netdev - RE: [Question]: TCP resets when using ECMP for load-balancing between multiple servers.

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]

Date: Wed, 16 Aug 2023 07:52:24 +0000
From: Sriram Yagnaraman <sriram.yagnaraman@....tech>
To: Stephen Hemminger <stephen@...workplumber.org>
CC: "netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: RE: [Question]: TCP resets when using ECMP for load-balancing between
 multiple servers.



> -----Original Message-----
> From: Stephen Hemminger <stephen@...workplumber.org>
> Sent: Tuesday, 15 August 2023 23:53
> To: Sriram Yagnaraman <sriram.yagnaraman@....tech>
> Cc: netdev@...r.kernel.org
> Subject: Re: [Question]: TCP resets when using ECMP for load-balancing
> between multiple servers.
> 
> On Tue, 15 Aug 2023 22:10:48 +0200
> Sriram Yagnaraman <sriram.yagnaraman@....tech> wrote:
> 
> > diff --git a/include/uapi/linux/in_route.h
> > b/include/uapi/linux/in_route.h index 0cc2c23b47f8..01ae06c7743b
> > 100644
> > --- a/include/uapi/linux/in_route.h
> > +++ b/include/uapi/linux/in_route.h
> > @@ -15,6 +15,7 @@
> >  #define RTCF_REDIRECTED	0x00040000
> >  #define RTCF_TPROXY	0x00080000 /* unused */
> >
> > +#define RTCF_MULTIPATH	0x00200000
> >  #define RTCF_FAST	0x00200000 /* unused */
> 
> This looks like a bad idea. Reusing bits that were unused in the past and
> exposed through uapi will likely cause an ABI breakage.

Thanks, I agree, the code I posted is just a hack to test the hypothesis if route hints are the root cause.

It would be nice to get some guidance/suggestion on
- Is this really a problem, and should be fixed?
- What is the best way to solve this for ipv4 and ipv6?

I can propose a proper fix after that.