Better than copying code from misoc, IMO.
No harm in also introducing config.py like on the curernt master. Compatibility isn't broken, and that's what's most important for backporting.
Does io_expander even make sense if there's no virtual_leds?
On standalone Kasli-SoCs that means the task isn't actually doing anything at the moment.
Maybe just move it and task::spawn of io_expander_service to the block with io expander settings?
A quick study of the Zynq7000 TRM in the topic of MIO, GPIO, etc. confirms that the reset is indeed not necessary.
Two things to note: An example on programming a MIO pin: https://docs.xilinx.…
Actually, it's even simpler than that. SD code was an inspiration.
Seems like reset_gpio is not necessary. I can boot normally and connect with DRTIO.
sys_clk_freq is the one that should be used, as grabber clock is based off the system clock domain ([deserializer_7series.py#L22](https://github.com/m-labs/artiq/blob/release-7/artiq/gateware/g…