Slides: https://github.com/pdp7/talks/blob/master/lpc-bof-fpga.pdf

Links also in:
https://linuxplumbersconf.org/event/7/contributions/705/

Here are the links:
[1] https://github.com/enjoy-digital/litex
[2] https://github.com/litex-hub/linux/commits/litex-vexriscv-rebase/drivers
[3] https://github.com/enjoy-digital/litex
[4] https://lkml.org/lkml/2020/6/4/303
[5] https://groups.google.com/d/msg/linux-litex/fJLlcsuBibY/3vP8_7nGAwAJ
[6] https://gitlab.freedesktop.org/mupuf/litedip/
[7] https://mupuf.org/blog/2020/06/09/FPGA-why-so-few-drivers/

Martin:

• we need to find a way to describe the hardware
• device tree works for something that is not discoverable
• device tree overlay could handle plugable modules
• alternative to DT: LiteDIP describes the hardware in a different way
• when we design SoC, Linux will be able to discover what IP is present and load the correct driver for that IP

jic (Jonathan Cameron):

• some devices like a power control chip will have various peripherals instantiate after it comes up
• this how MFD (multi-function device) works
• regmap should handle this use case
• the linux-litex thread said the overhead for regmap was 2 indirect calls
• jic and martin believe this is not a problem - there should be real data that shows there is a performance problem
• regmap can indicate le or be

jamie lokier:

• if regmap performance is bad, run time patching could be used 

Martin Peres: 

• I am interested in hotplugging for LiteX perihperals that is available of USB or PCIe
• issue with USB and PCIe, vendor and product id... would need new ID for each verison of LiteX device?
• use the buses 

Jason Sidebottom:

• there is a project that provides VID/PID for open source projects

jic:

• the non-discoverable children could be under a MFD, depends on your discoveribliity 

Jason Sidebottom:

•  we need to avoid having many different driver versions for similar IP, like a UART

Martin:

• every hardware IP block would have a version number
• but CSR are not "free" in terms of hw resources
• 32-bit should be enough

jic:

• create git repo to track LiteX module version numbers
• TODO: Drew should bring this idea to Flourent who created LiteX

Rich Persaud

• Is there a requirement for such devices to be accessible to virtual machines, e.g. via virtio?
• (Drew: here was discussion of this on litex github issues and linux-litex mailing list)