2021-05-11 07:20:23

by Íñigo Huguet

[permalink] [raw]
Subject: [PATCH] rtl8xxxu: avoid parsing short RX packet

One USB data buffer can contain multiple received network
packets. If that's the case, they're processed this way:
1. Original buffer is cloned
2. Original buffer is trimmed to contain only the first
network packet
3. This first network packet is passed to network stack
4. Cloned buffer is trimmed to eliminate the first network
packet
5. Repeat with the cloned buffer until there are no more
network packets inside

However, if the space remaining in original buffer after
the first network packet is not enough to contain at least
another network packet descriptor, it is not cloned.

The loop parsing this packets ended if remaining space == 0.
But if the remaining space was > 0 but < packet descriptor
size, another iteration of the loop was done, processing again
the previous packet because cloning didn't happen. Moreover,
the ownership of this packet had been passed to network
stack in the previous iteration.

This patch ensures that no extra iteration is done if the
remaining size is not enough for one packet, and also avoid
the first iteration for the same reason.

Probably this doesn't happen in practice, but can happen
theoretically.

Signed-off-by: Íñigo Huguet <[email protected]>
---
drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_core.c | 11 +++++++++--
1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_core.c b/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_core.c
index 9ff09cf7eb62..673961a82c40 100644
--- a/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_core.c
+++ b/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_core.c
@@ -5554,6 +5554,11 @@ int rtl8xxxu_parse_rxdesc16(struct rtl8xxxu_priv *priv, struct sk_buff *skb)
urb_len = skb->len;
pkt_cnt = 0;

+ if (urb_len < sizeof(struct rtl8xxxu_rxdesc16)) {
+ kfree_skb(skb);
+ return RX_TYPE_ERROR;
+ }
+
do {
rx_desc = (struct rtl8xxxu_rxdesc16 *)skb->data;
_rx_desc_le = (__le32 *)skb->data;
@@ -5581,7 +5586,7 @@ int rtl8xxxu_parse_rxdesc16(struct rtl8xxxu_priv *priv, struct sk_buff *skb)
* at least cover the rx descriptor
*/
if (pkt_cnt > 1 &&
- urb_len > (pkt_offset + sizeof(struct rtl8xxxu_rxdesc16)))
+ urb_len >= (pkt_offset + sizeof(struct rtl8xxxu_rxdesc16)))
next_skb = skb_clone(skb, GFP_ATOMIC);

rx_status = IEEE80211_SKB_RXCB(skb);
@@ -5627,7 +5632,9 @@ int rtl8xxxu_parse_rxdesc16(struct rtl8xxxu_priv *priv, struct sk_buff *skb)

pkt_cnt--;
urb_len -= pkt_offset;
- } while (skb && urb_len > 0 && pkt_cnt > 0);
+ next_skb = NULL;
+ } while (skb && pkt_cnt > 0 &&
+ urb_len >= sizeof(struct rtl8xxxu_rxdesc16));

