[PATCH 2/2] powerpc/85xx: describe the PAMU topology in the device tree

Yoder Stuart-B08248 B08248 at freescale.com
Fri Jan 25 09:57:08 EST 2013



> -----Original Message-----
> From: Gala Kumar-B11780
> Sent: Wednesday, January 23, 2013 11:27 AM
> To: Wood Scott-B07421; Yoder Stuart-B08248
> Cc: linuxppc-dev at ozlabs.org list; Timur Tabi
> Subject: Re: [PATCH 2/2] powerpc/85xx: describe the PAMU topology in the device tree
> 
> 
> On Jan 17, 2013, at 4:34 PM, Timur Tabi wrote:
> 
> > From: Timur Tabi <timur at freescale.com>
> >
> > The PAMU caches use the LIODNs to determine which cache lines hold the
> > entries for the corresponding LIODs.  The LIODNs must therefore be
> > carefully assigned to avoid cache thrashing -- two active LIODs with
> > LIODNs that put them in the same cache line.
> >
> > Currently, LIODNs are statically assigned by U-Boot, but this has
> > limitations.  LIODNs are assigned even for devices that may be disabled
> > or unused by the kernel.  Static assignments also do not allow for device
> > drivers which may know which LIODs can be used simultaneously.  In
> > other words, we really should assign LIODNs dynamically in Linux.
> >
> > To do that, we need to describe the PAMU device and cache topologies in
> > the device trees.
> >
> > Signed-off-by: Timur Tabi <timur at freescale.com>
> > ---
> > .../devicetree/bindings/powerpc/fsl/guts.txt       |   14 ++-
> > .../devicetree/bindings/powerpc/fsl/pamu.txt       |  142 ++++++++++++++++++++
> > arch/powerpc/boot/dts/fsl/p2041si-post.dtsi        |   87 +++++++++++--
> > arch/powerpc/boot/dts/fsl/p3041si-post.dtsi        |   87 +++++++++++--
> > arch/powerpc/boot/dts/fsl/p4080si-post.dtsi        |   68 +++++++++-
> > arch/powerpc/boot/dts/fsl/p5020si-post.dtsi        |   92 +++++++++++--
> > arch/powerpc/boot/dts/fsl/p5040si-post.dtsi        |   92 +++++++++++--
> > 7 files changed, 533 insertions(+), 49 deletions(-)
> > create mode 100644 Documentation/devicetree/bindings/powerpc/fsl/pamu.txt
> 
> Scott, Stuart, does this have your guys Ack?

Ack



More information about the Linuxppc-dev mailing list