Re: [PATCH net-next] net: airoha: Add missing filed to ppe_mbox_data struct

[Simon Horman <horms@kernel.org>]

On Wed, Apr 16, 2025 at 05:55:42PM +0200, Lorenzo Bianconi wrote:
> > On Tue, Apr 15, 2025 at 09:27:21AM +0200, Lorenzo Bianconi wrote:
> > > The official Airoha EN7581 firmware requires adding max_packet filed in
> > > ppe_mbox_data struct while the unofficial one used to develop the Airoha
> > > EN7581 flowtable offload does not require this field. This patch fixes
> > > just a theoretical bug since the Airoha EN7581 firmware is not posted to
> > > linux-firware or other repositories (e.g. OpenWrt) yet.
> > > 
> > > Fixes: 23290c7bc190d ("net: airoha: Introduce Airoha NPU support")
> > > Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
> > > ---
> > >  drivers/net/ethernet/airoha/airoha_npu.c | 1 +
> > >  1 file changed, 1 insertion(+)
> > > 
> > > diff --git a/drivers/net/ethernet/airoha/airoha_npu.c b/drivers/net/ethernet/airoha/airoha_npu.c
> > > index 7a5710f9ccf6a4a4f555ab63d67cb6b318de9b52..16201b5ce9f27866896226c3611b4a154d19bc2c 100644
> > > --- a/drivers/net/ethernet/airoha/airoha_npu.c
> > > +++ b/drivers/net/ethernet/airoha/airoha_npu.c
> > > @@ -104,6 +104,7 @@ struct ppe_mbox_data {
> > >  			u8 xpon_hal_api;
> > >  			u8 wan_xsi;
> > >  			u8 ct_joyme4;
> > > +			u8 max_packet;
> > >  			int ppe_type;
> > >  			int wan_mode;
> > >  			int wan_sel;
> > 
> > Hi Lorenzo,
> > 
> > I'm a little confused by this.
> > 
> > As I understand it ppe_mbox_data is sent as the data of a mailbox message
> > send to the device.  But by adding the max_packet field the layout is
> > changed. The size of the structure changes. And perhaps more importantly
> > the offset of fields after max_packet, e.g.  wan_mode, change.
> > 
> > Looking at how this is used, f.e. in the following code, I'm unclear on
> > how this change is backwards compatible.
> 
> you are right Simon, this change is not backwards compatible but the fw
> is not publicly available yet and the official fw version will use this
> new layout (the previous one was just a private version I used to develop
> the driver).  Can we use this simple approach or do you think we should
> differentiate the two firmware version in some way? (even if the previous
> one will never be used).

Hi Lorenzo,

Sorry, I misunderstood the commit message.

Yes, I think this simple approach is fine if there is no
compatibility issue wrt firmwares we can reasonably expect
in the wild. Which seems to be the case here.

Reviewed-by: Simon Horman <horms@kernel.org>