Re: CVS commit: src/sys/net

[Ryota Ozaki <ozaki-r%netbsd.org@localhost>]

On Mon, Sep 23, 2019 at 4:14 PM Kamil Rytarowski <n54%gmx.com@localhost> wrote:
>
> On 23.09.2019 06:53, Rin Okuyama wrote:
> > Hi,
> >
> > On 2019/09/22 18:30, Kamil Rytarowski wrote:
> >> On 12.04.2018 06:38, Ryota Ozaki wrote:
> >>> Module Name:    src
> >>> Committed By:    ozaki-r
> >>> Date:        Thu Apr 12 04:38:13 UTC 2018
> >>>
> >>> Modified Files:
> >>>     src/sys/net: if.h route.c route.h rtsock.c
> >>>
> >>> Log Message:
> >>> Resolve tangled lock dependencies in route.c
> >>>
> >>> This change sweeps remaining lock decisions based on if locked or not
> >>> by moving
> >>> utility functions of rtentry updates from rtsock.c and ensuring
> >>> holding the
> >>> rt_lock.  It also improves the atomicity of a update of a rtentry.
> >>>
> >>
> >>> +static struct ifaddr *
> >>> +rt_update_get_ifa(const struct rt_addrinfo info, const struct
> >>> rtentry *rt,
> >>> +    struct ifnet **ifp, struct psref *psref_ifp, struct psref *psref)
> >>> +{
> >>
> >>
> >> Do we need to pass info as a value? It is pretty large here (1024 bytes).
> >
> > Yeah, we were just discussing on this alert of LGTM bot.
> >
> > We can use const pointer here. I will commit the fix soon.
> >
> > Thanks,
> > rin

Thank you for the commit!

>
> Thanks for addressing it! I wonder whether there is performance impact
> here (is this hot-path code?).

The function is not used in any packet processing (hot paths)
and used only for route updates that are uncommon for most users.

  ozaki-r