Searched hist:"4 b0b72f7" (Results 1 – 4 of 4) sorted by relevance
/openbmc/linux/net/core/ |
H A D | datagram.c | 4b0b72f7 Wed Apr 28 16:35:48 CDT 2010 Eric Dumazet <eric.dumazet@gmail.com> net: speedup udp receive path
Since commit 95766fff ([UDP]: Add memory accounting.), each received packet needs one extra sock_lock()/sock_release() pair.
This added latency because of possible backlog handling. Then later, ticket spinlocks added yet another latency source in case of DDOS.
This patch introduces lock_sock_bh() and unlock_sock_bh() synchronization primitives, avoiding one atomic operation and backlog processing.
skb_free_datagram_locked() uses them instead of full blown lock_sock()/release_sock(). skb is orphaned inside locked section for proper socket memory reclaim, and finally freed outside of it.
UDP receive path now take the socket spinlock only once.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> 4b0b72f7 Wed Apr 28 16:35:48 CDT 2010 Eric Dumazet <eric.dumazet@gmail.com> net: speedup udp receive path Since commit 95766fff ([UDP]: Add memory accounting.), each received packet needs one extra sock_lock()/sock_release() pair. This added latency because of possible backlog handling. Then later, ticket spinlocks added yet another latency source in case of DDOS. This patch introduces lock_sock_bh() and unlock_sock_bh() synchronization primitives, avoiding one atomic operation and backlog processing. skb_free_datagram_locked() uses them instead of full blown lock_sock()/release_sock(). skb is orphaned inside locked section for proper socket memory reclaim, and finally freed outside of it. UDP receive path now take the socket spinlock only once. Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
|
/openbmc/linux/net/ipv4/ |
H A D | udp.c | 4b0b72f7 Wed Apr 28 16:35:48 CDT 2010 Eric Dumazet <eric.dumazet@gmail.com> net: speedup udp receive path
Since commit 95766fff ([UDP]: Add memory accounting.), each received packet needs one extra sock_lock()/sock_release() pair.
This added latency because of possible backlog handling. Then later, ticket spinlocks added yet another latency source in case of DDOS.
This patch introduces lock_sock_bh() and unlock_sock_bh() synchronization primitives, avoiding one atomic operation and backlog processing.
skb_free_datagram_locked() uses them instead of full blown lock_sock()/release_sock(). skb is orphaned inside locked section for proper socket memory reclaim, and finally freed outside of it.
UDP receive path now take the socket spinlock only once.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> 4b0b72f7 Wed Apr 28 16:35:48 CDT 2010 Eric Dumazet <eric.dumazet@gmail.com> net: speedup udp receive path Since commit 95766fff ([UDP]: Add memory accounting.), each received packet needs one extra sock_lock()/sock_release() pair. This added latency because of possible backlog handling. Then later, ticket spinlocks added yet another latency source in case of DDOS. This patch introduces lock_sock_bh() and unlock_sock_bh() synchronization primitives, avoiding one atomic operation and backlog processing. skb_free_datagram_locked() uses them instead of full blown lock_sock()/release_sock(). skb is orphaned inside locked section for proper socket memory reclaim, and finally freed outside of it. UDP receive path now take the socket spinlock only once. Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
|
/openbmc/linux/net/ipv6/ |
H A D | udp.c | 4b0b72f7 Wed Apr 28 16:35:48 CDT 2010 Eric Dumazet <eric.dumazet@gmail.com> net: speedup udp receive path
Since commit 95766fff ([UDP]: Add memory accounting.), each received packet needs one extra sock_lock()/sock_release() pair.
This added latency because of possible backlog handling. Then later, ticket spinlocks added yet another latency source in case of DDOS.
This patch introduces lock_sock_bh() and unlock_sock_bh() synchronization primitives, avoiding one atomic operation and backlog processing.
skb_free_datagram_locked() uses them instead of full blown lock_sock()/release_sock(). skb is orphaned inside locked section for proper socket memory reclaim, and finally freed outside of it.
UDP receive path now take the socket spinlock only once.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> 4b0b72f7 Wed Apr 28 16:35:48 CDT 2010 Eric Dumazet <eric.dumazet@gmail.com> net: speedup udp receive path Since commit 95766fff ([UDP]: Add memory accounting.), each received packet needs one extra sock_lock()/sock_release() pair. This added latency because of possible backlog handling. Then later, ticket spinlocks added yet another latency source in case of DDOS. This patch introduces lock_sock_bh() and unlock_sock_bh() synchronization primitives, avoiding one atomic operation and backlog processing. skb_free_datagram_locked() uses them instead of full blown lock_sock()/release_sock(). skb is orphaned inside locked section for proper socket memory reclaim, and finally freed outside of it. UDP receive path now take the socket spinlock only once. Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
|
/openbmc/linux/include/net/ |
H A D | sock.h | 4b0b72f7 Wed Apr 28 16:35:48 CDT 2010 Eric Dumazet <eric.dumazet@gmail.com> net: speedup udp receive path
Since commit 95766fff ([UDP]: Add memory accounting.), each received packet needs one extra sock_lock()/sock_release() pair.
This added latency because of possible backlog handling. Then later, ticket spinlocks added yet another latency source in case of DDOS.
This patch introduces lock_sock_bh() and unlock_sock_bh() synchronization primitives, avoiding one atomic operation and backlog processing.
skb_free_datagram_locked() uses them instead of full blown lock_sock()/release_sock(). skb is orphaned inside locked section for proper socket memory reclaim, and finally freed outside of it.
UDP receive path now take the socket spinlock only once.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> 4b0b72f7 Wed Apr 28 16:35:48 CDT 2010 Eric Dumazet <eric.dumazet@gmail.com> net: speedup udp receive path Since commit 95766fff ([UDP]: Add memory accounting.), each received packet needs one extra sock_lock()/sock_release() pair. This added latency because of possible backlog handling. Then later, ticket spinlocks added yet another latency source in case of DDOS. This patch introduces lock_sock_bh() and unlock_sock_bh() synchronization primitives, avoiding one atomic operation and backlog processing. skb_free_datagram_locked() uses them instead of full blown lock_sock()/release_sock(). skb is orphaned inside locked section for proper socket memory reclaim, and finally freed outside of it. UDP receive path now take the socket spinlock only once. Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
|