On Wednesday 06 June 2007, Segher Boessenkool wrote:

quoted

Is this a bugfix that needs to go in 2.6.22?

This potentially breaks lots of other platforms, so no
please, put it in early in the next cycle and give those
others time to recover from this change.

I would expect that the bugfix is relatively harmless, as the
driver is rather new and there can only be a very small set
of users. All the OF based systems should be working fine
either way, because they already have legacy_serial, and
the dtc sources that come with the kernel all list the 
serial ports as compatible to either "ns16550", "cpm_uart",
"marvell,mpsc", or "mpc5200-psc-uart". In the first case
of those four, the change is harmless, for the other three
it is a valid bug fix.

It probably only makes a difference for 52xx in the case that
you have CONFIG_SERIAL_OF_PLATFORM=y and CONFIG_SERIAL_MPC52xx=m.
In that configuration, you get the same bug that Kou Ishizaki
observed on celleb without this fix.

	Arnd <><