[Bug 276363] if_wg: Fix a code bug in calculate_padding() although impossible to hit

From: <bugzilla-noreply_at_freebsd.org>
Date: Tue, 16 Jan 2024 14:29:19 UTC
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=276363

--- Comment #1 from Aaron LI <aly@aaronly.me> ---
Well, after testing, I found 'pkt->p_mtu' is actually 0 for keepalive packets;
however, those packets also have a zero length, so calculate_padding() didn't
cause any real problems.

I suggest the following new patch:

--- if_wg.c.orig        2023-10-12 09:06:16.983637264 +0800
+++ if_wg.c     2024-01-16 22:25:04.878629478 +0800
@@ -1461,8 +1461,11 @@ calculate_padding(struct wg_packet *pkt)
 {
        unsigned int padded_size, last_unit = pkt->p_mbuf->m_pkthdr.len;

-       if (__predict_false(!pkt->p_mtu))
-               return (last_unit + (WG_PKT_PADDING - 1)) & ~(WG_PKT_PADDING -
1);
+       /* Keepalive packets don't set p_mtu, but also have a zreo length. */
+       if (__predict_false(pkt->p_mtu == 0)) {
+               padded_size = (last_unit + (WG_PKT_PADDING - 1)) &
~(WG_PKT_PADDING - 1);
+               return padded_size - last_unit;
+       }

        if (__predict_false(last_unit > pkt->p_mtu))
                last_unit %= pkt->p_mtu;

-- 
You are receiving this mail because:
You are the assignee for the bug.