Re: [Linux-gpib-general] New GPIO adapter / Patch submission guide?

[dave p. <dpe...@gm...>]

Hi Michael,
We have no guidelines for submitting patches, we are very flexible. Smaller
patches are of course preferred.

A lot of work has been done to avoid missing or double interrupts, although
we had not seen buffer overrun.
See also workaround <https://forums.raspberrypi.com/viewtopic.php?t=350295>
for a hardware issue.
What kernel version are you using?

What was your test setup when you encountered the problems you mentioned?

cheers,
-Dave

On Sun, 14 Jul 2024 at 19:07, Michael Schwingen <mi...@sc...>
wrote:

> Hi,
>
> are there any guidelines on how to submit patches, and the formatting style
> preferred?
>
> I have built another GPIO-based GPIB adapter:
>
> https://www.schwingen.org/yoga-gpib/
>
> I did the first development/testing on a simple adapter using only series
> resistors and noticed several problems - mostly hangs due to missing
> interrupts, but also kernel crashes due to the interrupt handlers sometimes
> transferring more bytes than the buffer size. While this mostly occurred on
> a non-working bus (device powered down), I got hangs and lost interrupts
> even on a simple setup with only 1 working device attached.
>
> After some search with a logic analyzer, I came to the conclusion that the
> use of edge-triggered interrupts on an asynchronous bus with slow edges
> does
> not really work reliably. I re-wrote the interrupt handlers in
> gpib_bitbang.c to be level-sensitive, enabling/disabling them as required
> on
> the fly.
>
> The resulting code works reliably in my tests both with the simple
> series-resistor-based adapter as well as my YOGA adapter with level
> shifters, and has been in use for some weeks now in our EMC lab at work.
>
> As a RFQ, the current code is at
>
>
> https://github.com/mschwingen/hardware/blob/master/YOGA_GPIB/software/gpib_bitbang.c
>
> I can provide patches as required. Currently, this is one big patch that
> contains
>  - the interrupt rewrite
>  - modified GPIO setup, since my adapter needs a different GPIO pin layout,
>    with gaps in the used GPIOs
>  - LED handling is changed from direct GPIO access to using kernel LED
>    triggers, allowing to use the LED from system bootup and then switching
>    to GPIB function as required.
>
> I am unsure about SRQ interrupt handling, as I don't understand how this is
> supposed to work and I can't test it, so I left that as-is.
>
> I am looking for guidance on how to go forward with this.
>
> thanks,
> Michael Schwingen
>
>
> _______________________________________________
> Linux-gpib-general mailing list
> Lin...@li...
> https://lists.sourceforge.net/lists/listinfo/linux-gpib-general
>