Re: [PATCH net v2] net/ipv6: don't route packets with unknown source address

From: Íñigo Huguet

Date: Mon Sep 07 2026 - 03:36:15 EST


On Thu, Sep 3, 2026 at 4:57 PM Ido Schimmel <idosch@xxxxxxxxxx> wrote:
> 1. This is a behavior change, not a bug fix, and should be targeted at
> net-next without a Fixes tag.

I see it as a bug. If you remove an address or its valid_lft expires,
the expectation is that it's not used any more. Anyway, even if we
consider it a bug, it's certainly not urgent, so it's OK for me to
send it to net-next instead.

> 2. What is the motivation for this drastic change beyond RFC conformance
> and parity with IPv4? IMO, these two are not a good enough reason to
> make such a change with a huge blast radius.

IMHO conforming to RFCs is always a strong reason. But beyond that, I
think that the current behavior is wrong because, as I've said above,
keeping using an address that has been removed or expired is not OK.

It's true that it's difficult to spot all the corner cases that may be
affected by the changes, so the tradeoff of fixing vs risk of
regressions is not that clear, I understand it.

> 3. See [1] for a list of regressions that AI flagged. Even if v3 fixes
> all of them (which means a much bigger diff), I don't think such a
> change will be merged without a proper real-world motivation beyond RFC
> conformance and IPv4 parity.

I will try to address all of them; thanks.

About the real-world motivation: this comes from a customer's bug
report, who is worried about packet leaks in their production
environment. I mentioned IPv4 parity and RFC compliance because I
thought that without those arguments it would be very difficult for
the patch to be accepted.


On Fri, Sep 4, 2026 at 12:38 AM Jakub Kicinski <kuba@xxxxxxxxxx> wrote:
> breaks a bunch of selftests, too, FWIW
> Please make sure to mend those if you want to forge ahead as a non-fix

Sure!


On Fri, Sep 4, 2026 at 12:22 PM Matthieu Baerts <matttbe@xxxxxxxxxx> wrote:
> Just in case, do not hesitate to look at the new countainers that might
> help you running those tests locally:

I didn't know about this, thank you!

--
Íñigo Huguet