netdev - Re: [PATCH net v1] net/sched: Don't print dump stack in event of transmission timeout

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 for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20200413105408.GJ334007@unreal>
Date:   Mon, 13 Apr 2020 13:54:08 +0300
From:   Leon Romanovsky <leon@...nel.org>
To:     Jose Abreu <Jose.Abreu@...opsys.com>
Cc:     "David S. Miller" <davem@...emloft.net>,
        Jakub Kicinski <kuba@...nel.org>,
        Arjan van de Ven <arjan@...ux.intel.com>,
        Cong Wang <xiyou.wangcong@...il.com>,
        Jamal Hadi Salim <jhs@...atatu.com>,
        Jiri Pirko <jiri@...nulli.us>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: [PATCH net v1] net/sched: Don't print dump stack in event of
 transmission timeout

On Mon, Apr 13, 2020 at 10:37:24AM +0000, Jose Abreu wrote:
> From: Leon Romanovsky <leon@...nel.org>
> Date: Apr/13/2020, 11:20:53 (UTC+00:00)
>
> > On Mon, Apr 13, 2020 at 09:01:32AM +0000, Jose Abreu wrote:
> > > From: Leon Romanovsky <leon@...nel.org>
> > > Date: Apr/12/2020, 07:08:54 (UTC+00:00)
> > >
> > > > [  281.170584] ------------[ cut here ]------------
> > >
> > > Not objecting to the patch it-self (because usually stack trace is
> > > useless), but just FYI we use this marker in our CI to track for timeouts
> > > or crashes. I'm not sure if anyone else is using it.
> >
> > I didn't delete the "NETDEV WATCHDOG .." message and it will be still
> > visible as a marker.
> >
> > >
> > > And actually, can you please explain why BQL is not suppressing your
> > > timeouts ?
> >
> > Driver can't distinguish between "real" timeout and "mixed traffic" timeout,
>
> The point is that you should not get any "mixed traffic" timeout if the
> driver uses BQL because Queue will be disabled long before timeout happens
> as per queue size usage ...

Sorry, if I misunderstood you, but you are proposing to count traffic, right?

If yes, RDMA traffic bypasses the SW stack and not visible to the kernel, hence
the BQL will count only ETH portion of that mixed traffic, while RDMA traffic
is the one who "blocked" transmission channel (QP in RDMA terminology).

Thanks

>
> ---
> Thanks,
> Jose Miguel Abreu