2019-12-28 10:37:43

by isidentical

Subject: [PATCH] mm/hugetlb: ensure signedness of large numbers

This change introduces a sanitity helper for
numbers that are larger than 2^5.

Signed-off-by: isidentical <[email protected]>
---
include/uapi/asm-generic/hugetlb_encode.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/uapi/asm-generic/hugetlb_encode.h b/include/uapi/asm-generic/hugetlb_encode.h
index b0f8e87235bd..42c06c62ae17 100644
--- a/include/uapi/asm-generic/hugetlb_encode.h
+++ b/include/uapi/asm-generic/hugetlb_encode.h
@@ -31,6 +31,6 @@
#define HUGETLB_FLAG_ENCODE_512MB (29 << HUGETLB_FLAG_ENCODE_SHIFT)
#define HUGETLB_FLAG_ENCODE_1GB (30 << HUGETLB_FLAG_ENCODE_SHIFT)
#define HUGETLB_FLAG_ENCODE_2GB (31 << HUGETLB_FLAG_ENCODE_SHIFT)
-#define HUGETLB_FLAG_ENCODE_16GB (34 << HUGETLB_FLAG_ENCODE_SHIFT)
+#define HUGETLB_FLAG_ENCODE_16GB (UINT32_C(34) << HUGETLB_FLAG_ENCODE_SHIFT)

#endif /* _ASM_GENERIC_HUGETLB_ENCODE_H_ */
--
2.20.1

2019-12-29 01:32:47

by James Bottomley

Subject: Re: [PATCH] mm/hugetlb: ensure signedness of large numbers

On Sat, 2019-12-28 at 13:33 +0300, isidentical wrote:
> This change introduces a sanitity helper for
> numbers that are larger than 2^5.

What's the rationale for this ... i.e. what problem are you trying to
solve? left to its own devices, gcc will assume int size for all
literals, so 34<<26 doesn't appear to have a problem, except that it
will be sign extended as a negative number into a 64 bit variable.
However, it's designed use is for an int flags in mmap, so its current
definition seems to be fine.

> Signed-off-by: isidentical <[email protected]>
> ---
> include/uapi/asm-generic/hugetlb_encode.h | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/include/uapi/asm-generic/hugetlb_encode.h
> b/include/uapi/asm-generic/hugetlb_encode.h
> index b0f8e87235bd..42c06c62ae17 100644
> --- a/include/uapi/asm-generic/hugetlb_encode.h
> +++ b/include/uapi/asm-generic/hugetlb_encode.h
> @@ -31,6 +31,6 @@
> #define HUGETLB_FLAG_ENCODE_512MB (29 <<
> HUGETLB_FLAG_ENCODE_SHIFT)
> #define HUGETLB_FLAG_ENCODE_1GB (30 <<
> HUGETLB_FLAG_ENCODE_SHIFT)
> #define HUGETLB_FLAG_ENCODE_2GB (31 <<
> HUGETLB_FLAG_ENCODE_SHIFT)
> -#define HUGETLB_FLAG_ENCODE_16GB (34 <<
> HUGETLB_FLAG_ENCODE_SHIFT)
> +#define HUGETLB_FLAG_ENCODE_16GB (UINT32_C(34) <<
> HUGETLB_FLAG_ENCODE_SHIFT)

And if there is a literal problem here, it can't be solved like this:
UINT32_C is a Cism which has no analogue in the kernel because we don't
pull in the /usr/include headers where it is defined. Usually in the
kernel we solve this by making the literal type explicit, like 34U, but
as I said above, I don't see a problem that this would solve.

James