Re: [PATCH v2 net-next 11/15] net: ethernet: mtk_eth_soc: introduce device register map

[Lorenzo Bianconi <lorenzo@kernel.org>]

[-- Attachment #1: Type: text/plain, Size: 1451 bytes --]

> On Wed, 18 May 2022 11:48:43 +0200 Lorenzo Bianconi wrote:
> > > On Mon, 16 May 2022 18:06:38 +0200 Lorenzo Bianconi wrote:  
> > > >  /* PDMA RX Base Pointer Register */
> > > > -#define MTK_PRX_BASE_PTR0	0x900
> > > > +#define MTK_PRX_BASE_PTR0	(eth->soc->reg_map[MTK_PDMA_BASE] + 0x100)
> > > >  #define MTK_PRX_BASE_PTR_CFG(x)	(MTK_PRX_BASE_PTR0 + (x * 0x10))  
> > > 
> > > Implicit macro arguments are really unpleasant for people doing
> > > tree-wide changes or otherwise unfamiliar with the driver.
> > > 
> > > Nothing we can do to avoid this?  
> > 
> > I used this approach in order to have just few changes in the codebase. I guess the best
> > option would be to explicitly add eth parameter to the register macros, what do you think?
> 
> I don't think there's a best known practice, you'll have to exercise
> your judgment. Taking a look at a random example of MTK_PDMA_INT_STATUS.
> Looks like that one is already assigned to eth->tx_int_status_reg.
> Maybe that can be generalized? Personally I'd forgo the macros
> completely and just use eth->soc->register_name in the code.

I personally think the code is easier to read if we use macros in this case.
Let's consider MTK_LRO_CTRL_DW1_CFG(), it depends on the particular soc based
on the register map and even on the ring index. I guess the best trade-off we
can get is to explicitly pass eth to the macros as parameter when needed.

Regards,
Lorenzo

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]