Hello Christopher,

Varlese, Christopher wrote:

(FYI I working on the kmeter1)

kmeter1.c reuses the same QE_ENET10 RGMII errata workaround code from mpc836x_mds.c (MPC8360EMDS eval board).

In my view errata nodes in the dts is overkill.   Maybe the errata code can go into a reusable function somewhere in 83xx/ or in ucc_geth.c?

To put an errata node in the dts was just an idea ;-)
I also mentioned putting this code in the ucc_geth.c driver ...

I also think the original errata code needs improving:
	- mask some SVR bits so activated for all matching CPU models, e.g. MPC8360 & MPC8360E.

Did a first try in my v2 patch, see:

http://ozlabs.org/pipermail/linuxppc-dev/2009-April/071384.html

but I got no response yet.

	- The code in mpc836x_mds.c and kmeter1.c does not do exactly what Freescale errata says!

:-(

Here the errata document:
	http://www.freescale.com/files/32bit/doc/errata/MPC8360ECE.pdf

Because kmeter1 is using an MPC8360 CPU model the workaround doesn't actually trigger.  So to kill 2 birds with 1 stone we tested a Uboot patch (below) doing what QE_ENET10 says.   It seemed to work fine for us.
        /* RGMII timing Errata workaround for rev 2.1 silicon
         * (ref: MPC8360ECE rev.1 12/2007 QE_ENET10 UCC2 option 1)
         */
        void *reg = (void *)(CONFIG_SYS_IMMR + 0x14ac);
        clrsetbits_be32 (reg, 0x000000F0, 0x000000A0);

quoted

From my point of view:

	- The workaround code in kmeter1.c could go for now.
	- An improved errata workaround for 836x boards would be nice (..who is motivated? :-))

I can make this errata, if someone gives advice, where to put ...
I vote for putting it into the ucc_geth.c driver, and activating it
maybe through the "phy-connection-type" if it activates the rgmii
mode ... ?

bye
Heiko
-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany