netdev - Re: [net-next 08/16] net/mlx5e: Add devlink fdb_large

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:   Mon, 2 Mar 2020 11:31:07 -0800
From:   Jakub Kicinski <kuba@...nel.org>
To:     Saeed Mahameed <saeedm@...lanox.com>
Cc:     Vlad Buslov <vladbu@...lanox.com>, Jiri Pirko <jiri@...lanox.com>,
        Roi Dayan <roid@...lanox.com>,
        "davem@...emloft.net" <davem@...emloft.net>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        Jianbo Liu <jianbol@...lanox.com>
Subject: Re: [net-next 08/16] net/mlx5e: Add devlink fdb_large_groups
 parameter

On Mon, 2 Mar 2020 18:48:22 +0000 Saeed Mahameed wrote:
> On Fri, 2020-02-28 at 11:10 -0800, Jakub Kicinski wrote:
> > On Thu, 27 Feb 2020 16:44:38 -0800 Saeed Mahameed wrote:  
> > > The size of each large group can be calculated according to the
> > > following
> > > formula: size = 4M / (fdb_large_groups + 1).
> >
> > Slicing memory up sounds like something that should be supported via
> > the devlink-resource API, not by params and non-obvious calculations
> > :(  
> 
> Hi Jakub, you have a point, but due to to the non-triviality of the
> internal mlnx driver and FW architecture of handling internal FDB table
> breakdown, we preferred to go with one driver-specific parameter to
> simplify the approach, instead of 3 or 4 generic params, which will not
> make any sense to other vendors for now.

Actually I was hoping this can be made into some resource attribute,
rather than a generic parameter. The formula in the commit message looks
very much like there is a resource of 4M "things" which is subdivided
into "large groups".

Maybe if Jiri acked it it's not a great fit.

> As always we will keep an eye on what other vendors are doing and will
> try to unify with a generic set of params once other vendors show
> interest of a similar thing.

Ah, yes :)