Re: [PATCH] arm64: kvm: avoid overflow in integer division

[Marc Zyngier]

On Wed, 17 May 2023 21:23:39 +0100,
Arnd Bergmann <arnd@xxxxxxxxxx> wrote:
> 
> From: Arnd Bergmann <arnd@xxxxxxxx>
> 
> The newly added kvm_mmu_split_nr_page_tables() function uses DIV_ROUND_DOWN_ULL()
> to divide 64-bit addresses, but this requires a 32-bit divisior, and PUD_SIZE
> may exceed that when 64KB pages are used:
> 
> arch/arm64/kvm/mmu.c: In function 'kvm_mmu_split_nr_page_tables':
> include/linux/math.h:42:64: error: conversion from 'long unsigned int' to 'u32' {aka 'unsigned int'} changes value from '68719476736' to '0' [-Werror=overflow]
>    42 |         DIV_ROUND_DOWN_ULL((unsigned long long)(ll) + (d) - 1, (d))
>       |                                                                ^~~
> include/linux/math.h:39:47: note: in definition of macro 'DIV_ROUND_DOWN_ULL'
>    39 | #define DIV_ROUND_DOWN_ULL(ll, d) div_u64(ll, d)
>       |                                               ^
> arch/arm64/kvm/mmu.c:95:22: note: in expansion of macro 'DIV_ROUND_UP_ULL'
>    95 |                 n += DIV_ROUND_UP_ULL(range, PUD_SIZE);
>       |                      ^~~~~~~~~~~~~~~~
> 
> Since this code is only used on 64-bit targets, DIV_ROUND_UP() can deal with this
> more easily, as it already takes 64-bit arguments.
> 
> Fixes: e7bf7a490c68 ("KVM: arm64: Split huge pages when dirty logging is enabled")
> Signed-off-by: Arnd Bergmann <arnd@xxxxxxxx>
> ---
>  arch/arm64/kvm/mmu.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
> index 3386bd28d267..6db9ef288ec3 100644
> --- a/arch/arm64/kvm/mmu.c
> +++ b/arch/arm64/kvm/mmu.c
> @@ -92,8 +92,8 @@ static int kvm_mmu_split_nr_page_tables(u64 range)
>  	int n = 0;
>  
>  	if (KVM_PGTABLE_MIN_BLOCK_LEVEL < 2)
> -		n += DIV_ROUND_UP_ULL(range, PUD_SIZE);
> -	n += DIV_ROUND_UP_ULL(range, PMD_SIZE);
> +		n += DIV_ROUND_UP(range, PUD_SIZE);
> +	n += DIV_ROUND_UP(range, PMD_SIZE);
>  	return n;
>  }

This is against -next, right? Oliver, I assume you'll take this as a
fix for Ricardo's series?

Thanks,

	M.

-- 
Without deviation from the norm, progress is not possible.