Date:	Wed, 8 Jul 2015 19:32:16 +0200
From:	Andrew Lunn <andrew@...n.ch>
To:	Vivien Didelot <vivien.didelot@...oirfairelinux.com>
Cc:	netdev <netdev@...r.kernel.org>, David <davem@...emloft.net>,
	Scott Feldman <sfeldma@...il.com>,
	Jiri Pirko <jiri@...nulli.us>,
	Florian Fainelli <f.fainelli@...il.com>,
	Guenter Roeck <linux@...ck-us.net>,
	linux-kernel <linux-kernel@...r.kernel.org>,
	kernel <kernel@...oirfairelinux.com>
Subject: Re: [PATCH v4 0/3] net: dsa: mv88e6xxx: add support for VLAN Table
 Unit

Vivien Didelot wrote:
> Hi Andrew,
> 
> On Jul 8, 2015, at 10:38 AM, Andrew Lunn andrew@...n.ch wrote:
> 
> > On Tue, Jul 07, 2015 at 05:18:17PM -0400, Vivien Didelot wrote:
> >> Hi all,
> >> 
> >> This patchset brings full support for hardware VLANs in DSA, and the Marvell
> >> 88E6xxx compatible switch chips.
> > 
> > Hi Vivien
> > 
> > I would like to do a proper review and testing of these patchset, but
> > i go on vacation this afternoon. So it will be in about 2 weeks time.
> > 
> > I spent 15 minutes tests just now. I spotted two things:
> > 
> > 1) I played with a configuration, and then rebooted the machine. After
> > login i see:
> > 
> > Debian GNU/Linux comes with ABSOLUTELY NO WARRANTY, to the extent
> > permitted by applicable law.
> > # cat /sys/kernel/debug/dsa0/vtu
> > VID  FID  SID  0  1  2  3  4  5  6
> >   1    1    0  u  u  u  u  x  x  t
> > 500  500    0  t  t  t  t  x  x  t
> > 550  550    0  t  x  x  x  x  x  t
> > # bridge vlan show
> > port    vlan ids
> > lan0     1 PVID Egress Untagged
> > 
> > lan0     1 PVID Egress Untagged
> > 
> > lan1
> > lan2
> > lan3
> > lan4
> > lan5
> > lan6
> > lan7
> > lan8     1 PVID Egress Untagged
> > 
> > lan8     1 PVID Egress Untagged
> > 
> > optical3
> > optical4
> > br0      1 PVID Egress Untagged
> > 
> > 
> > So the switch seems to have some VTU table entries, but the bridge
> > command does not show them. I suspect that a warm boot does not clear
> > out the VTU entries in the switch.
> > 
> > Until recently we had a similar problem with the statistics
> > counters. I wounder if we have the same problem with other tables? Do
> > static ATU entries get removed on a reboot?
> > 
> 
> You're right. There's a single operation to clear the STU and VTU. I
> will send a follow-up patch to send this command during the switch
> setup.
> 
> > 2) I cold booted the machine, to be sure to have a clean state. Then:
> > 
> > # cat /sys/kernel/debug/dsa0/vtu
> > VID  FID  SID  0  1  2  3  4  5  6
> >   1    1    0  u  x  x  x  x  x  t
> > 
> > So a good initial state. I then configure two bridges:
> > 
> > # brctl show
> > bridge name     bridge id               STP enabled     interfaces
> > br0             8000.92647a2160c4       yes             lan0
> >                                                        lan1
> > br1             8000.92647a2160c4       yes             lan2
> >                                                        lan3
> > 
> > and then add vlan 500 to the four interfaces.
> > 
> > # bridge vlan add vid 500 dev lan0 master
> > # bridge vlan add vid 500 dev lan1 master
> > # bridge vlan add vid 500 dev lan2 master
> > # bridge vlan add vid 500 dev lan3 master
> > 
> > # cat /sys/kernel/debug/dsa0/vtu
> > VID  FID  SID  0  1  2  3  4  5  6
> >   1    1    0  u  u  u  u  x  x  t
> > 500  500    0  t  t  t  t  x  x  t
> > 
> > Does this mean we have one hardware bridge? All four ports can talk to
> > each other? I've not actually sent any frames to test this, so i'm
> > just speculating. Given that i have two software bridges, this is not
> > what i would expect, if frames from lan0 or lan1, also went out lan2
> > or lan3.
> 
> Indeed, with the "master" keyword, we ask switchdev to populate the
> parent's (i.e. the switch chip) to create VLANs. Marvell switch such as
> the 88E66352 can only have a single VLAN table entry for a given VID.

Hi Vivien

We are using the switch to perform hardware acceleration of things
that Linux does already in software. We have to keep with the
semantics of what is already supported in software. The patch in its
current state breaks the accepted behaviour.

This is a limitation of the switch. So the driver needs to keep track
of which bridge a VLAN belongs to, if it is asked to accelerate the
same VLAN for a different bridge, it needs to say to the kernel,
sorry, cannot do that, and leave the kernel to do it in software.

       Andrew
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Powered by blists - more mailing lists