Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Hans Schultz <schultz.hans@gmail.com>
To:     Vladimir Oltean <olteanv@gmail.com>,
        Hans Schultz <schultz.hans@gmail.com>
Cc:     Andrew Lunn <andrew@lunn.ch>, davem@davemloft.net, kuba@kernel.org,
        netdev@vger.kernel.org, Vivien Didelot <vivien.didelot@gmail.com>,
        Florian Fainelli <f.fainelli@gmail.com>,
        Jiri Pirko <jiri@resnulli.us>,
        Ivan Vecera <ivecera@redhat.com>,
        Roopa Prabhu <roopa@nvidia.com>,
        Nikolay Aleksandrov <razor@blackwall.org>,
        Daniel Borkmann <daniel@iogearbox.net>,
        Ido Schimmel <idosch@nvidia.com>, linux-kernel@vger.kernel.org,
        bridge@lists.linux-foundation.org
Subject: Re: [PATCH net-next 3/3] net: dsa: mv88e6xxx: mac-auth/MAB
 implementation
In-Reply-To: <20220318121400.sdc4guu5m4auwoej@skbuf>
References: <86r17495gk.fsf@gmail.com>
 <20220316233447.kwyirxckgancdqmh@skbuf> <86lex9hsg0.fsf@gmail.com>
 <YjNDgnrYaYfviNTi@lunn.ch> <20220317153625.2ld5zgtuhoxbcgvo@skbuf>
 <86ilscr2a4.fsf@gmail.com> <20220317161808.psftauoz5iaecduy@skbuf>
 <8635jg5xe5.fsf@gmail.com> <20220317172013.rhjvknre5w7mfmlo@skbuf>
 <86tubvk24r.fsf@gmail.com> <20220318121400.sdc4guu5m4auwoej@skbuf>
Date:   Fri, 18 Mar 2022 14:10:26 +0100
Message-ID: <86pmmjieyl.fsf@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

On fre, mar 18, 2022 at 14:14, Vladimir Oltean <olteanv@gmail.com> wrote:
> On Fri, Mar 18, 2022 at 11:04:36AM +0100, Hans Schultz wrote:
>> On tor, mar 17, 2022 at 19:20, Vladimir Oltean <olteanv@gmail.com> wrote:
>> > On Thu, Mar 17, 2022 at 05:58:26PM +0100, Hans Schultz wrote:
>> >> On tor, mar 17, 2022 at 18:18, Vladimir Oltean <olteanv@gmail.com> wrote:
>> >> > On Thu, Mar 17, 2022 at 05:07:15PM +0100, Hans Schultz wrote:
>> >> >> On tor, mar 17, 2022 at 17:36, Vladimir Oltean <olteanv@gmail.com> wrote:
>> >> >> > On Thu, Mar 17, 2022 at 03:19:46PM +0100, Andrew Lunn wrote:
>> >> >> >> On Thu, Mar 17, 2022 at 09:52:15AM +0100, Hans Schultz wrote:
>> >> >> >> > On tor, mar 17, 2022 at 01:34, Vladimir Oltean <olteanv@gmail.com> wrote:
>> >> >> >> > > On Mon, Mar 14, 2022 at 11:46:51AM +0100, Hans Schultz wrote:
>> >> >> >> > >> >> @@ -396,6 +414,13 @@ static irqreturn_t mv88e6xxx_g1_atu_prob_irq_thread_fn(int irq, void *dev_id)
>> >> >> >> > >> >>  				    "ATU miss violation for %pM portvec %x spid %d\n",
>> >> >> >> > >> >>  				    entry.mac, entry.portvec, spid);
>> >> >> >> > >> >>  		chip->ports[spid].atu_miss_violation++;
>> >> >> >> > >> >> +		if (mv88e6xxx_port_is_locked(chip, chip->ports[spid].port))
>> >> >> >> > >> >> +			err = mv88e6xxx_switchdev_handle_atu_miss_violation(chip,
>> >> >> >> > >> >> +									    chip->ports[spid].port,
>> >> >> >> > >> >> +									    &entry,
>> >> >> >> > >> >> +									    fid);
>> >> >> >> > >> >
>> >> >> >> > >> > Do we want to suppress the ATU miss violation warnings if we're going to
>> >> >> >> > >> > notify the bridge, or is it better to keep them for some reason?
>> >> >> >> > >> > My logic is that they're part of normal operation, so suppressing makes
>> >> >> >> > >> > sense.
>> >> >> >> > >> >
>> >> >> >> > >>
>> >> >> >> > >> I have been seeing many ATU member violations after the miss violation is
>> >> >> >> > >> handled (using ping), and I think it could be considered to suppress the ATU member
>> >> >> >> > >> violations interrupts by setting the IgnoreWrongData bit for the
>> >> >> >> > >> port (sect 4.4.7). This would be something to do whenever a port is set in locked mode?
