[PATCH] tcp: Close timer fd when removing it from epoll

[PATCH] tcp: Close timer fd when removing it from epoll

[Laurent Vivier]

When tcp_epoll_ctl() handles a CLOSED connection, it removes the timer
from epoll but leaves the fd open until tcp_flow_defer() eventually
closes it.

Close the timer fd immediately when removing it from epoll, since
there's no reason to keep an fd around that's no longer monitored.
Set conn->timer to -1 to prevent a double-close in tcp_flow_defer().

Signed-off-by: Laurent Vivier <lvivier@redhat.com>
---
 tcp.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/tcp.c b/tcp.c
index 1db861705ddb..62064efcc1ef 100644
--- a/tcp.c
+++ b/tcp.c
@@ -535,8 +535,11 @@ static int tcp_epoll_ctl(struct tcp_tap_conn *conn)
 			int epollfd = flow_epollfd(&conn->f);
 
 			epoll_del(epollfd, conn->sock);
-			if (conn->timer != -1)
+			if (conn->timer != -1) {
 				epoll_del(epollfd, conn->timer);
+				close(conn->timer);
+				conn->timer = -1;
+			}
 		}
 
 		return 0;
-- 
2.52.0

Re: [PATCH] tcp: Close timer fd when removing it from epoll

[David Gibson]

[-- Attachment #1: Type: text/plain, Size: 1330 bytes --]

On Fri, Jan 16, 2026 at 01:55:57PM +0100, Laurent Vivier wrote:
> When tcp_epoll_ctl() handles a CLOSED connection, it removes the timer
> from epoll but leaves the fd open until tcp_flow_defer() eventually
> closes it.
> 
> Close the timer fd immediately when removing it from epoll, since
> there's no reason to keep an fd around that's no longer monitored.
> Set conn->timer to -1 to prevent a double-close in tcp_flow_defer().

This seems safe to me, but is there any actual advantage to it?

> 
> Signed-off-by: Laurent Vivier <lvivier@redhat.com>
> ---
>  tcp.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/tcp.c b/tcp.c
> index 1db861705ddb..62064efcc1ef 100644
> --- a/tcp.c
> +++ b/tcp.c
> @@ -535,8 +535,11 @@ static int tcp_epoll_ctl(struct tcp_tap_conn *conn)
>  			int epollfd = flow_epollfd(&conn->f);
>  
>  			epoll_del(epollfd, conn->sock);
> -			if (conn->timer != -1)
> +			if (conn->timer != -1) {
>  				epoll_del(epollfd, conn->timer);
> +				close(conn->timer);
> +				conn->timer = -1;
> +			}
>  		}
>  
>  		return 0;
> -- 
> 2.52.0
> 

-- 
David Gibson (he or they)	| 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/~dgibson

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH] tcp: Close timer fd when removing it from epoll

[Laurent Vivier]

On 1/19/26 05:40, David Gibson wrote:
> On Fri, Jan 16, 2026 at 01:55:57PM +0100, Laurent Vivier wrote:
>> When tcp_epoll_ctl() handles a CLOSED connection, it removes the timer
>> from epoll but leaves the fd open until tcp_flow_defer() eventually
>> closes it.
>>
>> Close the timer fd immediately when removing it from epoll, since
>> there's no reason to keep an fd around that's no longer monitored.
>> Set conn->timer to -1 to prevent a double-close in tcp_flow_defer().
> 
> This seems safe to me, but is there any actual advantage to it?
> 

No, you're right, there's no real advantage. Please drop this patch.

Thanks,
Laurent