Re: [PATCH] define vmemmap before pfn_to_page calls

[David Abdurachmanov <david.abdurachmanov@gmail.com>]

On Wed, Dec 18, 2019 at 5:46 AM Anup Patel <anup@brainfault.org> wrote:
>
> On Tue, Dec 17, 2019 at 6:45 PM David Abdurachmanov
> <david.abdurachmanov@gmail.com> wrote:
> >
> > pfn_to_page call depends on `vmemmap` being available before the call.
> > This caused compilation errors in Fedora/RISCV with 5.5-rc2 and was caused
> > by NOMMU changes which moved declarations after functions definitions.
> >
> > Signed-off-by: David Abdurachmanov <david.abdurachmanov@sifive.com>
> > Fixes: 6bd33e1ece52 ("riscv: add nommu support")
> > ---
> >  arch/riscv/include/asm/pgtable.h | 34 ++++++++++++++++----------------
> >  1 file changed, 17 insertions(+), 17 deletions(-)
> >
> > diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
> > index 7ff0ed4f292e..d8c89e6e6b3d 100644
> > --- a/arch/riscv/include/asm/pgtable.h
> > +++ b/arch/riscv/include/asm/pgtable.h
> > @@ -90,6 +90,23 @@ extern pgd_t swapper_pg_dir[];
> >  #define __S110 PAGE_SHARED_EXEC
> >  #define __S111 PAGE_SHARED_EXEC
> >
> > +#define VMALLOC_SIZE     (KERN_VIRT_SIZE >> 1)
> > +#define VMALLOC_END      (PAGE_OFFSET - 1)
> > +#define VMALLOC_START    (PAGE_OFFSET - VMALLOC_SIZE)
> > +
> > +/*
> > + * Roughly size the vmemmap space to be large enough to fit enough
> > + * struct pages to map half the virtual address space. Then
> > + * position vmemmap directly below the VMALLOC region.
> > + */
> > +#define VMEMMAP_SHIFT \
> > +       (CONFIG_VA_BITS - PAGE_SHIFT - 1 + STRUCT_PAGE_MAX_SHIFT)
> > +#define VMEMMAP_SIZE   BIT(VMEMMAP_SHIFT)
> > +#define VMEMMAP_END    (VMALLOC_START - 1)
> > +#define VMEMMAP_START  (VMALLOC_START - VMEMMAP_SIZE)
> > +
> > +#define vmemmap                ((struct page *)VMEMMAP_START)
> > +
> >  static inline int pmd_present(pmd_t pmd)
> >  {
> >         return (pmd_val(pmd) & (_PAGE_PRESENT | _PAGE_PROT_NONE));
> > @@ -400,23 +417,6 @@ static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
> >  #define __pte_to_swp_entry(pte)        ((swp_entry_t) { pte_val(pte) })
> >  #define __swp_entry_to_pte(x)  ((pte_t) { (x).val })
> >
> > -#define VMALLOC_SIZE     (KERN_VIRT_SIZE >> 1)
> > -#define VMALLOC_END      (PAGE_OFFSET - 1)
> > -#define VMALLOC_START    (PAGE_OFFSET - VMALLOC_SIZE)
> > -
> > -/*
> > - * Roughly size the vmemmap space to be large enough to fit enough
> > - * struct pages to map half the virtual address space. Then
> > - * position vmemmap directly below the VMALLOC region.
> > - */
> > -#define VMEMMAP_SHIFT \
> > -       (CONFIG_VA_BITS - PAGE_SHIFT - 1 + STRUCT_PAGE_MAX_SHIFT)
> > -#define VMEMMAP_SIZE   BIT(VMEMMAP_SHIFT)
> > -#define VMEMMAP_END    (VMALLOC_START - 1)
> > -#define VMEMMAP_START  (VMALLOC_START - VMEMMAP_SIZE)
> > -
> > -#define vmemmap                ((struct page *)VMEMMAP_START)
> > -
> >  #define PCI_IO_SIZE      SZ_16M
> >  #define PCI_IO_END       VMEMMAP_START
> >  #define PCI_IO_START     (PCI_IO_END - PCI_IO_SIZE)
> > --
> > 2.23.0
> >
>
> Can you add a comment for "#define vmemmap" about your findings ?

I send v2 in a few hours with extra comment. I will mention that this
is needed if CONFIG_SPARSEMEM_VMEMMAP=y
See https://github.com/torvalds/linux/blob/master/include/asm-generic/memory_model.h

>
> Otherwise looks good to me.
>
> Reviewed-by: Anup Patel <anup@brainfault.org>
>
> Regards,
> Anup