Re: [PATCH] mm/fadvise: Fix signed overflow UBSAN complaint

From: Andrew Morton
Date: Fri Jun 29 2018 - 21:38:01 EST


On Fri, 29 Jun 2018 21:44:53 +0300 Andrey Ryabinin <aryabinin@xxxxxxxxxxxxx> wrote:

> Signed integer overflow is undefined according to the C standard.
> The overflow in ksys_fadvise64_64() is deliberate, but since it is signed
> overflow, UBSAN complains:
> UBSAN: Undefined behaviour in mm/fadvise.c:76:10
> signed integer overflow:
> 4 + 9223372036854775805 cannot be represented in type 'long long int'
>
> Use unsigned types to do math. Unsigned overflow is defined so UBSAN
> will not complain about it. This patch doesn't change generated code.
>
> ...
>
> --- a/mm/fadvise.c
> +++ b/mm/fadvise.c
> @@ -73,7 +73,7 @@ int ksys_fadvise64_64(int fd, loff_t offset, loff_t len, int advice)
> }
>
> /* Careful about overflows. Len == 0 means "as much as possible" */
> - endbyte = offset + len;
> + endbyte = (u64)offset + (u64)len;
> if (!len || endbyte < len)
> endbyte = -1;
> else

Readers of this code will wonder "what the heck are those casts for".
Therefore:

--- a/mm/fadvise.c~mm-fadvise-fix-signed-overflow-ubsan-complaint-fix
+++ a/mm/fadvise.c
@@ -72,7 +72,11 @@ int ksys_fadvise64_64(int fd, loff_t off
goto out;
}

- /* Careful about overflows. Len == 0 means "as much as possible" */
+ /*
+ * Careful about overflows. Len == 0 means "as much as possible". Use
+ * unsigned math because signed overflows are undefined and UBSan
+ * complains.
+ */
endbyte = (u64)offset + (u64)len;
if (!len || endbyte < len)
endbyte = -1;
_