netdev - Re: [PATCH ethtool v2 1/2] ethtool: add support for get/set ethtool

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: <20200719234724.4dhsd547s7od4uyg@lion.mk-sys.cz>
Date:   Mon, 20 Jul 2020 01:47:24 +0200
From:   Michal Kubecek <mkubecek@...e.cz>
To:     Govindarajulu Varadarajan <gvaradar@...co.com>
Cc:     netdev@...r.kernel.org, edumazet@...gle.com,
        linville@...driver.com, govind.varadar@...il.com, benve@...co.com
Subject: Re: [PATCH ethtool v2 1/2] ethtool: add support for get/set
 ethtool_tunable

On Fri, Jul 17, 2020 at 07:59:49AM -0700, Govindarajulu Varadarajan wrote:
> Add support for ETHTOOL_GTUNABLE and ETHTOOL_STUNABLE options.
> 
> Tested rx-copybreak on enic driver. Tested ETHTOOL_TUNNABLE_STRING

A typo: TUNNABLE -> TUNABLE

> options with test/debug changes in kernel.
> 
> Signed-off-by: Govindarajulu Varadarajan <gvaradar@...co.com>
> ---

This looks good to me but I'm still not happy with the string tunables
handling. The reason I asked about it was to find out if I missed some
important piece of information. But it doesn't seem to be the case so
that the situation looks like this:

  - there is no documentation telling us how they should work
  - there is no kernel or userspace code yet (except this patch)
  - there is no string tunable yet
  - we don't even know if there is ever going to be any
  - proposed code is inconsistent (it allows passing value to kernel
    which it would not be able to receive back from kernel)
  - it adds extra complexity to do_gtunable() and do_stunable()
    (special handling and allocating a new buffer in each iteration)
  - it's dead code anyway: the way the interface is designed, current
    ethtool cannot get/set future tunables it does not recognize)

Therefore I suggest to drop handling of string tunables until there is
actually a string tunable and we get (preferrably documented) consensus
on how the interface should behave. (Which may very well never happen.)

Michal

Download attachment "signature.asc" of type "application/pgp-signature" (489 bytes)