Ubiquiti EdgeRouter Lite works multi-user with -CURRENT.

[Oleksandr Tymoshenko]

On 2013-05-17, at 10:52 AM, Milan Obuch <freebsd-mips at dino.sk> wrote:

> On Fri, 17 May 2013 13:28:06 -0400, Warner Losh <imp at bsdimp.com> wrote:
> 
>> 
>> On May 17, 2013, at 1:22 PM, Milan Obuch wrote:
>> 
>>> On Thu, 16 May 2013 19:10:26 +0100, Joe Holden <lists at rewt.org.uk>
>>> wrote:
>>> 
>>> [ snip ]
>>> 
>>>>> for this quick test I used OCTEON1 kernel config with some
>>>>> necessary tweaks, device gpio is commented out, maybe this will
>>>>> be enough for enable octeon_gpio.c (looking at files.octeon1 it
>>>>> seems to be possible). I will test it, definitely.
>>>> 
>>>> There is only one pin exposed via gpio:
>>>> 
>>>> root at erl1:~ # gpioctl -f /dev/gpioc0 -l -v
>>>> pin 07: 0       F/D<IN>, caps:<IN,OUT>
>>>> 
>>>> Not sure if it maps to anything usable, haven't played with it yet
>>>> - the other header is EJTAG according to Cavium docs.
>>> 
>>> I built new kernel with device gpio, something is not working - F/D
>>> should mean 'factory default' aka 'reset switch', but it does not
>>> sense status ot that switch. No matter whether pressed or not,
>>> gpioctl still reports 0.
>>> 
>>> Also, somehow strangely, in dmesg:
>>> 
>>> gpio0: <Cavium Octeon GPIO driver> on ciu0
>>> gpio0: [GIANT-LOCKED]
>>> gpio0: [GIANT-LOCKED]
>>> gpio0: [GIANT-LOCKED]
>>> gpio0: [GIANT-LOCKED]
>>> gpio0: [GIANT-LOCKED]
>>> gpio0: [GIANT-LOCKED]
>>> gpio0: [GIANT-LOCKED]
>>> gpio0: [GIANT-LOCKED]
>>> gpio0: [GIANT-LOCKED]
>>> gpio0: [GIANT-LOCKED]
>>> gpio0: [GIANT-LOCKED]
>>> gpio0: [GIANT-LOCKED]
>>> gpio0: [GIANT-LOCKED]
>>> gpio0: [GIANT-LOCKED]
>>> gpio0: [GIANT-LOCKED]
>>> gpio0: [GIANT-LOCKED]
>>> gpio0: [pin0] output=0, invinput=0, intr=0, intr_type=level
>>> gpio0: [pin1] output=0, invinput=0, intr=0, intr_type=level
>>> gpio0: [pin2] output=0, invinput=0, intr=0, intr_type=level
>>> gpio0: [pin3] output=0, invinput=0, intr=0, intr_type=level
>>> gpio0: [pin4] output=0, invinput=0, intr=0, intr_type=level
>>> gpio0: [pin5] output=0, invinput=0, intr=0, intr_type=level
>>> gpio0: [pin6] output=0, invinput=0, intr=0, intr_type=level
>>> gpio0: [pin7] output=0, invinput=0, intr=0, intr_type=level
>>> gpio0: [pin8] output=0, invinput=0, intr=0, intr_type=level
>>> gpio0: [pin9] output=0, invinput=0, intr=0, intr_type=level
>>> gpio0: [pin10] output=0, invinput=0, intr=0, intr_type=level
>>> gpio0: [pin11] output=0, invinput=0, intr=0, intr_type=level
>>> gpio0: [pin12] output=0, invinput=0, intr=0, intr_type=level
>>> gpio0: [pin13] output=0, invinput=0, intr=0, intr_type=level
>>> gpio0: [pin14] output=0, invinput=0, intr=0, intr_type=level
>>> gpio0: [pin15] output=0, invinput=0, intr=0, intr_type=level
>>> gpioc0: <GPIO controller> on gpio0
>>> gpiobus0: <GPIO bus> on gpio0
>>> 
>>> GIANT-LOCKED line is there 16 times, once for every one from 16
>>> pins - why?
>> 
>> Because it isn't marked as MPSAFE? Chances are something will need to
>> be done to make it mp safe though.
>> 
>> Warner
> 
> To be precise - why 16 times? And why 16 pins, when there is only one
> switch, and only one pin is reported in gpioctl output.

Each pin got its  own interrupt. I believe this message is printed when
bus_setup_intr is called. 

The discrepancy  between gpioctl and kernel is explained that only
some of actual pins are accessible from userland. Because some pins
might be reserved as signal lines by other drivers (e.g. card detect for SD 
interface).