[Skiboot] [PATCH 2/2] phb3/phb4/p7ioc: Document supported TCE sizes in DT

[Russell Currey]

On Mon, 2018-02-05 at 16:02 +1100, Alexey Kardashevskiy wrote:
> On 05/02/18 15:00, Russell Currey wrote:
> > Add a new property, "ibm,supported-tce-sizes", to advertise to
> > Linux how
> > big the available TCE sizes are.  Each value is a bit shift, from
> > smallest to largest.
> 
> Why would we possibly need such a limit? We already have an example
> of DDW
> in sPAPR where possible IOMMU page sizes are fixed and there is no
> way to
> have 2M IOMMU pages in the guest, now the same sort of limit is being
> added
> to skiboot...

There are no limits being imposed, this is purely documenting what the
hardware supports.
> 
> 
> > 
> > Signed-off-by: Russell Currey <ruscur at russell.cc>
> > ---
> >  hw/p7ioc-phb.c | 5 +++++
> >  hw/phb3.c      | 5 +++++
> >  hw/phb4.c      | 5 +++++
> >  3 files changed, 15 insertions(+)
> > 
> > diff --git a/hw/p7ioc-phb.c b/hw/p7ioc-phb.c
> > index 6c074518..b0a26e84 100644
> > --- a/hw/p7ioc-phb.c
> > +++ b/hw/p7ioc-phb.c
> > @@ -2596,6 +2596,11 @@ static void p7ioc_pcie_add_node(struct
> > p7ioc_phb *p)
> >  	tkill = reg[0] + PHB_TCE_KILL;
> >  	dt_add_property_cells(np, "ibm,opal-tce-kill",
> >  			      hi32(tkill), lo32(tkill));
> > +	dt_add_property_cells(np, "ibm,supported-tce-sizes",
> > +			      12, // 4K
> > +			      16, // 64K
> > +			      24, // 16M
> > +			      34); // 16G
> >  
> >  	/*
> >  	 * Linux may use this property to allocate the diag data
> > buffer, which
> > diff --git a/hw/phb3.c b/hw/phb3.c
> > index 0b4c3c32..0c4e1eb3 100644
> > --- a/hw/phb3.c
> > +++ b/hw/phb3.c
> > @@ -4469,6 +4469,11 @@ static void phb3_add_properties(struct phb3
> > *p)
> >  	tkill = reg + PHB_TCE_KILL;
> >  	dt_add_property_cells(np, "ibm,opal-tce-kill",
> >  			      hi32(tkill), lo32(tkill));
> > +	dt_add_property_cells(np, "ibm,supported-tce-sizes",
> > +			      12, // 4K
> > +			      16, // 64K
> > +			      24, // 16M
> > +			      28); // 256M
> >  
> >  	/*
> >  	 * Indicate to Linux that the architected IODA2 MSI EOI
> > method
> > diff --git a/hw/phb4.c b/hw/phb4.c
> > index 0a7e3065..946f9ab6 100644
> > --- a/hw/phb4.c
> > +++ b/hw/phb4.c
> > @@ -4688,6 +4688,11 @@ static void phb4_add_properties(struct phb4
> > *p)
> >  	/* M64 ranges start at 1 as MBT0 is used for M32 */
> >  	dt_add_property_cells(np, "ibm,opal-available-m64-ranges",
> >  			      1, p->mbt_size - 1);
> > +	dt_add_property_cells(np, "ibm,supported-tce-sizes",
> > +			      12, // 4K
> > +			      16, // 64K
> > +			      21, // 2M
> > +			      30); // 1G
> >  
> >  	/* Tell Linux about alignment limits for segment splits.
> >  	 *
> > 
> 
>