wifi: ath9k_htc: Use __skb_set_length() for resetting urb before resubmit

Syzbot points out that skb_trim() has a sanity check on the existing length of
the skb, which can be uninitialised in some error paths. The intent here is
clearly just to reset the length to zero before resubmitting, so switch to
calling __skb_set_length(skb, 0) directly. In addition, __skb_set_length()
already contains a call to skb_reset_tail_pointer(), so remove the redundant
call.

The syzbot report came from ath9k_hif_usb_reg_in_cb(), but there's a similar
usage of skb_trim() in ath9k_hif_usb_rx_cb(), change both while we're at it.

Reported-by: syzbot+98afa303be379af6cdb2@syzkaller.appspotmail.com
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://patch.msgid.link/20240812142447.12328-1-toke@toke.dk

diff --git a/drivers/net/wireless/ath/ath9k/hif_usb.c b/drivers/net/wireless/ath/ath9k/hif_usb.c
index 0c7841f952287f7915d2e92e04e277bfcbf76792..a3733c9b484e46a9890d740e8aceefa24216ccb7 100644 (file)
--- a/drivers/net/wireless/ath/ath9k/hif_usb.c
+++ b/drivers/net/wireless/ath/ath9k/hif_usb.c
@@ -716,8 +716,7 @@ static void ath9k_hif_usb_rx_cb(struct urb *urb)
        }
 
 resubmit:
-       skb_reset_tail_pointer(skb);
-       skb_trim(skb, 0);
+       __skb_set_length(skb, 0);
 
        usb_anchor_urb(urb, &hif_dev->rx_submitted);
        ret = usb_submit_urb(urb, GFP_ATOMIC);
@@ -754,8 +753,7 @@ static void ath9k_hif_usb_reg_in_cb(struct urb *urb)
        case -ESHUTDOWN:
                goto free_skb;
        default:
-               skb_reset_tail_pointer(skb);
-               skb_trim(skb, 0);
+               __skb_set_length(skb, 0);
 
                goto resubmit;
        }