Re: FreeBSD guest on Linux KVM, some vtnet(4) packets don't arrive

we can repro this in a ppc64be VM using the 2026-03-30 ppc64 (be) snapshot.

I'm running it on qemu on freebsd-x86 as a host.



-adrian

On Thu, 9 Apr 2026 at 06:53, Michael Tuexen
<[email protected]> wrote:
>
> > On 9. Apr 2026, at 15:27, Lexi Winter <[email protected]> wrote:
> >
> > Michael Tuexen wrote in <[email protected]>:
> >>> On 9. Apr 2026, at 13:39, Lexi Winter <[email protected]> wrote:
> >>> i was planning to test this, but i ran into a new problem which is that
> >>> we seem to have broken crypto (TLS, SSH) on ppc64 at some point, so once
> >>> i work out what's going on there, i'll come back to this issue :-)
> >>
> >> On big endian PPC?
> >
> > yes.  fetch(1) is broken:
> >
> > root@ppc64-main:~ # fetch https://bbc.co.uk
> > 0000000801C59010:error:0A000119:SSL routines:tls_get_more_records:decryption failed or bad record mac:/usr/src/crypto/openssl/ssl/record/methods/tls_common.c:869:
> > 0000000801C59010:error:0A000139:SSL routines::record layer failure:/usr/src/crypto/openssl/ssl/record/rec_layer_s3.c:703:
> >
> > as is openssl(1):
> >
> > root@ppc64-main:~ # openssl s_client -connect www.freebsd.org:443
> > Connecting to 2a02:80:0:3ffd::50:3
> > CONNECTED(00000003)
> > 0000000801D1E010:error:0A000119:SSL routines:tls_get_more_records:decryption failed or bad record mac:/usr/src/crypto/openssl/ssl/record/methods/tls_common.c:869:
> > 0000000801D1E010:error:0A000139:SSL routines::record layer failure:/usr/src/crypto/openssl/ssl/record/rec_layer_s3.c:703:
> >
> > and sshd(8):
> >
> > % ssh root@fd00:0:0:1::b
> > ssh_dispatch_run_fatal: Connection to fd00:0:0:1::b port 22: message authentication code incorrect
> >
> > using:
> >
> > FreeBSD 16.0-CURRENT #0 main-n284768-e6eba5076929: Mon Mar 30 16:07:28 UTC 2026
> >    [email protected]:/usr/obj/usr/src/powerpc.powerpc64/sys/GENERIC64 powerpc
> OK, I am updating my PPC machine.
>
> Please note that the above could result from a problem in the crypto code or
> in data corruption in the network stack...
>
> Best regards
> Michael
>
>