[PATCH v3] ARM: mvebu: rename ARM_ERRATA_753970

[pebolle@tiscali.nl (Paul Bolle)]

Hi Russell,

On Fri, 2014-09-12 at 14:31 +0100, Russell King - ARM Linux wrote:
> On Fri, Sep 12, 2014 at 01:10:05PM +0200, Paul Bolle wrote:
> > ARM_ERRATA_753970 was renamed to PL310_ERRATA_753970 in v3.2, through
> > commit fa0ce4035d48 ("ARM: 7162/1: errata: tidy up Kconfig options for
> > PL310 errata workarounds"). Two selects were added in v3.15 that still
> > use the previous name. Rename these. Make these select statements depend
> > on CACHE_L2X0, like all other select statements for PL310_ERRATA_753970
> > do, to be safe we don't inadvertently start to allow pointless
> > configurations in the future.
> > 
> > Signed-off-by: Paul Bolle <pebolle@tiscali.nl>
> > ---
> > v3: add CACHE_L2X0 dependency, as Russell requested.
> > 
> > Tested with "git grep" and fiddling with "make ARCH=arm [...]config".
> > Please shout if that's deemed insufficient.
> > 
> > Note that I promised this v3 over two months ago. Taking two months to
> > update this patch is not really something to be proud of. On the other
> > hand: no one bothered to prod me for this v3 or send in a proper fix
> > themselves. Add to this that I first reported these selects for an
> > outdated Kconfig symbol in April. So I begin to wonder whether these
> > selects statements are really needed to begin with.

Whatever will be done in the long run: no one else said anything about
my v3. It's now nearly seven months since I first reported this issue.
In the mean time three releases contained the bogus selects. Apparently
neither Marvell Armada 375 boards nor Marvell Armada 380/385 boards
actually need to select PL310_ERRATA_753970.

I think I'll send a trivial cleanup shortly.

> I wonder whether we should deal with this a different way: rather than
> having this stuff as a configuration option, have the work-arounds
> default to being enabled, and have kernel command parameters to disable
> them should that be necessary.
> 
> I expect the normal case is that everyone normally runs with these
> errata workarounds enabled (especially with a single zImage kernel), so
> there will be relatively few who need to disable them.  Plus, this gives
> flexibility on single zImage to turn the errata workarounds off without
> needing to rebuild.
> 
> The other major advantage is less Kconfig options with dependencies
> which seem all to easily to result in select abuse.

I have no idea what CACHE_L2X0, ie the L2x0 outer cache controller,
actually does nor what those PL310 errata are all about. So I can't
comment on the feasibility of your idea. Obviously, dropping Kconfig
symbols, and making configuration a bit easier would be one of the
benefits of that idea. But I wouldn't know whether it outweighs the
costs involved.

Thanks,


Paul Bolle