| Message ID | 20220505131811.3744503-3-asavkov@redhat.com (mailing list archive) |
|---|---|
| State | Changes Requested |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | Upper bound kernel timers | expand |
On Thu, May 05, 2022 at 03:18:11PM +0200, Artem Savkov wrote: > Make sure TCP keepalive timer does not expire late. Switching to upper > bound timers means it can fire off early but in case of keepalive > tcp_keepalive_timer() handler checks elapsed time and resets the timer > if it was triggered early. This results in timer "cascading" to a > higher precision and being just a couple of milliseconds off it's > original mark. > This adds minimal overhead as keepalive timers are never re-armed and > are usually quite long. > > Signed-off-by: Artem Savkov <asavkov@redhat.com> > --- > net/ipv4/inet_connection_sock.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/net/ipv4/inet_connection_sock.c b/net/ipv4/inet_connection_sock.c > index 1e5b53c2bb26..bb2dbfb6f5b5 100644 > --- a/net/ipv4/inet_connection_sock.c > +++ b/net/ipv4/inet_connection_sock.c > @@ -589,7 +589,7 @@ EXPORT_SYMBOL(inet_csk_delete_keepalive_timer); > > void inet_csk_reset_keepalive_timer(struct sock *sk, unsigned long len) > { > - sk_reset_timer(sk, &sk->sk_timer, jiffies + len); > + sk_reset_timer(sk, &sk->sk_timer, jiffies + upper_bound_timeout(len)); > } > EXPORT_SYMBOL(inet_csk_reset_keepalive_timer); As I mentioned before, there might be two sides to the same coin, depending on whether the keepalive is detecting vs preventing the disconnect. So this might possibly fix one case, while breaking another. Hopefully a networking expert can chime in.
On Thu, May 05, 2022 at 10:56:54AM -0700, Josh Poimboeuf wrote: > On Thu, May 05, 2022 at 03:18:11PM +0200, Artem Savkov wrote: > > Make sure TCP keepalive timer does not expire late. Switching to upper > > bound timers means it can fire off early but in case of keepalive > > tcp_keepalive_timer() handler checks elapsed time and resets the timer > > if it was triggered early. This results in timer "cascading" to a > > higher precision and being just a couple of milliseconds off it's > > original mark. > > This adds minimal overhead as keepalive timers are never re-armed and > > are usually quite long. > > > > Signed-off-by: Artem Savkov <asavkov@redhat.com> > > --- > > net/ipv4/inet_connection_sock.c | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > diff --git a/net/ipv4/inet_connection_sock.c b/net/ipv4/inet_connection_sock.c > > index 1e5b53c2bb26..bb2dbfb6f5b5 100644 > > --- a/net/ipv4/inet_connection_sock.c > > +++ b/net/ipv4/inet_connection_sock.c > > @@ -589,7 +589,7 @@ EXPORT_SYMBOL(inet_csk_delete_keepalive_timer); > > > > void inet_csk_reset_keepalive_timer(struct sock *sk, unsigned long len) > > { > > - sk_reset_timer(sk, &sk->sk_timer, jiffies + len); > > + sk_reset_timer(sk, &sk->sk_timer, jiffies + upper_bound_timeout(len)); > > } > > EXPORT_SYMBOL(inet_csk_reset_keepalive_timer); > > As I mentioned before, there might be two sides to the same coin, > depending on whether the keepalive is detecting vs preventing the > disconnect. So this might possibly fix one case, while breaking > another. But cascading is still there in the handler so it will fire off quite close to original timer in any case.
On Fri, May 06, 2022 at 08:39:08AM +0200, Artem Savkov wrote: > On Thu, May 05, 2022 at 10:56:54AM -0700, Josh Poimboeuf wrote: > > On Thu, May 05, 2022 at 03:18:11PM +0200, Artem Savkov wrote: > > > Make sure TCP keepalive timer does not expire late. Switching to upper > > > bound timers means it can fire off early but in case of keepalive > > > tcp_keepalive_timer() handler checks elapsed time and resets the timer > > > if it was triggered early. This results in timer "cascading" to a > > > higher precision and being just a couple of milliseconds off it's > > > original mark. > > > This adds minimal overhead as keepalive timers are never re-armed and > > > are usually quite long. > > > > > > Signed-off-by: Artem Savkov <asavkov@redhat.com> > > > --- > > > net/ipv4/inet_connection_sock.c | 2 +- > > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > > > diff --git a/net/ipv4/inet_connection_sock.c b/net/ipv4/inet_connection_sock.c > > > index 1e5b53c2bb26..bb2dbfb6f5b5 100644 > > > --- a/net/ipv4/inet_connection_sock.c > > > +++ b/net/ipv4/inet_connection_sock.c > > > @@ -589,7 +589,7 @@ EXPORT_SYMBOL(inet_csk_delete_keepalive_timer); > > > > > > void inet_csk_reset_keepalive_timer(struct sock *sk, unsigned long len) > > > { > > > - sk_reset_timer(sk, &sk->sk_timer, jiffies + len); > > > + sk_reset_timer(sk, &sk->sk_timer, jiffies + upper_bound_timeout(len)); > > > } > > > EXPORT_SYMBOL(inet_csk_reset_keepalive_timer); > > > > As I mentioned before, there might be two sides to the same coin, > > depending on whether the keepalive is detecting vs preventing the > > disconnect. So this might possibly fix one case, while breaking > > another. > > But cascading is still there in the handler so it will fire off quite > close to original timer in any case. Ah, indeed it does. Sorry, I should try actually reading the patch description next time :-/ Looks good to me. Reviewed-by: Josh Poimboeuf <jpoimboe@kernel.org>
diff --git a/net/ipv4/inet_connection_sock.c b/net/ipv4/inet_connection_sock.c index 1e5b53c2bb26..bb2dbfb6f5b5 100644 --- a/net/ipv4/inet_connection_sock.c +++ b/net/ipv4/inet_connection_sock.c @@ -589,7 +589,7 @@ EXPORT_SYMBOL(inet_csk_delete_keepalive_timer); void inet_csk_reset_keepalive_timer(struct sock *sk, unsigned long len) { - sk_reset_timer(sk, &sk->sk_timer, jiffies + len); + sk_reset_timer(sk, &sk->sk_timer, jiffies + upper_bound_timeout(len)); } EXPORT_SYMBOL(inet_csk_reset_keepalive_timer);
Make sure TCP keepalive timer does not expire late. Switching to upper bound timers means it can fire off early but in case of keepalive tcp_keepalive_timer() handler checks elapsed time and resets the timer if it was triggered early. This results in timer "cascading" to a higher precision and being just a couple of milliseconds off it's original mark. This adds minimal overhead as keepalive timers are never re-armed and are usually quite long. Signed-off-by: Artem Savkov <asavkov@redhat.com> --- net/ipv4/inet_connection_sock.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)