Re: [PATCH 1/3] i2c: aspeed: avoid new registers definition of AST2600

From: Joel Stanley
Date: Sun May 23 2021 - 22:35:13 EST


On Mon, 24 May 2021 at 01:53, Jamin Lin <jamin_lin@xxxxxxxxxxxxxx> wrote:
>
> The 05/21/2021 02:00, Tao Ren wrote:
> > Hi Jamin,
> >
> > On Thu, May 20, 2021 at 11:31:41AM +0800, Jamin Lin wrote:
> > > The 05/19/2021 22:59, Joel Stanley wrote:
> > > > On Wed, 19 May 2021 at 08:05, Jamin Lin <jamin_lin@xxxxxxxxxxxxxx> wrote:
> > > > >
> > > > > The register definition between AST2600 A2 and A3 is different.
> > > > > This patch avoid new registers definition of AST2600 to use
> > > > > this driver. We will submit the path for the new registers
> > > > > definition of AST2600.
> > > >
> > > > The AST2600 v9 datasheet says that bit 2 selects between old and new
> > > > register sets, and that the old register set is the default.
> > > >
> > > > Has the default changed for the A3?, and the datasheet is incorrect?
> > > >
> > > > Does the A3 still support the old register set?
> > > >
> > > We suggest user to use the new i2c driver for AST2600 and we will sumbit
> > > it. This driver is used to AST2500 and AST2400 SOCs. Change this
> > > driver to check global register of i2c to avoid user build the wrong driver.
> >
> > If I understand correctly, the answer implies old register set is still
> > supported in A3 although aspeed suggest people using the new driver/mode?
> >
> > Can you please share more context behind the suggestion? Such as new
> > register mode has better performance? Or some known issues that were
> > deteted in old mode are fixed in new register mode?
> >
> Yes, AST2600 A1, A2 and A3 support both old and new register set. The difference
> between old and new register set are the register address and supported registers.
> User can choose to use both old and new register set. However, ASPEED would like to
> change new register set by default for AST2600.

We can certainly make the driver for the new register set the default
for AST2600 when the new driver is merged.

I disagree that we should introduce a run time check to fail to probe
the old driver. Please do not merge this patch.

Please focus your effort on getting the new driver merged instead.

Cheers,

Joel