Re: [PATCH] mm/zsmalloc.c: Fix zsmalloc 32-bit PAE support

[Kirill A. Shutemov]

On Mon, Dec 10, 2018 at 12:21:05PM -0200, Rafael David Tinoco wrote:
> diff --git a/arch/x86/include/asm/pgtable_64_types.h b/arch/x86/include/asm/pgtable_64_types.h
> index 84bd9bdc1987..d808cfde3d19 100644
> --- a/arch/x86/include/asm/pgtable_64_types.h
> +++ b/arch/x86/include/asm/pgtable_64_types.h
> @@ -64,8 +64,6 @@ extern unsigned int ptrs_per_p4d;
>  #define P4D_SIZE		(_AC(1, UL) << P4D_SHIFT)
>  #define P4D_MASK		(~(P4D_SIZE - 1))
>  
> -#define MAX_POSSIBLE_PHYSMEM_BITS	52
> -
>  #else /* CONFIG_X86_5LEVEL */
>  
>  /*
> @@ -154,4 +152,6 @@ extern unsigned int ptrs_per_p4d;
>  
>  #define PGD_KERNEL_START	((PAGE_SIZE / 2) / sizeof(pgd_t))
>  
> +#define MAX_POSSIBLE_PHYSMEM_BITS (pgtable_l5_enabled() ? 52 : 46)
> +

...

>  #endif /* _ASM_X86_PGTABLE_64_DEFS_H */
> diff --git a/mm/zsmalloc.c b/mm/zsmalloc.c
> index 0787d33b80d8..132c20b6fd4f 100644
> --- a/mm/zsmalloc.c
> +++ b/mm/zsmalloc.c

...

> @@ -116,6 +100,25 @@
>   */
>  #define OBJ_ALLOCATED_TAG 1
>  #define OBJ_TAG_BITS 1
> +
> +/*
> + * MAX_POSSIBLE_PHYSMEM_BITS should be defined by all archs using zsmalloc:
> + * Trying to guess it from MAX_PHYSMEM_BITS, or considering it BITS_PER_LONG,
> + * proved to be wrong by not considering PAE capabilities, or using SPARSEMEM
> + * only headers, leading to bad object encoding due to object index overflow.
> + */
> +#ifndef MAX_POSSIBLE_PHYSMEM_BITS
> + #define MAX_POSSIBLE_PHYSMEM_BITS BITS_PER_LONG
> + #error "MAX_POSSIBLE_PHYSMEM_BITS HAS to be defined by arch using zsmalloc";
> +#else
> + #ifndef CONFIG_64BIT
> +  #if (MAX_POSSIBLE_PHYSMEM_BITS >= (BITS_PER_LONG + PAGE_SHIFT - OBJ_TAG_BITS))
> +   #error "MAX_POSSIBLE_PHYSMEM_BITS is wrong for this arch";
> +  #endif
> + #endif
> +#endif
> +
> +#define _PFN_BITS (MAX_POSSIBLE_PHYSMEM_BITS - PAGE_SHIFT)
>  #define OBJ_INDEX_BITS	(BITS_PER_LONG - _PFN_BITS - OBJ_TAG_BITS)
>  #define OBJ_INDEX_MASK	((_AC(1, UL) << OBJ_INDEX_BITS) - 1)

Have you tested it with CONFIG_X86_5LEVEL=y?

ASAICS, the patch makes OBJ_INDEX_BITS and what depends from it dynamic --
it depends what paging mode we are booting in. ZS_SIZE_CLASSES depends
indirectly on OBJ_INDEX_BITS and I don't see how struct zs_pool definition
can compile with dynamic ZS_SIZE_CLASSES.

Hm?

-- 
 Kirill A. Shutemov