2022-04-22 08:37:40

by baihaowen

Subject: [PATCH V5] staging: rtl8192e: Fix signedness bug in rtllib_rx_assoc_resp()

The rtllib_rx_assoc_resp() function has a signedness bug because it's
a declared as a u16 but it return -ENOMEM. When you look at it more
closely it returns a mix of error codes including 0xcafe, -ENOMEM, and
a->status which is WLAN_STATUS_NOT_SUPPORTED_AUTH_ALG. This is a mess.

Clean it up to just return standard kernel error codes. We can print
out the a->status before returning a regular error code. The printks
in the caller need to be adjusted as well.

Signed-off-by: Haowen Bai <[email protected]>
---
V1->V2: reduce return random value; print its own error message.
V2->V3: change commit message; change s16 -> int.
V3->V4:
1. change message suggested by Dan Carpenter;
2. hold a->status in auth_parse() and return error code or 0 on success.
3. print le16_to_cpu(errcode) -> int %d.
V4->V5: fix compile error.

drivers/staging/rtl8192e/rtllib_softmac.c | 17 +++++++++++------
1 file changed, 11 insertions(+), 6 deletions(-)

diff --git a/drivers/staging/rtl8192e/rtllib_softmac.c b/drivers/staging/rtl8192e/rtllib_softmac.c
index 82bf05eb1cbf..38ac733c3245 100644
--- a/drivers/staging/rtl8192e/rtllib_softmac.c
+++ b/drivers/staging/rtl8192e/rtllib_softmac.c
@@ -1764,7 +1764,7 @@ static void rtllib_softmac_check_all_nets(struct rtllib_device *ieee)
spin_unlock_irqrestore(&ieee->lock, flags);
}

-static inline u16 auth_parse(struct net_device *dev, struct sk_buff *skb,
+static inline int auth_parse(struct net_device *dev, struct sk_buff *skb,
u8 **challenge, int *chlen)
{
struct rtllib_authentication *a;
@@ -1773,7 +1773,7 @@ static inline u16 auth_parse(struct net_device *dev, struct sk_buff *skb,
if (skb->len < (sizeof(struct rtllib_authentication) -
sizeof(struct rtllib_info_element))) {
netdev_dbg(dev, "invalid len in auth resp: %d\n", skb->len);
- return 0xcafe;
+ return -EINVAL;
}
*challenge = NULL;
a = (struct rtllib_authentication *) skb->data;
@@ -1787,7 +1787,13 @@ static inline u16 auth_parse(struct net_device *dev, struct sk_buff *skb,
return -ENOMEM;
}
}
- return le16_to_cpu(a->status);
+
+ if (a->status) {
+ netdev_dbg(dev, "auth_parse() failed\n");
+ return -EINVAL;
+ }
+
+ return 0;
}

static int auth_rq_parse(struct net_device *dev, struct sk_buff *skb, u8 *dest)
@@ -2282,7 +2288,7 @@ rtllib_rx_assoc_resp(struct rtllib_device *ieee, struct sk_buff *skb,

static void rtllib_rx_auth_resp(struct rtllib_device *ieee, struct sk_buff *skb)
{
- u16 errcode;
+ int errcode;
u8 *challenge;
int chlen = 0;
bool bSupportNmode = true, bHalfSupportNmode = false;
@@ -2292,8 +2298,7 @@ static void rtllib_rx_auth_resp(struct rtllib_device *ieee, struct sk_buff *skb)
if (errcode) {
ieee->softmac_stats.rx_auth_rs_err++;
netdev_info(ieee->dev,
- "Authentication response status code 0x%x",
- errcode);
+ "Authentication response status code %d", errcode);
rtllib_associate_abort(ieee);
return;
}
--
2.7.4

2022-04-22 21:21:17

by Dan Carpenter

Subject: Re: [PATCH V5] staging: rtl8192e: Fix signedness bug in rtllib_rx_assoc_resp()

On Thu, Apr 21, 2022 at 04:21:17PM +0800, Haowen Bai wrote:
> The rtllib_rx_assoc_resp() function has a signedness bug because it's
> a declared as a u16 but it return -ENOMEM. When you look at it more
> closely it returns a mix of error codes including 0xcafe, -ENOMEM, and
> a->status which is WLAN_STATUS_NOT_SUPPORTED_AUTH_ALG. This is a mess.
>
> Clean it up to just return standard kernel error codes. We can print
> out the a->status before returning a regular error code. The printks
> in the caller need to be adjusted as well.
>
> Signed-off-by: Haowen Bai <[email protected]>
> ---
> V1->V2: reduce return random value; print its own error message.
> V2->V3: change commit message; change s16 -> int.
> V3->V4:
> 1. change message suggested by Dan Carpenter;
> 2. hold a->status in auth_parse() and return error code or 0 on success.
> 3. print le16_to_cpu(errcode) -> int %d.
> V4->V5: fix compile error.
>

Looks ok.

Reviewed-by: Dan Carpenter <[email protected]>

regards,
dan carpenter