On Tue, Feb 27, 2024 at 12:56:32PM +0100, Stefano Brivio wrote:[Sorry, I wrote this comment a while ago and just now realised I didn't send this out...] On Thu, 22 Feb 2024 10:21:11 +1100 David Gibson <david(a)gibson.dropbear.id.au> wrote:Ah, bother. Yes, my patch is entirely wrong, I'll drop.udp_timer_one() first checks for expiries of various sorts of sockets and if necessary sets the 'sockp' variable to trigger a cleanup at the end. If sockp is *not* set then, correctly, we don't attempt to close a non-existent socket. However, we also don't clear the flag in the udp_act[] map, in which case we'll come back here and there will, again, be nothing to be done.Well, but that's intended: if we didn't reach UDP_CONN_TIMEOUT for this port, the socket should still be in udp_act[], meaning we'll check it against activity timeouts and close on timeout. Otherwise, we'll never check that port for the activity timeout, right?-- David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibsonSo, clear the udp_act[] flag, even if we don't need to clean up a socket. Signed-off-by: David Gibson <david(a)gibson.dropbear.id.au> --- udp.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/udp.c b/udp.c index 03a5936e..4200f849 100644 --- a/udp.c +++ b/udp.c @@ -1123,8 +1123,9 @@ static void udp_timer_one(struct ctx *c, int v6, enum udp_act_type type, *sockp = -1; epoll_ctl(c->epollfd, EPOLL_CTL_DEL, s, NULL); close(s); - bitmap_clear(udp_act[v6 ? V6 : V4][type], port); } + + bitmap_clear(udp_act[v6 ? V6 : V4][type], port); } /**