>> >> >> >> > >
>> >> >> >> > > So the first packet with a given MAC SA triggers an ATU miss violation
>> >> >> >> > > interrupt.
>> >> >> >> > >
>> >> >> >> > > You program that MAC SA into the ATU with a destination port mask of all
>> >> >> >> > > zeroes. This suppresses further ATU miss interrupts for this MAC SA, but
>> >> >> >> > > now generates ATU member violations, because the MAC SA _is_ present in
>> >> >> >> > > the ATU, but not towards the expected port (in fact, towards _no_ port).
>> >> >> >> > >
>> >> >> >> > > Especially if user space decides it doesn't want to authorize this MAC
>> >> >> >> > > SA, it really becomes a problem because this is now a vector for denial
>> >> >> >> > > of service, with every packet triggering an ATU member violation
>> >> >> >> > > interrupt.
>> >> >> >> > >
>> >> >> >> > > So your suggestion is to set the IgnoreWrongData bit on locked ports,
>> >> >> >> > > and this will suppress the actual member violation interrupts for
>> >> >> >> > > traffic coming from these ports.
>> >> >> >> > >
>> >> >> >> > > So if the user decides to unplug a previously authorized printer from
>> >> >> >> > > switch port 1 and move it to port 2, how is this handled? If there isn't
>> >> >> >> > > a mechanism in place to delete the locked FDB entry when the printer
>> >> >> >> > > goes away, then by setting IgnoreWrongData you're effectively also
>> >> >> >> > > suppressing migration notifications.
>> >> >> >> >
>> >> >> >> > I don't think such a scenario is so realistic, as changing port is not
>> >> >> >> > just something done casually, besides port 2 then must also be a locked
>> >> >> >> > port to have the same policy.
>> >> >> >>
>> >> >> >> I think it is very realistic. It is also something which does not work
>> >> >> >> is going to cause a lot of confusion. People will blame the printer,
>> >> >> >> when in fact they should be blaming the switch. They will be rebooting
>> >> >> >> the printer, when in fact, they need to reboot the switch etc.
>> >> >> >>
>> >> >> >> I expect there is a way to cleanly support this, you just need to
>> >> >> >> figure it out.
>> >> >> >
>> >> >> > Hans, why must port 2 also be a locked port? The FDB entry with no
>> >> >> > destinations is present in the ATU, and static, why would just locked
>> >> >> > ports match it?
>> >> >> >
>> >> >> You are right of course, but it was more from a policy standpoint as I
>> >> >> pointed out. If the FDB entry is removed after some timeout and the
>> >> >> device in the meantime somehow is on another port that is not locked
>> >> >> with full access, the device will of course get full access.
>> >> >> But since it was not given access in the first instance, the policy is
>> >> >> not consistent.
>> >> >>
>> >> >> >> > The other aspect is that the user space daemon that authorizes catches
>> >> >> >> > the fdb add entry events and checks if it is a locked entry. So it will
>> >> >> >> > be up to said daemon to decide the policy, like remove the fdb entry
>> >> >> >> > after a timeout.
>> >> >> >
>> >> >> > When you say 'timeout', what is the moment when the timer starts counting?
>> >> >> > The last reception of the user space daemon of a packet with this MAC SA,
>> >> >> > or the moment when the FDB entry originally became unlocked?
>> >> >>
>> >> >> I think that if the device is not given access, a timer should be
>> >> >> started at that moment. No further FDB add events with the same MAC
>> >> >> address will come of course until the FDB entry is removed, which I
>> >> >> think would be done based on the said timer.
>> >> >> >
>> >> >> > I expect that once a device is authorized, and forwarding towards the
>> >> >> > devices that it wants to talk to is handled in hardware, that the CPU no
>> >> >> > longer receives packets from this device. In other words, are you saying
>> >> >> > that you're going to break networking for the printer every 5 minutes,
>> >> >> > as a keepalive measure?
>> >> >>
>> >> >> No, I don't think that would be a good idea, but as we are in userspace,
>> >> >> that is a policy decision of those creating the daemon. The kernel just
>> >> >> facilitates, it does not make those decisions as far as I think.
>> >> >> >
>> >> >> > I still think there should be a functional fast path for authorized
>> >> >> > station migrations.
>> >> >> >
>> >> >> I am not sure in what way you are suggesting that should be, if the
>> >> >> kernel should actively do something there? If a station is authorized,
>> >> >> and somehow is transferred to another port, if that port is not locked it
>> >> >> will get access, if the port is locked a miss violation will occur etc...
>> >> >
>> >> > Wait, if the new port is locked and the device was previously
>> >> > authorized, why will the new port trigger a miss violation? This is the
>> >> > part I'm not following. The authorization is still present in the form
>> >> > of an ATU entry on the old locked port, is it not?
