On Tue, 18 Feb 2020 15:47:13 -0800 (PST) David Miller wrote:

Really, mistakes happen and a poorly implemented or inserted fexit
module should not be a reason to not have access to accurate and
working statistics for fundamental events.

I am therefore totally against requiring fexit for this functionality.
If you want more sophisticated events or custome ones, sure, but not
for this baseline stuff.

I do, however, think we need a way to turn off these counter bumps if
the user wishes to do so for maximum performance.

Yes, this point plus the precedence you mentioned elsewhere are quite
hard to contend with.

In an ideal world I was wondering if we could have the kernel install
the fexit hook, a'la what we do with drop monitor using tracepoints
from within the kernel.

Then have a proper netlink stats group for them, instead of the ethtool
free-form endlessly bikesheddable strings.

But I guess it could be hard to easily recover the source interface
pointer without digging through NAPI instances or such :S