ACK/Cmnt: [SRU][F][J][PATCH 0/3] ip/nexthop: fix default address selection for connected nexthop

[Stefan Bader]

On 08.09.22 21:56, Luke Nowakowski-Krijger wrote:
> These patches are based on an SRU request from Nicolas Dichtel.
> 
> BugLink: https://bugs.launchpad.net/bugs/1988809
> 
> [Backport]
> 
> For both jammy and focal needed to replace dev_hold_track with
> dev_hold as dev_hold_track uses debugging infastructure that we don't
> have in our kernels/don't want to backport.
> 
> Focal needed slight context adjustments in makefile for the
> fib_nexthop_nongw selftest.
> 
> [Impact]
> 
> Packets sent by userland apps are rejected/dropped if the source address is not specified and the corresponding route is using a connected nexthop object.
> 
> This bug exists since linux v5.3 and has been fixed in v5.19 by the following upstream commits:
>   - 747c14307214 ("ip: fix dflt addr selection for connected nexthop")
>     https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=747c14307214
>   - cd72e61bad14 ("selftests/net: test nexthop without gw")
>     https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=cd72e61bad14
>   - eb55dc09b5dd ("ip: fix triggering of 'icmp redirect'")
>     https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=eb55dc09b5dd
> 
> The last commit (v6.0) fixes a regression introduced by the first commit.
> 
> [Test Case]
> 
> A detailed test case is explained in the first commit and a self-test is added in the second commit.
> 
> Lukenow: Ran selftest which fails before patches applied and passes
> after they are applied.
> 
> [Regression Potential]
> 
> The patch modifies some internal routing states. It has been living in the upstream trees for 2 months and the reported regression about icmp redirects has been fixed.
> The risk of regression should be contained.
> 
> 
> Nicolas Dichtel (3):
>    ip: fix dflt addr selection for connected nexthop
>    ip: fix triggering of 'icmp redirect'
>    selftests/net: test nexthop without gw
> 
>   net/ipv4/fib_frontend.c                       |   4 +-
>   net/ipv4/fib_semantics.c                      |   2 +-
>   tools/testing/selftests/net/Makefile          |   2 +-
>   .../selftests/net/fib_nexthop_nongw.sh        | 119 ++++++++++++++++++
>   4 files changed, 123 insertions(+), 4 deletions(-)
>   create mode 100755 tools/testing/selftests/net/fib_nexthop_nongw.sh
> 

Is there any difference between the two #1 and #2 patches? I slightly doubt it 
as #2 at least is a cherry pick in both cases. If there were any difference, it 
is impossible to tell which is for F and which for J since the subject is the 
same. For future reference this should better be prepared to have #1 and #2 only 
once.

Acked-by: Stefan Bader <stefan.bader at canonical.com>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <https://lists.ubuntu.com/archives/kernel-team/attachments/20220915/6a115dfa/attachment.sig>