netdev - Re: [PATCH net-next ct-offload v3 00/15] Introduce connection tracking offload

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: <20200311224415.GL3614@localhost.localdomain>
Date:   Wed, 11 Mar 2020 19:44:15 -0300
From:   Marcelo Ricardo Leitner <marcelo.leitner@...il.com>
To:     Paul Blakey <paulb@...lanox.com>
Cc:     Saeed Mahameed <saeedm@...lanox.com>,
        Oz Shlomo <ozsh@...lanox.com>,
        Vlad Buslov <vladbu@...lanox.com>,
        David Miller <davem@...emloft.net>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        Jiri Pirko <jiri@...lanox.com>, Roi Dayan <roid@...lanox.com>
Subject: Re: [PATCH net-next ct-offload v3 00/15] Introduce connection
 tracking offload

On Thu, Mar 12, 2020 at 12:27:37AM +0200, Paul Blakey wrote:
> 
> On 11/03/2020 21:13, Marcelo Ricardo Leitner wrote:
> > On Wed, Mar 11, 2020 at 04:33:43PM +0200, Paul Blakey wrote:
> >> Applying this patchset
> >> --------------------------
> >>
> >> On top of current net-next ("r8169: simplify getting stats by using netdev_stats_to_stats64"),
> >> pull Saeed's ct-offload branch, from git git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux.git
> >> and fix the following non trivial conflict in fs_core.c as follows:
> >> #define OFFLOADS_MAX_FT 2
> >> #define OFFLOADS_NUM_PRIOS 2
> >> #define OFFLOADS_MIN_LEVEL (ANCHOR_MIN_LEVEL + OFFLOADS_NUM_PRIOS)
> >>
> >> Then apply this patchset.
> > I did this and I couldn't get tc offloading (not ct) to work anymore.
> > Then I moved to current net-next (the commit you mentioned above), and
> > got the same thing.
> >
> > What I can tell so far is that 
> > perf script | head
> >        handler11  4415 [009]  1263.438424: probe:mlx5e_configure_flower__return: (ffffffffc094fa80 <- ffffffff93dc510a) arg1=0xffffffffffffffa1
> >
> > and that's EOPNOTSUPP. Not sure yet where that is coming from.
> 
> I guess you fail at what this patch "flow_offload: fix allowed types check" is suppose to fix
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next.git/commit/?id=a393daa8993fd7d6c9c33110d5dac08bc0dc2696
> 
> That was the last bug that caused what you decribe - all tc rules failed.
> 
> It should be before the patch I detailed as seen in git log here:
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next.git/log/
> 
> 
> Can you check you have it? im not sure if compiling just the driver is enough.

I had this one, in both tests.

> 
> if it is this, it should have fixed it.
> 
> if not try skipping flow_action_hw_stats_types_check() calls in mlx5 driver.

Ok. This one was my main suspect now after some extra printks. I could
confirm it parse_tc_fdb_actions is returning the error, but not sure
why yet.

> 
> there is a netlink error msg most of the cases, try skip_sw or verbose to see.

Yeah.. that probably would be easier, thanks. I'm using it under OvS
and without skip_sw, got just no logs for it.

> 
> >
> > Btw, it's prooving to be a nice exercise to find out why it is failing
> > to offload. Perhaps some netdev_err_once() is welcomed.
> >
> >   Marcelo