On Mon, Dec 14, 2015 at 09:53:05PM +0100, Linus Walleij wrote: > According to commit 2503a5ecd86c002506001eba432c524ea009fe7f > "ARM: 6201/1: RealView: Do not use outer_sync() on ARM11MPCore > boards with L220" Some PB11MPCore RealView core tiles have broken > outer_sync. > > We got rid of the custom barriers from the machine by disabling > outer sync, but that was just for the boardfile case. We have > to be able to do the same in the device tree case. > > Since __l2c_init() is cloning and copying the L2C vtable, > we pass an argument to this function to optionally numb > the outer sync operation if desired, before initializing > the cache. > > After this we can set up the cache correctly on the RealView > PB11MPCore. This was tested on a PB11MPCore known to have the > issue. Before this, spurious crashes would occur if we try to > set up the cache properly, after this it boots rock solid. > > Cc: Russell King <linux@xxxxxxxxxxxxxxxx> > Cc: Arnd Bergmann <arnd@xxxxxxxx> > Cc: devicetree@xxxxxxxxxxxxxxx > Signed-off-by: Linus Walleij <linus.walleij@xxxxxxxxxx> > --- > ChangeLog v2->v3: > - Update description, reference Catalins initial commit in the > commit blurb. > --- > Documentation/devicetree/bindings/arm/l2cc.txt | 3 +++ > arch/arm/mm/cache-l2x0.c | 13 ++++++++++--- > 2 files changed, 13 insertions(+), 3 deletions(-) > > diff --git a/Documentation/devicetree/bindings/arm/l2cc.txt b/Documentation/devicetree/bindings/arm/l2cc.txt > index d181b7c4c522..416864e9dc92 100644 > --- a/Documentation/devicetree/bindings/arm/l2cc.txt > +++ b/Documentation/devicetree/bindings/arm/l2cc.txt > @@ -75,6 +75,9 @@ Optional properties: > specified to indicate that such transforms are precluded. > - arm,parity-enable : enable parity checking on the L2 cache (L220 or PL310). > - arm,parity-disable : disable parity checking on the L2 cache (L220 or PL310). > +- arm,outer-sync-disable : disable the outer sync operation on the L2 cache. > + Some core tiles, especially ARM PB11MPCore have a faulty L220 cache that > + will randomly hang unless outer sync operations are disabled. This is also useful on Highbank as a feature where all DMA goes thru the ACP port and there should not be a need to flush the L2 write buffer. But I already optimized the performance critical register accesses with _relaxed variants, so it doesn't really matter. I don't think any other platform cares, therefore: Acked-by: Rob Herring <robh@xxxxxxxxxx> Rob -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html