>> >> >
>> >> I am sure (have not tested) that a miss violation will occur. It might
>> >> be a member violation in this instance though.
>> >> When thinking of it, afaik there is no way today of having fine control
>> >> over the DPV when adding a FDB entry.
>> >> If the DPV could be finer controlled the entry could cover several
>> >> possible ports and the fast (immediate migration) will be accomplished?
>> >
>> > I'm not sure I understand this, either.
>> >
>> > You're saying we should configure the authorizations as de-facto
>> > multicast ATU entries towards all locked ports, so that there wouldn't
>> > be any violation when a station migrates, because the new port is still
>> > in the destination port mask of the ATU entry?
>> >
>> > Yes, but... this leaks traffic between ports to a significant degree.
>> > Any packet that targets your printer now targets your colleague's printer too.
>> >
>> > I was expecting you'd say that when the cable is unplugged from the
>> > switch, the authorization daemon is notified through rtnetlink of the
>> > link state change, and it flushes the port of addresses it has added
>> > (because the kernel surely does not do this).
>> >
>> So, my HW tests show that when the link is removed, the FDB entries
>> related to the port are flushed automatically.
>
> Don't get me wrong, some addresses are flushed, this is handled by
> dsa_port_fast_age() as a result of the fact that the bridge port
> transitions to the DISABLED state when the link is lost. But those
> should only be the dynamically learned addresses. I'm not sure what
> flags your user space daemon uses to program ATU entries to hardware,
> but if those entries are anything other than dynamic, I don't think that
> dsa_port_fast_age() should flush them.

In the offloaded case there is no difference between static and dynamic
flags, which I see as a general issue. (The resulting ATU entry is static
in either case.)

These FDB entries are removed when link goes down (soft or hard). The
zero DPV entries that the new code introduces age out after 5 minutes,
while the locked flagged FDB entries are removed by link down (thus the
FDB and the ATU are not in sync in this case).

I need to get hold of a couple hubs to test the migration issue when no
link goes down, which I can do in the weekend, and test on Monday. :-)

>
> root@debian:~# ip link add br0 type bridge
> root@debian:~# ip link set swp3 master br0
> [46343.691692] br0: port 1(swp3) entered blocking state
> [46343.696791] br0: port 1(swp3) entered disabled state
> [46343.703962] device swp3 entered promiscuous mode
> root@debian:~# bridge fdb add dev swp3 00:01:02:03:04:05 master static
> root@debian:~# bridge fdb show dev swp3
> 00:01:02:03:04:05 vlan 1 offload master br0 static
> 00:01:02:03:04:05 offload master br0 static
> 00:04:9f:05:f6:28 vlan 1 master br0 permanent
> 00:04:9f:05:f6:28 master br0 permanent
> 00:01:02:03:04:05 vlan 1 self
> 00:01:02:03:04:05 self
> root@debian:~# ip link set swp3 up
> [46391.626362] fsl_enetc 0000:00:00.2 eno2: configuring for fixed/internal link mode
> [46391.634110] fsl_enetc 0000:00:00.2 eno2: Link is Up - 2.5Gbps/Full - flow control rx/tx
> [46391.635256] mscc_felix 0000:00:00.5 swp3: configuring for inband/qsgmii link mode
> root@debian:~# [46395.751268] mscc_felix 0000:00:00.5 swp3: Link is Up - 1Gbps/Full - flow control rx/tx
> [46395.759613] IPv6: ADDRCONF(NETDEV_CHANGE): swp3: link becomes ready
> root@debian:~# bridge fdb show dev swp3
> 00:01:02:03:04:05 vlan 1 offload master br0 static
> 00:01:02:03:04:05 offload master br0 static
> 00:04:9f:05:f6:28 vlan 1 master br0 permanent
> 00:04:9f:05:f6:28 master br0 permanent
> 00:01:02:03:04:05 vlan 1 self
> 00:01:02:03:04:05 self
> root@debian:~# ip link set swp3 down
> [46403.978832] mscc_felix 0000:00:00.5 swp3: Link is Down
> root@debian:~# bridge fdb show dev swp3
> 00:01:02:03:04:05 vlan 1 offload master br0 static
> 00:01:02:03:04:05 offload master br0 static
> 00:04:9f:05:f6:28 vlan 1 master br0 permanent
> 00:04:9f:05:f6:28 master br0 permanent
> 00:01:02:03:04:05 vlan 1 self
> 00:01:02:03:04:05 self
> root@debian:~# ip link set swp3 up
> [46410.499445] mscc_felix 0000:00:00.5 swp3: configuring for inband/qsgmii link mode
> root@debian:~# [46414.597381] mscc_felix 0000:00:00.5 swp3: Link is Up - 1Gbps/Full - flow control rx/tx
> [46414.605775] IPv6: ADDRCONF(NETDEV_CHANGE): swp3: link becomes ready
> root@debian:~# bridge fdb show dev swp3
> 00:01:02:03:04:05 vlan 1 offload master br0 static
> 00:01:02:03:04:05 offload master br0 static
> 00:04:9f:05:f6:28 vlan 1 master br0 permanent
> 00:04:9f:05:f6:28 master br0 permanent
> 00:01:02:03:04:05 vlan 1 self
> 00:01:02:03:04:05 self
>
> I've searched for the call paths of br_fdb_delete_by_port().
