Issues with usb gpio support

[vespaman]

Hi all,

I have this USB/serial devkit from exar/maxlinear with XR21B1424, that I would like to use the gpio pins from. As far as I can understand the kernel module (xr_serial.c) should support gpio for this and many other chips in the same series.
But:- it does not show up in /dev/

Only an ftdi device that I also have connected shows up as gpiochip0, but not XR21B1424. There's also no sign of it in /sys/.

The kernel modules (xr_serial and cdc_acm) are loaded, there's no error in kernel log. The ttyUSBx shows up as they should.
Just no /dev/gpiochipX, and (of course) gpiodetect does not list it.

Does anyone have any idea or suggestion how to proceed, or what to look for? I must miss some dependency or something.

I am so stuck at this point ..

[NeddySeagoon]

vespaman,

Start reading at XR21B1424
In particular the data sheet. That's a PDF.

Page 1 says

[vespaman]

NeddySeagoon,

Thanks for digging, really appreciate!

The way I read the datasheet, is that each pin can be configured with either some control/status signal for UART or a GPIO, apart from RxD and TxD, which are fixed for UART. This is configured in GPIO_MODE register (p. 2.
I think I also read in some patch message in the kernel mailing list (post was about 3 yrs old), that the intention was for the libgpio to work simultaneously while the ttyUSB was open, since the set/get pin state was/is always accessible (but this is not true for the set-up registers, as I understood it).

And:- the grep line you used, actually gives that xr_port_probe() is returning it to the kernel, which kind of makes sense. xr_port_probe is setup as a function pointer in the usb_serial_driver struct. So I guess it must be run when loading the module(?).

[NeddySeagoon]

vespaman,

I'm a retired hardware engineer - software mostly goes over my head.

You can certainly swap between GPIO mode and UART mode for whole UARTs
It's even possible to do that 'on the fly'.

You may be able to mix GPIO and UART in the same UART, say that you used the UART with no handshaking, the handshake pins may be usable as GPIO.
It's not possible to use that same pins for different functions an the same time. That would need some external multiplexing.

I don't see how to tell the kernel to set up either UART mode or GPIO mode. That's software, which is not my thing. :)
I'm not above adding some print statements to see what's happening though.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

[vespaman]

NeddySeagoon,

While I'm not retired yet, I am also a hardware guy (intention was to use this board (or a couple of them) in a production test equipment for a product I am about to start produce). The XR21B1424 has a very low latency for receiving bytes on the serial port, which other don't (e.g. FTDI et al), and all the extra GPIO is also nice to use in the production test for testing/setting inputs and outputs of the boards. But of course I have to find out why it does not show up, otherwise I'll have to use some other solution for the GPIOs.
Print statements might be the way, if nothing else surfaces.

Thanks anyway, I'll follow the trace of the usb_* functions tomorrow, to see if they lead somewhere. Maybe libgpio has to have specific support for this IC, and that is what is missing. Who knows.

- Micael

[NeddySeagoon]

vespaman,

I suspect that the default settings give you four /dev/ttyUSBx ports, with no pins left over for GPIO.
To have some GPIO, you need take some pins away from the UART(s)

That's just the start. Then the kernel GPIO driver needs to know about them.

What does

[vespaman]

So in /dev I have all 4 of the ttyUSB* and boot log shows no error, but also no mention about gpio (which I think might be normal).
Maybe you are right in that there are no default gpio available, and this is why nothing shows up. But then I would have thought it must be configurable somehow, or maybe this is a user space setup-app that needs to be run, before kernel picks up the gpios?

I have a FDTI USB-UART device also on this computer, and it shows its GPIO without any configuration from my side.

The lsusb -vvv gives;

[vespaman]

.. so I added printk's here and there. And it all checks out ok.

xr_gpio_init() sets most of the port functions to generic gpio, only leaving GPIO8 and GPIO8 as other function when plugging in the board (module loaded). Even reading out gpio mode register when opening the ttyUSBx gives that all are set to GPIO except the GPIO8/9, if I am opening the port without HW flow control.

So there's no new clues here.

Not sure what's the link to libgio and/or the gpiochip device in /dev is.

[NeddySeagoon]

vespaman,

In ./drivers/usb/serial/xr_serial.c, the function

[vespaman]

Hi NeddySeagoon,

So what I did this morning, was to add debug outputs (printk) in xr_set_flow_mode() amongst others. Hereby verifying that the mode is always gpio (apart from 8/9), as long as flow is off. If RTS/CTS is activated in e.g. minicom, the driver will take these from the gpio set.But there's still 6 gpio even if 8/9 is for the LED toggle, and RTS/CTS is activated.

I actually print out the register content from the chip in the beginning of xr_set_flow_mode, and what exactly is programmed into it, later on in the same function.

Here's dmesg with my debug printk's;

[NeddySeagoon]

vespaman,

That patch is a new file

[vespaman]

Ah, yes, but I was talking about this patch-set https://lore.kernel.org/all/20210226113709.733f6526@coco.lan/ which I guess was dropped somewhere.

But maybe you are correct in that the xrusb_serial.c is a complete (hopefully still) working file. Another route to try out, thanks! Will see if it compiles, and what happens when loaded.

[vespaman]