netdev - Re: mpls_xmit() calls skb

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

Message-ID: <20231208154446.0feee63f@kernel.org>
Date: Fri, 8 Dec 2023 15:44:46 -0800
From: Jakub Kicinski <kuba@...nel.org>
To: Christoph Paasch <cpaasch@...le.com>
Cc: netdev <netdev@...r.kernel.org>, Roopa Prabhu <roopa@...dia.com>, Craig
 Taylor <cmtaylor@...le.com>
Subject: Re: mpls_xmit() calls skb_orphan()

Updating Roopa's email.

On Fri, 08 Dec 2023 13:15:04 -0800 Christoph Paasch wrote:
> > On Dec 8, 2023, at 1:06 PM, Christoph Paasch <cpaasch@...le.com> wrote:
> > 
> > Hello,
> > 
> > we observed an issue when running a TCP-connection with BBR on top of an MPLS-tunnel in that we saw a lot of CPU-time spent coming from tcp_pace_kick(), although sch_fq was configured on this host.
> > 
> > The reason for this seems to be because mpls_xmit() calls skb_orphan(), thus settings skb->sk to NULL, preventing the qdisc to set sk_pacing_status (which would allow to avoid the call to tcp_pace_kick()).
> > 
> > The question is: Why is this call to skb_orphan in mpls_xmit necessary ?
> > 
> > 
> > Thanks,
> > Christoph  
> 
> Resend in plain-text. Sorry about that!