2010-08-22 22:47:34

by Arnaud Patard

Subject: [patch 1/1] rt2x00: fix "buffer size not provably correct" build warning

rt2x00 debugfs interface doesn't check the size of the data coming
from userspace, leading to a build warning. Fix That.

Signed-off-by: Arnaud Patard <[email protected]>

Index: rt2x00/drivers/net/wireless/rt2x00/rt2x00debug.c
===================================================================
--- rt2x00.orig/drivers/net/wireless/rt2x00/rt2x00debug.c 2010-08-23 00:21:42.637941244 +0200
+++ rt2x00/drivers/net/wireless/rt2x00/rt2x00debug.c 2010-08-23 00:33:41.237941022 +0200
@@ -480,6 +480,8 @@ static ssize_t rt2x00debug_write_##__nam
if (index >= debug->__name.word_count) \
return -EINVAL; \
\
+ if (length > sizeof(line)) \
+ length = sizeof(line); \
if (copy_from_user(line, buf, length)) \
return -EFAULT; \
\

2010-08-23 09:38:19

by Ivo Van Doorn

Subject: Re: [patch 1/1] rt2x00: fix "buffer size not provably correct" build warning

On Mon, Aug 23, 2010 at 12:37 AM, Arnaud Patard
<[email protected]> wrote:
> rt2x00 debugfs interface doesn't check the size of the data coming
> from userspace, leading to a build warning. Fix That.
>
> Signed-off-by: Arnaud Patard <[email protected]>
>
> Index: rt2x00/drivers/net/wireless/rt2x00/rt2x00debug.c
> ===================================================================
> --- rt2x00.orig/drivers/net/wireless/rt2x00/rt2x00debug.c ? ? ? 2010-08-23 00:21:42.637941244 +0200
> +++ rt2x00/drivers/net/wireless/rt2x00/rt2x00debug.c ? ?2010-08-23 00:33:41.237941022 +0200
> @@ -480,6 +480,8 @@ static ssize_t rt2x00debug_write_##__nam
> ? ? ? ?if (index >= debug->__name.word_count) ? ? ? ? ? ? ? ? ?\
> ? ? ? ? ? ? ? ?return -EINVAL; ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? \
> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?\
> + ? ? ? if (length > sizeof(line)) ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?\
> + ? ? ? ? ? ? ? length = sizeof(line); ? ? ? ? ? ? ? ? ? ? ? ? ?\
> ? ? ? ?if (copy_from_user(line, buf, length)) ? ? ? ? ? ? ? ? ?\
> ? ? ? ? ? ? ? ?return -EFAULT; ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? \
> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?\

I think it should actually return -EINVAL to clearly indicate that the
input is invalid,
rather then alter the value. rt2x00debug_write_* alters registers, so
we should not
assume anything when writing into it.

Ivo