netdev - Re: [PATCH ipsec] xfrm: check DST_NOPOLICY as well as DST

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:   Thu, 05 Dec 2019 10:05:33 +0000
From:   Mark Gillott <mgillott@...tta.att-mail.com>
To:     nicolas.dichtel@...nd.com, netdev@...r.kernel.org
Cc:     steffen.klassert@...unet.com, herbert@...dor.apana.org.au
Subject: Re: [PATCH ipsec] xfrm: check DST_NOPOLICY as well as DST_NOXFRM

On Thu, 2019-12-05 at 09:52 +0100, Nicolas Dichtel wrote:
> Le 05/12/2019 à 09:10, Mark Gillott a écrit :
> > On Wed, 2019-12-04 at 17:57 +0100, Nicolas Dichtel wrote:
> > > Le 04/12/2019 à 16:17, Mark Gillott a écrit :
> > > > Before performing a policy bundle lookup, check the
> > > > DST_NOPOLICY
> > > > option, as well as DST_NOXFRM. That is, skip further processing
> > > > if
> > > > either of the disable_policy or disable_xfrm sysctl attributes
> > > > are
> > > > set.
> > > 
> > > Can you elaborate why this change is needed?
> > 
> > We have a separate DPDK-based dataplane that is responsible for all
> > IPsec processing - policy handing/encryption/decryption.
> > Consequently
> > we set the net.ipv[4|6].conf.<if>.disable_policy sysctl to 1 for
> > all
> > "interesting" interfaces. That is we want the kernel to ignore any
> > IPsec policies.
> > 
> > Despite the above & depending on configuration, we found that
> > originating traffic was ending up deep inside XFRM where it would
> > get
> > dropped because of a route lookup problem.
> 
> And why don't you set disable_xfrm to thoses interfaces also?
> disable_policy means no xfrm policy lookup on output, disable_xfrm
> means no xfrm
> policy check on input.
> 

True, setting disable_xfrm=1 would solve the issue. Except this is
output - the test case is a ping from a peer, the corresponding ICMP
response is discarded by the kernel. Feels like disable_policy is the
right check (the kernel is doing XFRM output).

Cheers,
Mark