Using information from pages 26-14 and 26-15 of Freescale's MCF547x Reference Manual (Rev 2.1) I issue the following commands to losh:

> losh> w /b 0x10008f20 0
> losh> w /b 0x10008f08 0
> losh> w /b 0x10008f20 0x0a
> losh> w /b 0x10008f0c 0
> losh> w /b 0x10008f08 0
> losh> w /b 0x10008f20 1

This should theoretically issue a STOP to all slaves and make sure the bus is idle.

At this point, I dump the registers to see what state everything's in:
> losh> x 0x10008f00 4
> 0x10008f00 00000000 00000000 00000000 81000000

This looks right; Transfer Complete (i.e. not in progress) and no ACK received. This is the default state of I2SR. Now I want to set up to write something to a slave:
> losh> w /b 0x10008f04 0x10
> losh> w /b 0x10008f00 0
> losh> w /b 0x10008f08 0x80
> losh> w /b 0x10008f08 0xb0

Again I check the registers:
> losh> x 0x10008f00 4
> 0x10008f00 00000000 10000000 90000000 93000000

Wha...?

I wrote a b0 to I2ICR, but what stuck was a 90. Over in I2SR I find the reason: Arbitration lost.

Who the double-hockey-sticks did I lose arbitration to? There's nobody on the bus except the CF and the RTC! Plus, I have a scope on SCL and SDA and they're both still high so nobody's talking anyway. (I see a burst of I2C data when losh starts up so I know the scope is on the right pins.) (losh has set the two bits in the I2C PAR for me so we're OK there too.)

Is this normal behavior?

Anyway, I bull on ahead and run through paragraph 26.5.1, thus:

If IBB is set, 0x1000f80c will be =>0xa0 and it's not (it's 93, see above), so the bus is not busy.
> losh> w /b 0x10008f08 0x10
> losh> w /b 0x10008f08 0x20
> losh> w /b 0x10008f10 0xa0

0xa0 is the address of the EEPROM I will eventually put out there; even though it's not connected at the moment I should still see its address go out over the bus, but... no activity!

So I'm missing something. What?