On Tue, May 26, 2009 at 00:47, gyang wrote:
> On Wed, 2009-05-20 at 06:08 -0400, Mike Frysinger wrote:
>> On Wed, May 20, 2009 at 05:50, gyang wrote:
>> > On Wed, 2009-05-20 at 05:36 -0400, Mike Frysinger wrote:
>> >> On Thu, Jan 22, 2009 at 04:04,  <[email protected]> wrote:
>> >>> Modified: trunk/drivers/net/bfin_mac.c (6031 => 6032)
>> >>>
>> >>>     /* search for connect PHY device */
>> >>> -   for (i = 0; i < PHY_MAX_ADDR; i++) {
>> >>> +   for (i = PHY_MAX_ADDR - 1; i >= 0; i--) {
>> >>
>> >> why did you change it from searching 0...31 to 31...0 ?  the vast
>> >> majority of boards out there will be sitting at address 0, not 31.
>> >> this incurs unnecessary overhead and looks like a hack rather than a
>> >> real fix.
>> >
>> > I want to use the PHY at address 2, rather than 0 for KSZ8893M.
>> >
>> > I may need to add a Kconfig options to assign the PHY port when enable
>> > DSA, and then revert this change.
>>
>> board resource you mean ...
>>
>> maybe we're looking at this the wrong way.  there are three devices
>> listening on MII bus with KSZ8893M.  shouldnt they all show up ?  if i
>> look at how wireless is architected, there is a "wmaster#" and a
>> "wlan#" -- the wmaster# isnt accessed directly, but the wlan# is.  or
>> maybe this isnt relevant at all.
>>
>> do we have any examples of how other people are using DSAs ?
>
> I'm looking at Marvell's kirkwood rd88f6281 board resource...
> When DSA is enabled, mv643xx_eth driver would work at phy-less mode --
> network link is always on. An valid/invalid phy address in board
> resource told the mac driver whether to work at phy or phy-less mode.
>
> For blackfin, a Kconfig option is better, and can avoid moving
> bfin_mac.h or import another head file to include/linux/.

until mainline settles on DSA support, that should be fine
-mike

_______________________________________________
Linux-kernel-commits mailing list
[email protected]
https://blackfin.uclinux.org/mailman/listinfo/linux-kernel-commits

Reply via email to