[PATCH net-next] net: dsa: mt7530: support MDB and bridge flag operations

Vladimir Oltean olteanv at gmail.com
Mon Mar 15 21:15:41 GMT 2021


On Mon, Mar 15, 2021 at 01:44:02PM -0700, Florian Fainelli wrote:
> On 3/15/2021 1:09 PM, Vladimir Oltean wrote:
> > On Mon, Mar 15, 2021 at 01:03:10PM -0700, Florian Fainelli wrote:
> >>
> >>
> >> On 3/15/2021 10:09 AM, DENG Qingfang wrote:
> >>> Support port MDB and bridge flag operations.
> >>>
> >>> As the hardware can manage multicast forwarding itself, offload_fwd_mark
> >>> can be unconditionally set to true.
> >>>
> >>> Signed-off-by: DENG Qingfang <dqfext at gmail.com>
> >>> ---
> >>> Changes since RFC:
> >>>   Replaced BR_AUTO_MASK with BR_FLOOD | BR_LEARNING
> >>>
> >>>  drivers/net/dsa/mt7530.c | 124 +++++++++++++++++++++++++++++++++++++--
> >>>  drivers/net/dsa/mt7530.h |   1 +
> >>>  net/dsa/tag_mtk.c        |  14 +----
> >>>  3 files changed, 122 insertions(+), 17 deletions(-)
> >>>
> >>> diff --git a/drivers/net/dsa/mt7530.c b/drivers/net/dsa/mt7530.c
> >>> index 2342d4528b4c..f765984330c9 100644
> >>> --- a/drivers/net/dsa/mt7530.c
> >>> +++ b/drivers/net/dsa/mt7530.c
> >>> @@ -1000,8 +1000,9 @@ mt753x_cpu_port_enable(struct dsa_switch *ds, int port)
> >>>  	mt7530_write(priv, MT7530_PVC_P(port),
> >>>  		     PORT_SPEC_TAG);
> >>>  
> >>> -	/* Unknown multicast frame forwarding to the cpu port */
> >>> -	mt7530_rmw(priv, MT7530_MFC, UNM_FFP_MASK, UNM_FFP(BIT(port)));
> >>> +	/* Disable flooding by default */
> >>> +	mt7530_rmw(priv, MT7530_MFC, BC_FFP_MASK | UNM_FFP_MASK | UNU_FFP_MASK,
> >>> +		   BC_FFP(BIT(port)) | UNM_FFP(BIT(port)) | UNU_FFP(BIT(port)));
> >>
> >> It's not clear to me why this is appropriate especially when the ports
> >> operated in standalone mode, can you expand a bit more on this?
> > 
> > We are in the function called "mt753x_cpu_port_enable" here. It's ok to
> > apply this config for the CPU port.
> 
> Because the user ports will flood unknown traffic and we have mediatek
> tags enabled presumably, so all traffic is copied to the CPU port, OK.

Actually this is just how Qingfang explained it:
https://patchwork.kernel.org/project/netdevbpf/patch/20210224081018.24719-1-dqfext@gmail.com/

I just assume that MT7530/7531 switches don't need to enable flooding on
user ports when the only possible traffic source is the CPU port - the
CPU port can inject traffic into any port regardless of egress flooding
setting. If that's not true, I don't see how traffic in standalone ports
mode would work after this patch.



More information about the Linux-mediatek mailing list