburg]

On 3/27/06, J.A. Magallon <jamagallon@xxxxxxx> wrote:
> Corrected:
>
> --- linux/drivers/net/e100.c.orig       2006-01-24 09:20:44.000000000 +0100
> +++ linux/drivers/net/e100.c    2006-01-24 09:21:55.000000000 +0100
> @@ -884,10 +884,10 @@
>          * procedure it should be done under lock.
>          */
>         spin_lock_irqsave(&nic->mdio_lock, flags);
> -       for (i = 100; i; --i) {
> +       for (i = 1000; i; --i) {
>                 if (readl(&nic->csr->mdi_ctrl) & mdi_ready)
>                         break;
> -               udelay(20);
> +               udelay(2);

what is the purpose of this patch?  what bug is it solving?  Are we
trying to achieve some goal?  A comment at the very least is
necessary.  I don't like changing timing stuff unless we have some
clear reason.  In fact I sent a patch a while back to a guy who was
complaining about latency in a -RT kernel with e100 and he said this
kind of change made things worse: see the end of:

http://marc.theaimsgroup.com/?l=linux-kernel&m=113808831932769&w=2

The problem in this case is the mii library calling back into our
mdio_read.  eepro100's mdio read hard spins with no delay besides the
ioread32 delay created by reading from an i/o port.  This could
explain the glitching in an RT kernel.

Is this the kind of problem this patch tries to solve?

Thanks,
  Jesse
-
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/