RE: [PATCH v5] wifi: mwifiex: Fix OOB and integer underflow when rx packets

[David Laight <David.Laight@ACULAB.COM>]

From: Polaris Pi
> Sent: 14 July 2023 13:02
> 
> Make sure mwifiex_process_mgmt_packet and its callers
> mwifiex_process_sta_rx_packet and mwifiex_process_uap_rx_packet
> not out-of-bounds access the skb->data buffer.
> 
...
> diff --git a/drivers/net/wireless/marvell/mwifiex/sta_rx.c
> b/drivers/net/wireless/marvell/mwifiex/sta_rx.c
> index 13659b02ba88..88aaec645291 100644
> --- a/drivers/net/wireless/marvell/mwifiex/sta_rx.c
> +++ b/drivers/net/wireless/marvell/mwifiex/sta_rx.c
> @@ -194,7 +194,8 @@ int mwifiex_process_sta_rx_packet(struct mwifiex_private *priv,
> 
>  	rx_pkt_hdr = (void *)local_rx_pd + rx_pkt_offset;
> 
> -	if ((rx_pkt_offset + rx_pkt_length) > (u16) skb->len) {
> +	if ((rx_pkt_offset + rx_pkt_length) > (u16)skb->len ||
> +	    skb->len - rx_pkt_offset < sizeof(*rx_pkt_hdr)) {

What is the (u16) cast for?
The domain of both tx_pkt_offset and rx_pkt_length is 16bits.
Their sum is of type 'int', masking skb->len to 16 bits just
discards high bits (which are probably zero).

The second test should probably be:
	rx_pkt_offset + sizeof(*rx_pkt_hdr) > skb->len
although the first test includes the 'offset > len' check.

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)