Re: [PATCH v3 1/5] powerpc/mm: set the radix linear page mapping size

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

On Mon, 2016-12-19 at 14:28 +0530, Aneesh Kumar K.V wrote:
> Reza Arbab <arbab@linux.vnet.ibm.com> writes:
> 
> > This was defaulting to 4K, regardless of PAGE_SIZE.
> > 
> > Signed-off-by: Reza Arbab <arbab@linux.vnet.ibm.com>
> > ---
> >  arch/powerpc/mm/pgtable-radix.c | 2 ++
> >  1 file changed, 2 insertions(+)
> > 
> > diff --git a/arch/powerpc/mm/pgtable-radix.c
> > b/arch/powerpc/mm/pgtable-radix.c
> > index 623a0dc..54bd70e 100644
> > --- a/arch/powerpc/mm/pgtable-radix.c
> > +++ b/arch/powerpc/mm/pgtable-radix.c
> > @@ -351,8 +351,10 @@ void __init radix__early_init_mmu(void)
> >  #ifdef CONFIG_PPC_64K_PAGES
> >  	/* PAGE_SIZE mappings */
> >  	mmu_virtual_psize = MMU_PAGE_64K;
> > +	mmu_linear_psize = MMU_PAGE_64K;
> 
> That is not clearly correct, we map the linear address with either
> 64K,
> 2M or 1G depending on the memory available. Take a look at
> static void __init radix_init_pgtable(void)
> 

So should we fix that initialization regardless or take it out ?
> 
> >  #else
> >  	mmu_virtual_psize = MMU_PAGE_4K;
> > +	mmu_linear_psize = MMU_PAGE_4K;
> >  #endif
> > 
> >  #ifdef CONFIG_SPARSEMEM_VMEMMAP
> > -- 
> > 1.8.3.1