I (think) the AR8327 switch support now works

[Milan Obuch]

On Sat, 8 Mar 2014 14:50:14 -0800
Adrian Chadd <adrian at freebsd.org> wrote:

> Ok, well now let's figure out if the reset is working or not.
> 
> look at arswitch_attach():
> 
> * It sets up the default methods;
> * it calls ar8327_attach();
> * it resets the switch via arswitch_reset();
> * it calls the HAL methods to set things up;
> * it attaches the PHYs;
> * it attaches said PHYs;
> * it starts the periodic checks via arswitch_tick().
> 
> So, I'd add some printf()s in arswitch_attach() to see where it's
> getting to before it calls return (ENXIO) or return (err).
> 
> 
> -a
>

OK, so I patched said function with some printf's (in attachment),
built new kernel and got following result:

mdio0: <MDIO> on argemdio0
mdioproxy0: <MII/MDIO proxy, MDIO side> on mdio0
arswitch0: readreg 0x00000000: 0x12041204
arswitch0: readreg 0x00000000: 0x00000000
arswitch0: <Atheros AR8327 Ethernet Switch> on mdio0
arswitch0: arswitch_attach 1
device_attach: arswitch0 attach returned 6

Now I put there some more printf's (in arswitch_probe) and things go a
bit stranger. I've got:

mdio0: <MDIO> on argemdio0
mdioproxy0: <MII/MDIO proxy, MDIO side> on mdio0
unknown: arswitch_probe 1
arswitch0: readreg 0x00000000: 0x00000000
unknown: arswitch_probe 2 0

and for the second time, just doing one more reboot:

mdio0: <MDIO> on argemdio0
mdioproxy0: <MII/MDIO proxy, MDIO side> on mdio0
unknown: arswitch_probe 1
arswitch0: readreg 0x00000000: 0x12041204
unknown: arswitch_probe 2 6
unknown: arswitch_probe 1
arswitch0: readreg 0x00000000: 0x00000000
unknown: arswitch_probe 2 0
arswitch0: <Atheros AR8327 Ethernet Switch> on mdio0
arswitch0: arswitch_attach 1 0
device_attach: arswitch0 attach returned 6

Next reboot brings again the first result, next one the second, again
the second, again the first one etc. It looks like two two
possibilities are alternating with some level of randomness. To me this
tells that mdio does not work reliably in this case - am I right
thinking that reading register 0 should always return the same value,
0x12041204, at least when no other register is touched in between?

Well, this was with soft reset - I got every time mountroot promp, so
just pressing enter gives me panic and then I typed reboot. when doing
power-cycle, things are a bit deterministic - three time in a row I got
the first result.

Next question - why is arswitch_probe function called second time? It
is called twice for both switches, so it is probably by design, and as
the second switch is only 'hint probed', it does not matter. But for
8327, register is read, unfortunatelly not the same result is given
both times.

Also, I played with if_arge.c - arge_fetch_mdiobus_clock_rate()
function a bit, but no matter the clock divider I used, the results are
still the same. I see in next function, arge_reset_miibus a comment
about resetting the mdio block(s), so I suspect it is something to be
done, maybe, but is not yet. I have no idea how to try it, however. I
can just wait for next hint, now.

Regards,
Milan
-------------- next part --------------
A non-text attachment was scrubbed...
Name: arswitch.c-diff
Type: application/octet-stream
Size: 2206 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/freebsd-mips/attachments/20140309/16dd73a2/attachment.obj>