On Tue, 8 Oct 2024 10:12:34 -0400, Luiz Augusto von Dentz wrote:
> On Tue, Oct 8, 2024 at 6:47 AM Edward Adam Davis <eadavis@xxxxxx> wrote:
> >
> > Syzbot reported a uninit-value in hci_rx_work.This is because l2cap didn't
> > execute the corresponding connection request to call l2cap_send_cmd() or
> > l2cap_do_send(), and ultimately called hci_add_acl_hdr() to set hdr->handle.
>
> What are you talking about here, what these functions have to do with
> a local handle variable?
>
> > Therefore, when calling the thread callback function hci_rx_work() to call
> > hci_acldata_packet, hdr->handle should not be used directly.
>
> It is not being used directly, the handle is a local variable which
> get assigned:
>
> handle = __le16_to_cpu(hdr->handle);
>
> If what you are saying is that there is no guarantee that skb->len >=
> HCI_ACL_HDR_SIZE then we probably want to replace skb_pull with
> skb_pull_data.
You are right, skb->len is too small.
But, skb_pull is ok, maybe just need to add a check.
#syz test
diff --git a/net/bluetooth/hci_core.c b/net/bluetooth/hci_core.c
index d6976db02c06..cfb828452a13 100644
--- a/net/bluetooth/hci_core.c
+++ b/net/bluetooth/hci_core.c
@@ -3769,7 +3769,11 @@ static void hci_acldata_packet(struct hci_dev *hdev, struct sk_buff *skb)
struct hci_conn *conn;
__u16 handle, flags;
- skb_pull(skb, HCI_ACL_HDR_SIZE);
+ if (!skb_pull(skb, HCI_ACL_HDR_SIZE)) {
+ BT_ERR("ACL data packet %d, smaller than ACL HEADER Size %d\n",
+ skb->len, HCI_ACL_HDR_SIZE);
+ goto out;
+ }
handle = __le16_to_cpu(hdr->handle);
flags = hci_flags(handle);
@@ -3795,6 +3799,7 @@ static void hci_acldata_packet(struct hci_dev *hdev, struct sk_buff *skb)
handle);
}
+out:
kfree_skb(skb);
}
