Re: git: d6c4cea7740d - main - loopback: improve checksum offloading

> On 1. Jul 2026, at 06:16, Gleb Smirnoff <glebius@FreeBSD.org> wrote:
> 
> On Sun, Jun 28, 2026 at 10:57:58AM +0000, Michael Tuexen wrote:
> M> commit d6c4cea7740d5c5c673a06ba37e4f1bdcddb2ece
> M> Author:     Timo Völker <timo.voelker@fh-muenster.de>
> M> AuthorDate: 2026-06-28 10:50:15 +0000
> M> Commit:     Michael Tuexen <tuexen@FreeBSD.org>
> M> CommitDate: 2026-06-28 10:50:15 +0000
> M> 
> M>     loopback: improve checksum offloading
> M>     
> M>     * Allow disabling IFCAP_RXCSUM_IPV6 or IFCAP_TXCSUM_IPV6.
> M>     * Do not pretend the checksum is correct by setting the LO_CSUM_SET
> M>       flags if IFCAP_RXCSUM_IPV6 or IFCAP_RXCSUM is enabled. Instead,
> M>       remove the LO_CSUM_SET flags (in case they have been set somehow)
> M>       if IFCAP_RXCSUM_IPV6 or IFCAP_RXCSUM is disabled.
> M>     * Do not unset the transmit checksum offload flags LO_CSUM_FEATURES or
> M>       LO_CSUM_FEATURES6 since they now have a meaning for the receive path.
> M>     
> M>     Reviewed by:            glebius, pouria, tuexen
> M>     Okayed by:              bz
> M>     MFC after:              1 week
> M>     Differential Revision:  https://reviews.freebsd.org/D57518
> 
> I'm guessing this change  is responsible for breaking the
> netinet/ip_reass_test. The tests inject packets to localhost.  Now all
> subtests fail, but after each test that creates single packet I see in 'netstat
> -sp ip' increase in 'bad header checksums'.
> 
> Can you please check this?
We will do. Thanks for notifying.

Best regards
Michael
> 
> -- 
> Gleb Smirnoff