Den 09.09.2015 03:48, skrev Eric Anholt:
kernel@xxxxxxxxxxxxxxxx writes:
From: Martin Sperl <kernel@xxxxxxxxxxxxxxxx>
The BCM2835 contains 3 auxiliar devices:
* spi1
* spi2
* uart1
All of those 3 devices are enabled/disabled via a shared register,
which is set by default to be disabled.
Access to this register needs to get serialized.
So after several iterations of discussions with the following ideas:
* syscon - device tree should describe HW not drivers to use -
'compatiblity = "brcm,bcm2835-aux-enable", "syscon";'
is not acceptable
* regulator - it is not necessarily a regulator or a power gate
that is implemented in HW, so it is not valid to use
this framework
The recommendation was made to create a new minimal API in soc
just for access to this shared enable/disable register.
This patch-series implements:
* the bcm2835-auxiliar device enable/disable api in soc.
* the bcm2835-auxiliar spi device driver
The uart1 device driver (ns16550 based) is not implemented so far
but would be using the same API.
Both spi and uart drivers can run with shared interrupts,
so there is no need for an interrupt-controller to get implemented.
I finally had a chance to sit down and look at what the hardware's doing
with the enable bit (also, I've read a whole lot more of the hardware
now, so I'm a lot faster at answering questions like this). The enable
bits are a clock gate off of the VPU clock.
I knocked together the enable bits as a clock gate driver, since I'd
just written very similar code for the audio domain clock driver (and I
assume you are grumpy about how much time you've spent on this one
stupid register). It's up at
https://github.com/anholt/linux/tree/bcm2835-clock-aux and I can submit
it if you like the result. I've compile tested it only, but I'm hoping
you could just drop your aux SPI driver on top of it and have things
work.
This looks interesting.
But there's a challenge with the uart1 and the 8250 driver.
Phil Elwell has this to say:
This means that that UART1 isn't an exact clone of a 8250 UART.
In a particular, the clock divisor is calculated differently.
A standard 8250 derives the baud rate as clock/(divisor16),
whereas the BCM2835 mini UART uses clock/(divisor8). This means
that if you want to use the standard driver then you need to lie
about the clock frequency, providing a value is twice the real
value, in order for a suitable divisor to be calculated.
Ref: https://github.com/raspberrypi/linux/pull/1008#issuecomment-139234607
So either we need a new uart1 driver or a doubled clock freq. somehow.
Noralf.
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html