On Wed, 2003-07-16 at 05:39, Liam Girdwood wrote:
> I've found some function/type naming errors within the Alchemy serial
> driver. Patch below.
> Index: drivers/serial/au1x00_uart.c
> RCS file: /home/cvs/linux/drivers/serial/au1x00_uart.c,v
> retrieving revision 1.1
> diff -r1.1 au1x00_uart.c
> < up->port.irq = irq_cannonicalize(old_serial_port[i].irq);
> > up->port.irq = irq_canonicalize(old_serial_port[i].irq);
> < int __init early_serial_setup(struct serial_struct *req)
> > int __init early_serial_setup(struct uart_port *req)
> There is also a reference to a non existant member called change_speed
> of struct uart_ops on line 1055 i.e.
> .change_speed = serial_au1x00_change_speed,
> This builds if it is commented out, however serial_au1x00_change_speed()
> is not called anywhere in the driver and the change_speed member is not
> used anywhere else in the kernel!
> It looks like this file is still in transition between 2.4 and 2.6 as
> the 2.4 driver works fine. Does anyone have a working 2.5/2.6 Alchemy
> uart driver ?
The serial driver worked fine for me about a month or more ago. Then
there was a large update with the latest 2.5.x bits and it's very
possible that the driver got broken. There are other Au1x problems in
2.5 as well. The 36 bit address patch is not there yet. The last time I
played with 2.5 the kernel mounted the root fs and then hung trying to
start a userland process ... and I haven't had any time since then. I
think you're in new territory here but, hey, someone needs to clean up