[PATCH] ARM: cache-l2x0.c: Apply the workaround to r3p0 only

[srinidhi.kasagar@stericsson.com (Srinidhi Kasagar)]

On Wed, Dec 19, 2012 at 12:56:41 +0100, Will Deacon wrote:
> On Wed, Dec 19, 2012 at 05:10:53AM +0000, Srinidhi Kasagar wrote:
> > Hello Will,
> > 
> > On Tue, Oct 30, 2012 at 11:02:32 +0100, srinidhi kasagar wrote:
> > > The PL310 bug 753970 should be applied only to r3p0
> > > 
> > > Signed-off-by: srinidhi kasagar <srinidhi.kasagar@stericsson.com>
> > > ---
> > >  arch/arm/mm/cache-l2x0.c |    9 +++++++--
> > >  1 files changed, 7 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/arch/arm/mm/cache-l2x0.c b/arch/arm/mm/cache-l2x0.c
> > > index 2a8e380..7925734 100644
> > > --- a/arch/arm/mm/cache-l2x0.c
> > > +++ b/arch/arm/mm/cache-l2x0.c
> > > @@ -311,6 +311,7 @@ void __init l2x0_init(void __iomem *base, u32 aux_val, u32 aux_mask)
> > >  	u32 aux;
> > >  	u32 cache_id;
> > >  	u32 way_size = 0;
> > > +	u32 l2x0_revision;
> > >  	int ways;
> > >  	const char *type;
> > >  
> > > @@ -331,8 +332,12 @@ void __init l2x0_init(void __iomem *base, u32 aux_val, u32 aux_mask)
> > >  			ways = 8;
> > >  		type = "L310";
> > >  #ifdef CONFIG_PL310_ERRATA_753970
> > > -		/* Unmapped register. */
> > > -		sync_reg_offset = L2X0_DUMMY_REG;
> > > +		 /* FIXME: make sense to have dt support for L2 Erratas? */
> 
> Don't bother adding a FIXME. A more useful comment would be describing the
> affected revisions.

Ok.

> 
> > > +		l2x0_revision = readl_relaxed(l2x0_base + L2X0_CACHE_ID) &
> > > +				L2X0_CACHE_ID_RTL_MASK;
> > > +		if (l2x0_revision == L2X0_CACHE_ID_RTL_R3P0)
> > > +			/* Unmapped register. */
> > > +			sync_reg_offset = L2X0_DUMMY_REG;
> > >  #endif
> > >  		outer_cache.set_debug = pl310_set_debug;
> > >  		break;
> > 
> > What is your thought on this? The platforms like x540 shares mostly the same code as 
> > of 8500 but Cortex and PL310 revisions are upgraded, and consequently since we manage the
> > the single defconfig for both these platforms, the compile time options for enabling
> > ARM/PL310 ERRATAs are not really working out.
> 
> Can you be more specific about `not working out' please? Are there other
> PL310 errata which would benefit from the same thing?

What i mean is, we need to apply the same method of detecting the PL310/ARM revisions
dynamically at run time rather than compile time switches.

For example, as I mentioned, another instance is PL310_ERRATA_769419 should also be handled
by detecting the PL310 revision. I will send out similar patch if you wish..

regards/srinidhi