On Tue, Sep 09, 2025 at 11:02:36AM -0400, Jon Maloy wrote:
> 
> 
> On 2025-09-07 22:42, David Gibson wrote:
> > On Fri, Sep 05, 2025 at 10:11:46PM -0400, Jon Maloy wrote:
> > > We add a cache table to keep partial contents of the kernel ARP/NDP
> 
> [...]
> 
> > > +	return mac_undefined(e->mac);
> > > +}
> > > +
> > > +/**
> > > + * mac_entry_expired() - Check if a cache entry has expired
> > > + * @e:		Cache entry
> > > + *
> > > + * Return: True if the entry has expired, false otherwise
> > > + */
> > > +static bool mac_entry_expired(const struct mac_cache_entry *e)
> > > +{
> > > +	struct timespec now;
> > > +
> > > +	clock_gettime(CLOCK_MONOTONIC, &now);
> > 
> > Mostly we try to keep to a single clock_gettime() call per epoll
> > cycle, passing 'now' down to the things we call there.
> > 
> > > +	return timespec_before(&e->expiry, &now);
> > > +}
> 
> Passing this value along the whole call chain just for the unlikely case
> that we may need it here seems like a bad idea, especially since the
> performance gain is minimal.
> 
> It would make some sense if we add it to the context struct or even as a
> global variable the set it in the main loop, and pass it along that way.
> What do you think?

Honestly, I prefer calling clock_gettime() again to adding another
global or pseudo-global.

-- 
David Gibson (he or they)	| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you, not the other way
				| around.
http://www.ozlabs.org/~dgibson