Re: [PATCH] staging: rtl8192e: initializing the wep buffer

[Dan Carpenter]

On Thu, Oct 17, 2019 at 11:57:58PM -0500, Kangjie Lu wrote:
> The "wep" buffer is not initialized. To avoid memory disclosures,
> the fix initializes it, as peer functions like rtllib_ccmp_set_key
> do.
> 
> Signed-off-by: Kangjie Lu 
> ---
>  drivers/staging/rtl8192e/rtllib_crypt_wep.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/staging/rtl8192e/rtllib_crypt_wep.c 
> b/drivers/staging/rtl8192e/rtllib_crypt_wep.c
> index b1ea650036d2..0931777ed157 100644
> --- a/drivers/staging/rtl8192e/rtllib_crypt_wep.c
> +++ b/drivers/staging/rtl8192e/rtllib_crypt_wep.c
> @@ -232,6 +232,7 @@ static int prism2_wep_set_key(void *key, int len, u8 
> *seq, void *priv)
>   if (len < 0 || len > WEP_KEY_LEN)
>   return -1;
>  
> + memset(wep, 0, sizeof(*wep));
>   memcpy(wep->key, key, len);
>   wep->key_len = len;

If we read beyond wep->key_len then it's probably a bug, right?  It
doesn't matter whether it's zeroed out or not.  Fortunately we never
do:

memcpy(key, wep->key, wep->key_len);

So this change isn't required.

regards,
dan carpenter

[PATCH] staging: rtl8192e: initializing the wep buffer

[Kangjie Lu]

The "wep" buffer is not initialized. To avoid memory disclosures,
the fix initializes it, as peer functions like rtllib_ccmp_set_key
do.

Signed-off-by: Kangjie Lu 
---
 drivers/staging/rtl8192e/rtllib_crypt_wep.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/staging/rtl8192e/rtllib_crypt_wep.c 
b/drivers/staging/rtl8192e/rtllib_crypt_wep.c
index b1ea650036d2..0931777ed157 100644
--- a/drivers/staging/rtl8192e/rtllib_crypt_wep.c
+++ b/drivers/staging/rtl8192e/rtllib_crypt_wep.c
@@ -232,6 +232,7 @@ static int prism2_wep_set_key(void *key, int len, u8 *seq, 
void *priv)
if (len < 0 || len > WEP_KEY_LEN)
return -1;
 
+   memset(wep, 0, sizeof(*wep));
memcpy(wep->key, key, len);
wep->key_len = len;
 
-- 
2.17.1