| Message ID | 20240515163358.4105915-1-edumazet@google.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 581073f626e387d3e7eed55c48c8495584ead7ba |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net] af_packet: do not call packet_read_pending() from tpacket_destruct_skb() | expand |
Eric Dumazet wrote: > trafgen performance considerably sank on hosts with many cores > after the blamed commit. > > packet_read_pending() is very expensive, and calling it > in af_packet fast path defeats Daniel intent in commit > b013840810c2 ("packet: use percpu mmap tx frame pending refcount") > > tpacket_destruct_skb() makes room for one packet, we can immediately > wakeup a producer, no need to completely drain the tx ring. > > Fixes: 89ed5b519004 ("af_packet: Block execution of tasks waiting for transmit to complete in AF_PACKET") > Signed-off-by: Eric Dumazet <edumazet@google.com> > Cc: Neil Horman <nhorman@tuxdriver.com> > Cc: Daniel Borkmann <daniel@iogearbox.net> > Cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com> Reviewed-by: Willem de Bruijn <willemb@google.com> Thanks!
Hello: This patch was applied to netdev/net.git (main) by Jakub Kicinski <kuba@kernel.org>: On Wed, 15 May 2024 16:33:58 +0000 you wrote: > trafgen performance considerably sank on hosts with many cores > after the blamed commit. > > packet_read_pending() is very expensive, and calling it > in af_packet fast path defeats Daniel intent in commit > b013840810c2 ("packet: use percpu mmap tx frame pending refcount") > > [...] Here is the summary with links: - [net] af_packet: do not call packet_read_pending() from tpacket_destruct_skb() https://git.kernel.org/netdev/net/c/581073f626e3 You are awesome, thank you!
On 5/15/24 6:33 PM, Eric Dumazet wrote: > trafgen performance considerably sank on hosts with many cores > after the blamed commit. > > packet_read_pending() is very expensive, and calling it > in af_packet fast path defeats Daniel intent in commit > b013840810c2 ("packet: use percpu mmap tx frame pending refcount") > > tpacket_destruct_skb() makes room for one packet, we can immediately > wakeup a producer, no need to completely drain the tx ring. > > Fixes: 89ed5b519004 ("af_packet: Block execution of tasks waiting for transmit to complete in AF_PACKET") > Signed-off-by: Eric Dumazet <edumazet@google.com> > Cc: Neil Horman <nhorman@tuxdriver.com> > Cc: Daniel Borkmann <daniel@iogearbox.net> > Cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com> Acked-by: Daniel Borkmann <daniel@iogearbox.net> Thanks Eric!
diff --git a/net/packet/af_packet.c b/net/packet/af_packet.c index 8c6d3fbb4ed87f17c2e365810106a05fe9b8ff0c..ea3ebc160e25cc661901717a755f47db927c304d 100644 --- a/net/packet/af_packet.c +++ b/net/packet/af_packet.c @@ -2522,8 +2522,7 @@ static void tpacket_destruct_skb(struct sk_buff *skb) ts = __packet_set_timestamp(po, ph, skb); __packet_set_status(po, ph, TP_STATUS_AVAILABLE | ts); - if (!packet_read_pending(&po->tx_ring)) - complete(&po->skb_completion); + complete(&po->skb_completion); } sock_wfree(skb);
trafgen performance considerably sank on hosts with many cores after the blamed commit. packet_read_pending() is very expensive, and calling it in af_packet fast path defeats Daniel intent in commit b013840810c2 ("packet: use percpu mmap tx frame pending refcount") tpacket_destruct_skb() makes room for one packet, we can immediately wakeup a producer, no need to completely drain the tx ring. Fixes: 89ed5b519004 ("af_packet: Block execution of tasks waiting for transmit to complete in AF_PACKET") Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Neil Horman <nhorman@tuxdriver.com> Cc: Daniel Borkmann <daniel@iogearbox.net> Cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com> --- net/packet/af_packet.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)