On 01/02/2014 11:03 PM, Cody P Schafer wrote:
> On Thu, Jan 2, 2014 at 1:35 PM, Rafał Miłecki <email@example.com> wrote:
>> 2014/1/2 Hauke Mehrtens <firstname.lastname@example.org>:
>>> From: Cody P Schafer <email@example.com>
>>> Add a few Belkin F7Dxxxx entries, with F7D4401 sourced from online
>>> documentation and the "F7D7302" being observed. F7D3301, F7D3302, and
>>> F7D4302 are reasonable guesses which are unlikely to cause
>>> It also appears that at least the F7D3302, F7D3301, F7D7301, and F7D7302
>>> have a shared boardtype and boardrev, so use that as a fallback to a
>>> "generic" F7Dxxxx board.
>> Cody, Hauke: I'm starring at this patch for 10 minutes now and it's
>> still unclear for me.
>> You say 3301, 3302, 7301 and 7302 have the same board_* entries
>> stating they can be treated with a generic ID entry.
> I included the generic BCM47XX_BOARD_BELKIN_F7DXXXX entry to catch
> those boards that we don't yet have specific entries for. It allows us
> to get the leds and buttons working mostly correctly.
> The specific names are included so that one can determine a more exact
> board. The stock CFE requires different images for different boards
> even though they are very similar. Hardware variations are simply
> gigabit vs 100MB switches, usb port population, led population, and
> 5Ghz radio population (none of which truly require the greater detail
> in board type).
>> At the same time
>> you define BELKIN_F7D3301 and BELKIN_F7D3302... so they are not
>> identical after all?
> [rehash of above] They have the same boardtype & boardrev, but
> (unfortunately) have different image requirements from the stock CFE.
>> Finally what about 4302? I can see it's untested,
>> but for some reason you assign it to the separated enum entry. Is this
>> not going to share config with the generic ones?
> Sorry, I've had this patch go though a couple revisions (adding more
> boards), and not all of them made it into the patch description. (4302
> is just another variation on the generic f7dxxxx board).
>> Sorry, but it looks really messy to me.
> We can thank Belkin for that (see CFE issues mentioned above that
> cause us to want these more specific BCM47XX_BOARD_* macros).
> As an alternate to exposing the specific board names via this
> interface, we (openwrt) could use the nvram userspace tool to look for
> the specific type (the kernel really only needs the generic one,
> unless we want to give a more accurate picture of which LEDs are
> populated). Hauke - thoughts?
If it is possible to detect the specific board I would go with that. At
least when the led configuration is different we have to do different
things for the different boards. In the current way it does not take a
lot of memory to add a new board to the detect code just some bytes <
50 in init ram. I would remove the generic entry now and leave the
others in, if someone has a different board we can add it.