On 23.8.2012 15:07, Jakub Jermar wrote:
>> - consider being port IO / memory mapped IO neutral
> 
> mmmm, I don't remember what the problem was with the current
> implementation...
> Can you refresh my mind please?

The ns8250 driver (and also your rtc-clock dirver) refuses its
hw_resources if it turns out the I/O registers are memory mapped (and
hence not from the I/O port space). The driver should not make
differences between an I/O port and a memory mapped registers in this
case because the pio_enable(), pio_read*/write*() functions can deal
with both in a uniform manner.

We agreed with Jiri that there may be some situations when it may be
useful to know what kind of a hw resource the driver is passed (for
example to be able to tell the resources apart), but this is not the
case. The pio_* interfaces are there to unify the distinct notions of
I/O ports and memory mapped registers and especially the different
access mechanisms so that the drivers can be platform neutral and
portable. Reinforcing the differences in the drivers would be against
this principle and will also result in non-portable drivers.

Jakub

_______________________________________________
HelenOS-devel mailing list
[email protected]
http://lists.modry.cz/cgi-bin/listinfo/helenos-devel

Reply via email to