diff options
| author | Eric Dumazet <eric.dumazet@gmail.com> | 2011-10-13 07:28:54 +0000 |
|---|---|---|
| committer | David S. Miller <davem@davemloft.net> | 2011-10-13 16:05:07 -0400 |
| commit | 87fb4b7b533073eeeaed0b6bf7c2328995f6c075 (patch) | |
| tree | be4b37f08d7fe2d018ae68bae4577b1b2bafd0fc /net/ipv4 | |
| parent | 97ba0eb64ca690165f945a83e609102fcefa99cb (diff) | |
| download | op-kernel-dev-87fb4b7b533073eeeaed0b6bf7c2328995f6c075.zip op-kernel-dev-87fb4b7b533073eeeaed0b6bf7c2328995f6c075.tar.gz | |
net: more accurate skb truesize
skb truesize currently accounts for sk_buff struct and part of skb head.
kmalloc() roundings are also ignored.
Considering that skb_shared_info is larger than sk_buff, its time to
take it into account for better memory accounting.
This patch introduces SKB_TRUESIZE(X) macro to centralize various
assumptions into a single place.
At skb alloc phase, we put skb_shared_info struct at the exact end of
skb head, to allow a better use of memory (lowering number of
reallocations), since kmalloc() gives us power-of-two memory blocks.
Unless SLUB/SLUB debug is active, both skb->head and skb_shared_info are
aligned to cache lines, as before.
Note: This patch might trigger performance regressions because of
misconfigured protocol stacks, hitting per socket or global memory
limits that were previously not reached. But its a necessary step for a
more accurate memory accounting.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Andi Kleen <ak@linux.intel.com>
CC: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net/ipv4')
| -rw-r--r-- | net/ipv4/icmp.c | 5 | ||||
| -rw-r--r-- | net/ipv4/tcp_input.c | 14 |
2 files changed, 9 insertions, 10 deletions
diff --git a/net/ipv4/icmp.c b/net/ipv4/icmp.c index 23ef31b..ab188ae 100644 --- a/net/ipv4/icmp.c +++ b/net/ipv4/icmp.c @@ -1152,10 +1152,9 @@ static int __net_init icmp_sk_init(struct net *net) net->ipv4.icmp_sk[i] = sk; /* Enough space for 2 64K ICMP packets, including - * sk_buff struct overhead. + * sk_buff/skb_shared_info struct overhead. */ - sk->sk_sndbuf = - (2 * ((64 * 1024) + sizeof(struct sk_buff))); + sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024); /* * Speedup sock_wfree() diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 81cae64..c1653fe 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -265,8 +265,7 @@ static inline int TCP_ECN_rcv_ecn_echo(struct tcp_sock *tp, struct tcphdr *th) static void tcp_fixup_sndbuf(struct sock *sk) { - int sndmem = tcp_sk(sk)->rx_opt.mss_clamp + MAX_TCP_HEADER + 16 + - sizeof(struct sk_buff); + int sndmem = SKB_TRUESIZE(tcp_sk(sk)->rx_opt.mss_clamp + MAX_TCP_HEADER); if (sk->sk_sndbuf < 3 * sndmem) { sk->sk_sndbuf = 3 * sndmem; @@ -349,7 +348,7 @@ static void tcp_grow_window(struct sock *sk, struct sk_buff *skb) static void tcp_fixup_rcvbuf(struct sock *sk) { struct tcp_sock *tp = tcp_sk(sk); - int rcvmem = tp->advmss + MAX_TCP_HEADER + 16 + sizeof(struct sk_buff); + int rcvmem = SKB_TRUESIZE(tp->advmss + MAX_TCP_HEADER); /* Try to select rcvbuf so that 4 mss-sized segments * will fit to window and corresponding skbs will fit to our rcvbuf. @@ -540,8 +539,7 @@ void tcp_rcv_space_adjust(struct sock *sk) space /= tp->advmss; if (!space) space = 1; - rcvmem = (tp->advmss + MAX_TCP_HEADER + - 16 + sizeof(struct sk_buff)); + rcvmem = SKB_TRUESIZE(tp->advmss + MAX_TCP_HEADER); while (tcp_win_from_space(rcvmem) < tp->advmss) rcvmem += 128; space *= rcvmem; @@ -4950,8 +4948,10 @@ static void tcp_new_space(struct sock *sk) struct tcp_sock *tp = tcp_sk(sk); if (tcp_should_expand_sndbuf(sk)) { - int sndmem = max_t(u32, tp->rx_opt.mss_clamp, tp->mss_cache) + - MAX_TCP_HEADER + 16 + sizeof(struct sk_buff); + int sndmem = SKB_TRUESIZE(max_t(u32, + tp->rx_opt.mss_clamp, + tp->mss_cache) + + MAX_TCP_HEADER); int demanded = max_t(unsigned int, tp->snd_cwnd, tp->reordering + 1); sndmem *= 2 * demanded; |
