Re: [PATCH net-next v3 1/2] drivers: net: stmmac: handle start time set in the past for flexible PPS

[Jakub Kicinski <kuba@kernel.org>]

On Fri, 29 Aug 2025 12:51:40 +0200 Gatien CHEVALLIER wrote:
> On 8/28/25 04:31, Jakub Kicinski wrote:
> > On Wed, 27 Aug 2025 13:04:58 +0200 Gatien Chevallier wrote:  
> >> +		curr_time = ns_to_timespec64(ns);
> >> +		if (target_ns < ns + PTP_SAFE_TIME_OFFSET_NS) {
> >> +			cfg->start = timespec64_add_safe(cfg->start, curr_time);  
> > 
> > Is there a strong reason to use timespec64_add_safe()?
> > It's not exported to modules:
> > ERROR: modpost: "timespec64_add_safe" [drivers/net/ethernet/stmicro/stmmac/stmmac.ko] undefined!  
> 
> Hello Jakub,
> 
> you're absolutely right. I don't know how I did not encounter the build
> error while performing some tests, that I'm getting now as well.
> 
> The handling of overflows is already done in that function. Either
> I can make a patch to export the symbol or handle the computation in the
> driver. What do you think is best?

The odds of me being right about time related code are only slightly
better than 50/50, and I don't know what "flexible PPS" is :)
But in principle, if the reason you need to check for overflow is valid
-- add the export. The time maintainers will tell us if they don't
want it.