netdev - Re: [net-next 8/8] net/mlx5: Add vf ACL access via tc flower

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: <20191113121943.4df01958@cakuba>
Date:   Wed, 13 Nov 2019 12:19:43 -0800
From:   Jakub Kicinski <jakub.kicinski@...ronome.com>
To:     Saeed Mahameed <saeedm@....mellanox.co.il>
Cc:     Saeed Mahameed <saeedm@...lanox.com>,
        "David S. Miller" <davem@...emloft.net>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        Ariel Levkovich <lariel@...lanox.com>
Subject: Re: [net-next 8/8] net/mlx5: Add vf ACL access via tc flower

On Tue, 12 Nov 2019 16:31:19 -0800, Saeed Mahameed wrote:
> On Tue, Nov 12, 2019 at 3:41 PM Jakub Kicinski wrote:
> > On Tue, 12 Nov 2019 17:13:53 +0000, Saeed Mahameed wrote:  
> > > From: Ariel Levkovich <lariel@...lanox.com>
> > >
> > > Implementing vf ACL access via tc flower api to allow
> > > admins configure the allowed vlan ids on a vf interface.
> > >
> > > To add a vlan id to a vf's ingress/egress ACL table while
> > > in legacy sriov mode, the implementation intercepts tc flows
> > > created on the pf device where the flower matching keys include
> > > the vf's mac address as the src_mac (eswitch ingress) or the
> > > dst_mac (eswitch egress) while the action is accept.
> > >
> > > In such cases, the mlx5 driver interpets these flows as adding
> > > a vlan id to the vf's ingress/egress ACL table and updates
> > > the rules in that table using eswitch ACL configuration api
> > > that is introduced in a previous patch.  
> >
> > Nack, the magic interpretation of rules installed on the PF is a no go.  
> 
> PF is the eswitch manager it is legit for the PF to forward rules to
> the eswitch FDB,
> we do it all over the place, this is how ALL legacy ndos work, why
> this should be treated differently ?

It's not a legacy NDO, there's little precedent for it, and you're
inventing a new meaning for an operation.

> Anyway just for the record, I don't think you are being fair here, you
> just come up with rules on the go just to block anything related to
> legacy mode.

I tried to block everything related to legacy NDOs for a while now, and
I'm not the only one (/me remembers Or in netdevconf 1.1). I'm sorry but
I won't go and dig out the links now, it's a waste of time.

Maybe we differ on the definition of fairness. I'm against this exactly
_because_ I'm fair, nobody gets a free pass, no matter how much we
otherwise appreciate given company contributing to the kernel...