> One caller is br_stp_disable_port(), but this passes 0 to "do_all", so
> this can't be it. Also, both IFLA_BRPORT_FLUSH and the "flush" sysfs
> pass 0 to "do_all".
>
> If it's not the user space daemon who is deleting these entries, this is strange.
> More details on my suspicion below.
>
>> > This could work to an extent, but it wouldn't handle the case where the
>> > printer isn't connected directly to the 802.1X port, but through
>> > another dumb switch. I don't know enough about 802.1X, but I don't see
>> > why this isn't a valid configuration.
>> >
>> ATM, the dynamic flag (bridge fdb add MAC dev DEV master dynamic)
>> doesn't create an ageing FDB entry in the offloaded case. Maybe if that
>> was solved, it would be a good enough solution, as for a noisy device,
>> it would lose some packets every 5 minutes, which higher layers should
>> be able to handle?
>
> Before we explore this too deeply, can you first double-check whether
> station migrations will trigger a miss or a member violation?
>
> When I look at the documentation, I see that the switch triggers a member
> violation, unless the ATU age interrupt is enabled and the entry's state
> is less than 4 (which it should be if RefreshLocked is false). In this
> latter case, the documentation says that a miss violation will be
> triggered, not a membership one. So if you satisfy these requirements,
> you should be able to safely set IgnoreWrongData, and not miss out on
> migrations.
>
> My suspicion is that, since you add ATU authorizations on locked ports
> with RefreshLocked as false, they will eventually age out. But the
> driver doesn't enable the ATU age interrupt through Global 2 register 5,
> so software never finds out when these ATU entries expire. And the idea
> of triggering an ATU age interrupt when the entry state becomes less
> than 4 (i.e. they are aged half-way) is precisely done such that you
> don't break networking for the printer every 5 minutes. The switch
> notifies you in advance of the complete entry expiration. I think you
> should service this interrupt in the driver. Note that my belief is that
> such ATU entries on locked ports aren't truly 'dynamic' as far as user
> space is concerned. Sure, they age, but this is just a way for software
> to keep them in check. But user space doesn't have to know this. To the
> application, the entry is static, and the driver just refreshes the
> entry unless told by the application to delete it.
>
> Basically what I'm saying is that more testing needs to be done to make
> sure that the hardware is configured to do something reasonable in all
> cases, and doesn't just work by coincidence.
>
> Makes sense?
>
>> > To explain what I'm thinking of. At office, IT gave one Ethernet port to
>> > each desk, but I have multiple devices. I have a PC, a printer, and a
>> > development board, each with a single Ethernet port, so I use a dumb
>> > 4-port switch to connect all these devices to the 802.1X port beneath my
>> > desk. I talked to IT, brought my printer to them, they agreed to bypass
>> > 802.1X authorization for it based on the MAC address on its label.
>> >
>> > I've been working from home for the past few years, but now I need to
>> > return to office. But since years have passed, some colleagues left,
>> > some new colleagues came, and I need to change my desk. The new one
>> > belonged to a co-worker who also had a dumb switch on his desk, so I see
>> > no reason to move mine too. I unplug the printer from my dumb switch,
>> > plug it into the new one, but it doesn't work. What do I do, open a
>> > ticket to IT asking for halp?
>> >
>> > To be honest this is purely fictional and I haven't tried it, but it
>> > sounds like I should when I get the chance, to get a better image of how
>> > things are supposed to work.
>> >
>> >> >> >> > > Oh, btw, my question was: could you consider suppressing the _prints_ on
>> >> >> >> > > an ATU miss violation on a locked port?
>> >> >> >> >
>> >> >> >> > As there will only be such on the first packet, I think it should be
>> >> >> >> > logged and those prints serve that purpose, so I think it is best to
>> >> >> >> > keep the print.
>> >> >> >> > If in the future some tests or other can argue for suppressing the
>> >> >> >> > prints, it is an easy thing to do.
>> >> >> >>
>> >> >> >> Please use a traffic generator and try to DOS one of your own
>> >> >> >> switches. Can you?
>> >> >> >>
>> >> >> >> 	  Andrew