return RX_TYPE_DATA_PKT;
}
--
2.31.1


2021-06-01 09:01:06

by Íñigo Huguet

[permalink] [raw]
Subject: Re: [PATCH] rtl8xxxu: avoid parsing short RX packet

On Tue, May 11, 2021 at 9:19 AM Íñigo Huguet <[email protected]> wrote:
>
> One USB data buffer can contain multiple received network
> packets. If that's the case, they're processed this way:
> 1. Original buffer is cloned
> 2. Original buffer is trimmed to contain only the first
> network packet
> 3. This first network packet is passed to network stack
> 4. Cloned buffer is trimmed to eliminate the first network
> packet
> 5. Repeat with the cloned buffer until there are no more
> network packets inside
>
> However, if the space remaining in original buffer after
> the first network packet is not enough to contain at least
> another network packet descriptor, it is not cloned.
>
> The loop parsing this packets ended if remaining space == 0.
> But if the remaining space was > 0 but < packet descriptor
> size, another iteration of the loop was done, processing again
> the previous packet because cloning didn't happen. Moreover,
> the ownership of this packet had been passed to network
> stack in the previous iteration.
>
> This patch ensures that no extra iteration is done if the
> remaining size is not enough for one packet, and also avoid
> the first iteration for the same reason.
>
> Probably this doesn't happen in practice, but can happen
> theoretically.
>
> Signed-off-by: Íñigo Huguet <[email protected]>
> ---
> drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_core.c | 11 +++++++++--
> 1 file changed, 9 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_core.c b/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_core.c
> index 9ff09cf7eb62..673961a82c40 100644
> --- a/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_core.c
> +++ b/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_core.c
> @@ -5554,6 +5554,11 @@ int rtl8xxxu_parse_rxdesc16(struct rtl8xxxu_priv *priv, struct sk_buff *skb)
> urb_len = skb->len;
> pkt_cnt = 0;
>
> + if (urb_len < sizeof(struct rtl8xxxu_rxdesc16)) {
> + kfree_skb(skb);
> + return RX_TYPE_ERROR;
> + }
> +
> do {
> rx_desc = (struct rtl8xxxu_rxdesc16 *)skb->data;
> _rx_desc_le = (__le32 *)skb->data;
> @@ -5581,7 +5586,7 @@ int rtl8xxxu_parse_rxdesc16(struct rtl8xxxu_priv *priv, struct sk_buff *skb)
> * at least cover the rx descriptor
> */
> if (pkt_cnt > 1 &&
> - urb_len > (pkt_offset + sizeof(struct rtl8xxxu_rxdesc16)))
> + urb_len >= (pkt_offset + sizeof(struct rtl8xxxu_rxdesc16)))
> next_skb = skb_clone(skb, GFP_ATOMIC);
>
> rx_status = IEEE80211_SKB_RXCB(skb);
> @@ -5627,7 +5632,9 @@ int rtl8xxxu_parse_rxdesc16(struct rtl8xxxu_priv *priv, struct sk_buff *skb)
>
> pkt_cnt--;
> urb_len -= pkt_offset;
> - } while (skb && urb_len > 0 && pkt_cnt > 0);
> + next_skb = NULL;
> + } while (skb && pkt_cnt > 0 &&
> + urb_len >= sizeof(struct rtl8xxxu_rxdesc16));
>
> return RX_TYPE_DATA_PKT;
> }
> --
> 2.31.1
>

Hello,

About 3 weeks ago I sent this patch, but received no response. Any
feedback would be appreciated.

Thanks!
--
Íñigo Huguet

2021-06-03 09:54:32

by Kalle Valo

[permalink] [raw]
Subject: Re: [PATCH] rtl8xxxu: avoid parsing short RX packet

Íñigo Huguet <[email protected]> writes:

> On Tue, May 11, 2021 at 9:19 AM Íñigo Huguet <[email protected]> wrote:
>>
>> One USB data buffer can contain multiple received network
>> packets. If that's the case, they're processed this way:
>> 1. Original buffer is cloned
>> 2. Original buffer is trimmed to contain only the first
>> network packet
>> 3. This first network packet is passed to network stack
>> 4. Cloned buffer is trimmed to eliminate the first network
>> packet
>> 5. Repeat with the cloned buffer until there are no more
>> network packets inside
>>
>> However, if the space remaining in original buffer after
>> the first network packet is not enough to contain at least
>> another network packet descriptor, it is not cloned.
>>
>> The loop parsing this packets ended if remaining space == 0.
>> But if the remaining space was > 0 but < packet descriptor
>> size, another iteration of the loop was done, processing again
>> the previous packet because cloning didn't happen. Moreover,
>> the ownership of this packet had been passed to network
>> stack in the previous iteration.
>>
>> This patch ensures that no extra iteration is done if the
>> remaining size is not enough for one packet, and also avoid
>> the first iteration for the same reason.
>>
>> Probably this doesn't happen in practice, but can happen
>> theoretically.
>>
>> Signed-off-by: Íñigo Huguet <[email protected]>
>
> About 3 weeks ago I sent this patch, but received no response. Any
> feedback would be appreciated.

Maintainers are sometimes so busy that review takes extra long, but you
can always check the state in patchwork:

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches#checking_state_of_patches_from_patchwork

--
https://patchwork.kernel.org/project/linux-wireless/list/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches