Currently on high rate SCTP streams the heartbeat timer refresh can
consume quite a lot of resources as timer updates are costly and it
contains a random factor, which a) is also costly and b) invalidates
mod_timer() optimization for not editing a timer to the same value.
It may even cause the timer to be slightly advanced, for no good reason.
There are 3 places that call sctp_transport_reset_timers(), including
one that calls it for every single chunk added to a packet on
sctp_outq_flush(), even if they are bundled. As there is the possibility
of working on several transports, it's not trivial to just post-poned
this call as it would require to at least remember which transports were
used.
This change then moves the random factor to outside
sctp_transport_timeout(), allowing sctp_transport_reset_timers() to
check if a timer update is really necessary. For that, the random factor
is considered 0. If timer expires is still after it, the update is not
necessary as it's a possible random value and there is no
security/functional loss in doing so.
On loopback with MTU of 65535 and data chunks with 1636, so that we
have a considerable amount of chunks without stressing system calls,
netperf -t SCTP_STREAM -l 30, perf looked like this before:
Samples: 103K of event 'cpu-clock', Event count (approx.): 25833000000
Overhead Command Shared Object Symbol
+ 6,15% netperf [kernel.vmlinux] [k] copy_user_enhanced_fast_string
- 5,43% netperf [kernel.vmlinux] [k] _raw_write_unlock_irqrestore
- _raw_write_unlock_irqrestore
- 96,54% _raw_spin_unlock_irqrestore
- 36,14% mod_timer
+ 97,24% sctp_transport_reset_timers
+ 2,76% sctp_do_sm
+ 33,65% __wake_up_sync_key
+ 28,77% sctp_ulpq_tail_event
+ 1,40% del_timer
- 1,84% mod_timer
+ 99,03% sctp_transport_reset_timers
+ 0,97% sctp_do_sm
+ 1,50% sctp_ulpq_tail_event
And after this patch:
Samples: 120K of event 'cpu-clock', Event count (approx.): 30002250000
Overhead Command Shared Object Symbol
+ 7,34% netperf [kernel.vmlinux] [k] memcpy_erms
+ 6,67% netperf [kernel.vmlinux] [k] copy_user_enhanced_fast_string
- 5,34% netperf [kernel.vmlinux] [k] _raw_write_unlock_irqrestore
- _raw_write_unlock_irqrestore
- 98,00% _raw_spin_unlock_irqrestore
+ 54,24% __wake_up_sync_key
+ 41,54% sctp_ulpq_tail_event
- 2,61% mod_timer
+ 79,88% sctp_transport_update_pmtu
+ 20,12% sctp_do_sm
+ 1,61% del_timer
+ 1,83% sctp_ulpq_tail_event
+ 2,34% netperf [kernel.vmlinux] [k] pvclock_clocksource_read
+ 2,31% netperf [sctp] [k] sctp_sendmsg
+ 1,93% netperf [kernel.vmlinux] [k] __slab_free
+ 1,92% netperf [sctp] [k] sctp_packet_transmit
+ 1,85% netperf [kernel.vmlinux] [k] kfree
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@xxxxxxxxx>
---
net/sctp/transport.c | 10 +++++++---
1 file changed, 7 insertions(+), 3 deletions(-)
diff --git a/net/sctp/transport.c b/net/sctp/transport.c
index 9b6b48c7524e4b441a151b80f0babec81f539d49..c9d6c61f5a511f96f38a0f2c275e418e158e0632 100644
--- a/net/sctp/transport.c
+++ b/net/sctp/transport.c
@@ -185,6 +185,8 @@ static void sctp_transport_destroy(struct sctp_transport *transport)
*/
void sctp_transport_reset_timers(struct sctp_transport *transport)
{
+ unsigned long expires;
+
/* RFC 2960 6.3.2 Retransmission Timer Rules
*
* R1) Every time a DATA chunk is sent to any address(including a
@@ -199,8 +201,10 @@ void sctp_transport_reset_timers(struct sctp_transport *transport)
sctp_transport_hold(transport);
/* When a data chunk is sent, reset the heartbeat interval. */
- if (!mod_timer(&transport->hb_timer,
- sctp_transport_timeout(transport)))
+ expires = sctp_transport_timeout(transport);
+ if (time_before(transport->hb_timer.expires, expires) &&
+ !mod_timer(&transport->hb_timer,
+ expires + prandom_u32_max(transport->rto)))
sctp_transport_hold(transport);
}
@@ -595,7 +599,7 @@ void sctp_transport_burst_reset(struct sctp_transport *t)
unsigned long sctp_transport_timeout(struct sctp_transport *trans)
{
/* RTO + timer slack +/- 50% of RTO */
- unsigned long timeout = (trans->rto >> 1) + prandom_u32_max(trans->rto);
+ unsigned long timeout = trans->rto >> 1;
if (trans->state != SCTP_UNCONFIRMED &&
trans->state != SCTP_PF)
--
2.5.0
--
To unsubscribe from this list: send the line "unsubscribe linux-sctp" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
