[PATCH 3/3] mmc: sdhci-of-esdhc: add workaround for T4240 incorrect HOSTVER value

[Scott Wood]

On Wed, 2015-08-26 at 02:49 -0500, Lu Yangbo-B47093 wrote:
> > > > > > For T4240-R1.0-R2.0, the HOSTVER register has incorrcet vender
> > > > > > version value and sdhc spec version value. This will break down
> > > > > > the ADMA data transfer. So add workaround to get right value
> > > > > > VVN=0x13, SVN = 0x1.
> > > > > 
> > > > > So T4240-R1.0-R2.0 is the version of the controller, right?
> > > > > 
> > > > > If I understand correct you are checking what CPU/SoC you are
> > > > > running on, to figure out which controller version you are using,
> > > > > as that can't be fetched (trusted) from the registers of the esdhc
> > > > > controller itself!?
> > > > > 
> > > > > Instead, you could deal with this directly in the DTS files. I
> > > > > assume you have some DTS file for each SoC/board variant, right?
> > > > 
> > > > No, we do not have a separate DTS file for each revision of an SoC
> > > > -- and if we did, we'd constantly have people using the wrong one.
> > > > 
> > > > > In principle, in your DTS file specific for the board/SoC that
> > > > > holds the T4240-R1.0-R2.0 version of the controller, should add a
> > > > > specific esdhc DT property to indicate this errata.
> > > > 
> > > > No, because (in addition to the above issue about chip revisions)
> > > > the device tree is stable ABI and errata are often discovered after
> > > > device trees are deployed.
> > > 
> > > Fair enough. Then what is your suggestion for the solution here?
> > 
> > As I said in my comment on patch 2/3, read SVR from the device-
> > config/guts MMIO block, which works on both PPC and ARM.
> > 
> > -Scott
> 
> Thanks, Scott.
> I checked the device nodes of device-config/guts, finding all the platforms 
> has not a uniform compatible name.
> Could I add a compatible name called "fsl, dcfg"("fsl, guts" if using guts) 
> for PowerPC P series, T series, B series boards, so that I could get this 
> device node by compatible without knowing platform? 

No.  It's too vague.  Linux should keep a list of all the nodes it can match 
against for this.  Centralize it in one place rather than having each driver 
that cares about SVR do it.

-Scott