Re: [CFT/Review] DPAA2 SFP+, LX2160A eSDHC, Thunderbolt bridge, FreeBSD on Apple Silicon (vz), arm64 16K pages

Warner Losh <[email protected]> writes:

> On Tue, Jul 14, 2026 at 1:28=E2=80=AFPM Dmitry Salychev <[email protected]>=
 wrote:
>
>  Adrian Chadd <[email protected]> writes:
>
>  > On Mon, 6 Jul 2026 at 16:39, =E5=AD=94=E7=A5=A5=E6=B3=A2 <yarshure@gma=
il.com> wrote:
>  >>
>  >>
>  >> Hi hackers,
>  >>
>  >> I'd like to share a few things I've been working on recently around
>  >> FreeBSD/arm64, and I'm looking for testers and code review before
>  >> attempting to upstream the relevant parts.
>  >>
>  >> 1) DPAA2: SFP+ now works on LX2160A (runtime configuration support)
>  >>
>  >>    https://github.com/freebsd/freebsd-src/compare/main...networkexten=
sion:freebsd-src:dpaa2-runtime-config
>  >>
>  >
>  > hi!
>  >
>  > This looks good (and dsl@ already replied, which is a good sign.)
>  >
>  > I think it'd be good to get this up as a specific review candidate /
>  > pull request so dsl@, jhibbits@, bz@ and others can
>  > dig into it some more.
>  >
>  >
>  >
>  > -adrian
>
>  I'm pretty sure that code was generated by LLM and the project's opinion
>  regarding such "implementation" is clear IIRC. Personally, I wouldn't
>  want to spend my time reviewing changes which won't be accepted anyway.
>
> I don't think the project has a coherent view about LLM genreated code. T=
here's
> no prohibition, core hasn't announced any bans, etc. There was much fear =
about
> issues that turned out to not be issues.
>
> I've personally landed a dozen changes that were at least assisted by LLM
> that I drove to completion. I've landed a bunch more where different code=
 completion
> tools were used, which is a weaker version of LLM generated. IMHO, The cu=
rrent
> 'center of opinion' seems to be that thoughtful use of LLM is OK, as long=
 as the
> results meet our minimum quality standards. There's also a strong undercu=
rrent of
> "and the submitter has to understand it well enough to maintain the code"=
 as well,
> which is something the review process should probe (but that's no difffer=
ent than
> what I do with completely human generated code: why did you do this and n=
ot
> that? etc).
>
> So the code could be accepted, if it meets the stringent quality
> demands of FreeBSD.

Okay, it makes sense to me. Generally, I'd try to look at phab review if
there will be any regarding DPAA2 stuff.

Regards,
Dmitry

>
> Warner
>=20=20
>  However, I'm open to consider other points of view.
>
>  Regards,
>  Dmitry
>
>  --=20
>  https://wiki.freebsd.org/DmitrySalychev

--=20
https://wiki.freebsd.org/DmitrySalychev