[PATCH net-next] net: airoha: Fix GDM2 loopback configuration for USB serdes

[Lorenzo Bianconi]

> On Sat, Feb 28, 2026 at 01:20:33PM +0100, Lorenzo Bianconi wrote:
> > Flow Control source port mapping for USB serdes needs to be configured
> > according to the GDM port nbd parameter. This is a preliminary patch
> > since nbd parameter is specific for the given port serdes and needs to
> > be read from the DTS (in the current codebase is assigned statically).
> > 
> > Signed-off-by: Lorenzo Bianconi <lorenzo at kernel.org>
> 
> Hi Lorenzo,
> 
> Is this a fix? If so, should it have a fixes tag? And be targeted at net?
> Otherwise, perhaps drop the "Fix" from the subject.

Hi Simon,

This is a preliminary patch in order to completely fix the loopback
configuration for USB serdes but the point is we are not supporting
USB serdes upstream yet (there is only OpenWrt support available at
the moment). Do you think we should target net-next or net in this
case?

> 
> Also, I think it would be worth explaining the situation
> before and after this patch.
> 
> > ---
> >  drivers/net/ethernet/airoha/airoha_eth.c  | 10 ++++++----
> >  drivers/net/ethernet/airoha/airoha_regs.h |  5 +----
> >  2 files changed, 7 insertions(+), 8 deletions(-)
> > 
> > diff --git a/drivers/net/ethernet/airoha/airoha_eth.c b/drivers/net/ethernet/airoha/airoha_eth.c
> > index 3779f93b47bc13ae21d4ae088dece4712dc214e5..37fc2b33d3af8a190a74efc69ed54f163afa6412 100644
> > --- a/drivers/net/ethernet/airoha/airoha_eth.c
> > +++ b/drivers/net/ethernet/airoha/airoha_eth.c
> > @@ -1729,10 +1729,12 @@ static int airhoha_set_gdm2_loopback(struct airoha_gdm_port *port)
> >  		      SP_CPORT_MASK(val),
> >  		      FE_PSE_PORT_CDM2 << __ffs(SP_CPORT_MASK(val)));
> >  
> > -	if (port->id != AIROHA_GDM3_IDX && airoha_is_7581(eth))
> > -		airoha_fe_rmw(eth, REG_SRC_PORT_FC_MAP6,
> > -			      FC_ID_OF_SRC_PORT24_MASK,
> > -			      FIELD_PREP(FC_ID_OF_SRC_PORT24_MASK, 2));
> > +	if (port->id == AIROHA_GDM4_IDX && airoha_is_7581(eth)) {
> > +		u32 mask = FC_ID_OF_SRC_PORT_MASK(nbq);
> 
> AI generated code review flags that the commit message talks about nbd.
> But here we have nbq.

yes, my mistake. It is supposed to be nbq and not nbd.

> 
> > +
> > +		airoha_fe_rmw(eth, REG_SRC_PORT_FC_MAP6, mask,
> > +			      AIROHA_GDM2_IDX << __ffs(mask));
> 
> Perhaps naïvely I would have used FIELD_PREP here.
> Maybe that leads to an incorrect result. But if not,
> is a shift used here for consistency with the code at
> the beginning of this hunk?

ack, I think we can FIELD_PREP() now. In the past we can't use FIELD_PREP()
with non-compile constant mask, but now it is doable.

Regards,
Lorenzo

> 
> > +	}
> >  
> >  	return 0;
> >  }
> 
> ...
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 228 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/linux-mediatek/attachments/20260304/a01b019d/attachment-0001.sig>