Re: [PATCH net-next v3] net: af_packet: Use hrtimer to do the retire operation

From: Willem de Bruijn
Date: Sun Aug 17 2025 - 11:55:50 EST


Xin Zhao wrote:
> On Sun, 2025-08-17 at 21:28 +0800, Willem wrote:
>
> > Here we cannot use hrtimer_add_expires for the same reason you gave in
> > the second version of the patch:
> >
> > > Additionally, I think we cannot avoid using ktime_get, as the retire
> > > timeout for each block is not fixed. When there are a lot of network packets,
> > > a block can retire quickly, and if we do not re-fetch the time, the timeout
> > > duration may be set incorrectly.
> >
> > Is that right?
> >
> > Otherwise patch LGTM.
>
>
> I'll think about whether there's a better way to implement the logic.
>
> Additionally, regarding the previous email where you mentioned replacing retire_blk_tov
> with the interval_ktime field, do we still need to make that change?
> I noticed you didn't respond to my latest patch that replaces retire_blk_tov with
> interval_ktime, and I'm wondering if we should make that change.
> So we remain the retire_blk_tov field?

Sorry, this response was intended to v4. Yes, let's keep that change.
If hrtimer_add_expires cannot be used, then that patch is good as is.