Re: Bug report connect to VM with Vagrant

[Leon Romanovsky <leon@kernel.org>]

On Tue, Jan 02, 2024 at 10:49:58AM +0100, Eric Dumazet wrote:
> On Wed, Dec 27, 2023 at 9:33 AM Leon Romanovsky <leon@kernel.org> wrote:
> >
> > On Fri, Dec 15, 2023 at 10:55:05AM +0100, Linux regression tracking (Thorsten Leemhuis) wrote:
> > > On 08.12.23 11:49, Eric Dumazet wrote:
> > > > On Thu, Dec 7, 2023 at 2:03 PM Shachar Kagan <skagan@nvidia.com> wrote:
> > > >>>> On Thu, Nov 30, 2023 at 2:55 PM Shachar Kagan <skagan@nvidia.com> wrote:
> > > >>>>
> > > >>>> I have an issue that bisection pointed at this patch:
> > > >>>> commit 0a8de364ff7a14558e9676f424283148110384d6
> > > >>>> tcp: no longer abort SYN_SENT when receiving some ICMP
> > > >>>
> > > >>> Please provide tcpdump/pcap captures.
> > > >>>
> > > >>>  It is hard to say what is going on just by looking at some application logs.
> > > >>
> > > >> I managed to capture the tcpdump of ‘Vagrant up’ step over old kernel and new kernel where this step fails. Both captures are attached.
> > > >> The tcpdump is filtered by given IP of the nested VM.
> > > >
> > > > I do not see any ICMP messages in these files, can you get them ?
> > > >
> > > > Feel free to continue this exchange privately, no need to send MB
> > > > email to various lists.
> > >
> > > Here this thread died, so I assume this turned out to be not a
> > > regression at all or something like that? If not please speak up!
> >
> > No, it wasn't fixed and/or reverted. Right now, Vagrant is broken and
> > all our regressions around nested VM functionality doesn't run.
> >
> > Eric, can you please revert the bisected patch while you are continuing
> > your offline discussion with Shachar?
> >
> 
> This is not how things work.
> 
> I have not received any evidence yet, only partial packet dumps with
> no ICMP messages that could be related to the 'Vagrant issue'

Revert of the original patch worked, so it is strong enough evidence to do
not break very popular orchestration software.

> 
> Patch is adhering to the RFC.
> 
> If an application wants to have fast reaction to ICMP, it must use
> appropriate socket options instead of relying on a prior
> implementation detail.

Maybe yes, maybe not. Right now, Vagrant is broken.

Thanks