old tulip patch

old tulip patch

[David Mosberger]

I'd like to remove the attached old patch from my linux-ia64-2.5 tree.
If anybody thinks it's still needed, please speak up now.

Thanks,

	--david

diff -Nru a/drivers/net/tulip/media.c b/drivers/net/tulip/media.c
--- a/drivers/net/tulip/media.c	Mon Jan 12 15:37:12 2004
+++ b/drivers/net/tulip/media.c	Mon Jan 12 15:37:12 2004
@@ -278,6 +278,10 @@
 				for (i = 0; i < init_length; i++)
 					outl(init_sequence[i], ioaddr + CSR12);
 			}
+
+			(void) inl(ioaddr + CSR6); /* flush CSR12 writes */
+			udelay(500);		/* Give MII time to recover */
+
 			tmp_info = get_u16(&misc_info[1]);
 			if (tmp_info)
 				tp->advertising[phy_num] = tmp_info | 1;

Re: old tulip patch

[Grant Grundler]

On Mon, Jan 12, 2004 at 04:13:02PM -0800, David Mosberger wrote:
> I'd like to remove the attached old patch from my linux-ia64-2.5 tree.
> If anybody thinks it's still needed, please speak up now.

Sorry - that's my mess. Feel free to remove it.
I expect tulip will stop working for some folks.

Turns out that while the simple udelay() usually works, the driver
ignores the 802.3 spec.  Newer patch is here:
	ftp://ftp.parisc-linux.org/patches/diff-tulip_phy_reset-03

Jeff Garzik rejected this patch because it introduces a potential
1ms delay while holding off interrupts. I agree this is not a good
thing. But it's better than not working.

Jeff wanted me to avoid this by adding schedule_timeout() to the code
path someplace. I refused since the above Works For Me and the init
sequence in tulip is, uhm, well, "difficult". It covers too many
different NICs for me to even begin to understand all the variations.
I felt my patch was fairly low risk change to a "mature" product.

I'll maintain the patch for SuSE/RHAS/Debian on ia64 and for parisc-linux.

thanks,
grant

Re: old tulip patch

[David Mosberger]

I'll drop it from my tree then and let you and Jeff figure out what to
do about it.

Thanks,

	--david

>>>>> On Mon, 12 Jan 2004 17:00:59 -0800, Grant Grundler <iod00d@hp.com> said:

  Grant> On Mon, Jan 12, 2004 at 04:13:02PM -0800, David Mosberger wrote:
  >> I'd like to remove the attached old patch from my linux-ia64-2.5 tree.
  >> If anybody thinks it's still needed, please speak up now.

  Grant> Sorry - that's my mess. Feel free to remove it.
  Grant> I expect tulip will stop working for some folks.

  Grant> Turns out that while the simple udelay() usually works, the driver
  Grant> ignores the 802.3 spec.  Newer patch is here:
  Grant> ftp://ftp.parisc-linux.org/patches/diff-tulip_phy_reset-03

  Grant> Jeff Garzik rejected this patch because it introduces a potential
  Grant> 1ms delay while holding off interrupts. I agree this is not a good
  Grant> thing. But it's better than not working.

  Grant> Jeff wanted me to avoid this by adding schedule_timeout() to the code
  Grant> path someplace. I refused since the above Works For Me and the init
  Grant> sequence in tulip is, uhm, well, "difficult". It covers too many
  Grant> different NICs for me to even begin to understand all the variations.
  Grant> I felt my patch was fairly low risk change to a "mature" product.

  Grant> I'll maintain the patch for SuSE/RHAS/Debian on ia64 and for parisc-linux.

  Grant> thanks,
  Grant> grant

Re: old tulip patch

[Grant Grundler]

On Mon, Jan 12, 2004 at 05:09:03PM -0800, David Mosberger wrote:
> I'll drop it from my tree then

ok

> and let you and Jeff figure out what to do about it.

We've agreed to disagree. And since Jeff is the maintainer, my patch
is not going in and I'm not rewriting the tulip init sequence.
If anyone is looking for a challenging, fairly well defined kernel driver
project, this would be a good one.  Jeff told me several times how
simple he thinks it is...it just never looked that simple to me.

grant