netdev - Re: IPv6 TCP seems broken on 32bit x86 (bisected)

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <CABq1_vgE_nzcGPUf0r76xrNchcKUXSHfHCZs4EsXedXCTu-nNg@mail.gmail.com>
Date: Thu, 24 Aug 2023 14:09:27 +0200
From: Klara Modin <klarasmodin@...il.com>
To: Pavel Begunkov <asml.silence@...il.com>
Cc: netdev@...r.kernel.org, edumazet@...gle.com, davem@...emloft.net, 
	simon.horman@...igine.com
Subject: Re: IPv6 TCP seems broken on 32bit x86 (bisected)

Den tors 24 aug. 2023 kl 13:28 skrev Pavel Begunkov <asml.silence@...il.com>:
>
> On 8/24/23 01:00, Klara Modin wrote:
> > Hi,
> >
> > I recently noticed that IPv6 stopped working properly on my 32 bit x86
> > machines with recent kernels. Bisecting leads to
> > "fe79bd65c819cc520aa66de65caae8e4cea29c5a net/tcp: refactor
> > tcp_inet6_sk()". Reverting this commit on top of
> > a5e505a99ca748583dbe558b691be1b26f05d678 fixes the issue.
> > 64 bit x86 seems unaffected.
>
> It should be same to what was discussed here
>
> https://lore.kernel.org/netdev/CANn89iLTn6vv9=PvAUccpRNNw6CKcXktixusDpqxqvo+UeLviQ@mail.gmail.com/T/#eb8dca9fc9491cc78fca691e64e601301ec036ce7
>
> Thanks to Eric it should be fixed upstream. Let me check if
> it's queued for stable kernels.

That does solve the issue, thanks!

>
>
> > One of the symptoms seems to be that the hop limit is sometimes set to
> > zero. Attached is tcpdump output from trying to ssh into the machine
> > from a different subnet.
> >
> > I also tried http which seems even weirder, the source address is cut
> > off or offsetted by 32 bits (results in
> > "a5c:1204:84c4:847e:950e:2b3d::", should be
> > "2001:678:a5c:1204:84c4:847e:950e:2b3d"). Trying to connect using
> > netcat sees the same result as with http.
> >
> > Tested on two machines running Gentoo, one with a locally compiled
> > kernel and the other compiled from a x86_64 host, both using gcc 13.2.
> > If I remember correctly I was using gcc 12.3.1 on the first machine
> > when I initially noticed the issue but didn't troubleshoot further at
> > the time.
> >
> > Please tell me if there's any other information you need.
> >
> > Regards,
> > Klara Modin
>
> --
> Pavel Begunkov