netdev - Re: [Intel-wired-lan] [PATCH 12/15] ice: Add stats and ethtool support

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: <20180313141414.00007d11@intel.com>
Date:   Tue, 13 Mar 2018 14:14:14 -0700
From:   Jesse Brandeburg <jesse.brandeburg@...el.com>
To:     Eric Dumazet <eric.dumazet@...il.com>, davem@...emloft.net
Cc:     "Venkataramanan, Anirudh" <anirudh.venkataramanan@...el.com>,
        "kubakici@...pl" <kubakici@...pl>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        "intel-wired-lan@...ts.osuosl.org" <intel-wired-lan@...ts.osuosl.org>,
        jesse.brandeburg@...el.com
Subject: Re: [Intel-wired-lan] [PATCH 12/15] ice: Add stats and ethtool
 support

On Tue, 13 Mar 2018 12:17:10 -0700
Eric Dumazet <eric.dumazet@...il.com> wrote:
> 
> Yes, this is a recurring mistake
> 
> See commit
> bf909456f6a89654cb65c01fe83a4f9b133bf978 Revert "net: hns3: Add packet 
> statistics of netdev"

Thanks for the pointer, that was a useful thread to review.  I
understand the point that was made about not having the netdev stats
shown in ethtool -S.  We definitely do provide per-queue stats as well
as these regular stats in ethtool -S.

I do remember from the past discussions that it *is* useless for the
driver to keep internally any stats that were already stored via the
get_stats NDO, and we missed it in the internal review that this driver
was doing that, so that will be fixed.

Maybe it's just that I've been doing this too long, but I regularly
(and many other customers/users do as well) depend on the ethtool stats
being atomically updated w.r.t. each other.  This means that if I'm
getting the over rx_packets, as well as the per-queue rx_packets, and I
read them all at once from the driver with ethtool, then I can check
that things are working as expected.

If I have to gather the netdev stats from /proc/net/dev (which iproute2
tool shows the /proc/net/dev stats?) and somehow atomically gather the
ethtool -S stats.  What's the user to do in this brave new world where
ethtool doesn't at least have rx/tx bytes and packets?