Re: [PATCH v6 10/18] sh/tlb: Convert SH to generic mmu_gather

From: Geert Uytterhoeven
Date: Thu Dec 05 2019 - 10:27:12 EST


Hoi Peter,

On Wed, Dec 4, 2019 at 5:42 PM Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> On Wed, Dec 04, 2019 at 04:07:53PM +0100, Geert Uytterhoeven wrote:
> > On Wed, Dec 4, 2019 at 2:35 PM Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> > > Does this fare better?
> >
> > Yes. Migo-R is happy again.
> > Tested-by: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>
> >
> > > --- a/arch/sh/include/asm/pgalloc.h
> > > +++ b/arch/sh/include/asm/pgalloc.h
> > > @@ -36,9 +36,7 @@ do { \
> > > #if CONFIG_PGTABLE_LEVELS > 2
> > > #define __pmd_free_tlb(tlb, pmdp, addr) \
> > > do { \
> > > - struct page *page = virt_to_page(pmdp); \
> > > - pgtable_pmd_page_dtor(page); \
> > > - tlb_remove_page((tlb), page); \
> > > + pmd_free((tlb)->mm, (pmdp)); \
> > > } while (0);
> > > #endif
>
> OK, so I was going to write a Changelog to go with that, but then I
> realized that while this works and is similar to before the patch, I'm
> not sure this is in fact correct.
>
> With this on (and also before) we're freeing the PMD before we've done
> the TLB invalidate, that seems wrong!
>
> Looking at the size of that pmd_cache, that looks to be 30-(12+12-3)+3
> == 12, which is exactly 1 page, for PAGE_SIZE_4K, less for the larger
> pages.
>
> I'm thinking perhaps we should do something like the below instead?

Your advice is better when in close vicinity of an SH cross compiler,
though ;-)

> --- a/arch/sh/mm/pgtable.c
> +++ b/arch/sh/mm/pgtable.c
> @@ -5,9 +5,6 @@
> #define PGALLOC_GFP GFP_KERNEL | __GFP_ZERO
>
> static struct kmem_cache *pgd_cachep;
> -#if PAGETABLE_LEVELS > 2
> -static struct kmem_cache *pmd_cachep;
> -#endif
>
> void pgd_ctor(void *x)
> {
> @@ -23,11 +20,6 @@ void pgtable_cache_init(void)
> pgd_cachep = kmem_cache_create("pgd_cache",
> PTRS_PER_PGD * (1<<PTE_MAGNITUDE),
> PAGE_SIZE, SLAB_PANIC, pgd_ctor);
> -#if PAGETABLE_LEVELS > 2
> - pmd_cachep = kmem_cache_create("pmd_cache",
> - PTRS_PER_PMD * (1<<PTE_MAGNITUDE),
> - PAGE_SIZE, SLAB_PANIC, NULL);
> -#endif
> }
>
> pgd_t *pgd_alloc(struct mm_struct *mm)
> @@ -48,11 +40,7 @@ void pud_populate(struct mm_struct *mm, pud_t *pud, pmd_t *pmd)
>
> pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long address)
> {
> - return kmem_cache_alloc(pmd_cachep, PGALLOC_GFP);
> -}
> -
> -void pmd_free(struct mm_struct *mm, pmd_t *pmd)

mm/memory.o: In function `__pmd_alloc':
memory.c:(.text+0x1d74): undefined reference to `pmd_free'

> -{
> - kmem_cache_free(pmd_cachep, pmd);
> + BUILD_BUG_ON(PTRS_PER_PMD * (1<<PTE_MAGNITUDE) <= PAGE_SIZE);

... > PAGE_SIZE ?

Else it triggers all the time.

> + return (pmd_t *)__get_free_page(PGALLOC_GFP);
> }
> #endif /* PAGETABLE_LEVELS > 2 */

BTW, I'm still running Willy's fix that never made it upstream
to kill an ugly boot warning, which also touches this code:
https://patchwork.kernel.org/patch/10549883/#22166333

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds