Not needed anymore.

Could move this down in this case given it's not committed yet.

use SPDX?

While this is probably sufficient, does that mean we can leave actual data behind in the buffer? Should we zero the entire buffer? I do not know how this is used, hence asking as there is a cost to zeroing...

I would prefer if the public KPI functions would be named linuxkpi_ and "internal" functions get the "lkpi_" prefix.
I think during the last year we slowly got to that and it would help if we try to converge to that for everything new.

You probably want to print to buffer + len and only do that if len < size (not overflowed yet) and then the s->size needs to be reduced by len as well?

An overflow of the buffer is also considered an error so we'll need to check if len + ret < size and only if so update len and return success.

This likely needs to be += then.

According to documentation this returns 0 on success and -1 on error, not the length it would have printed.