[PATCH net-next 0/2] net: Namespace-ify sysctl_tcp_rmem and sysctl_tcp_wmem

[PATCH net-next 0/2] net: Namespace-ify sysctl_tcp_rmem and sysctl_tcp_wmem

[Eric Dumazet]

We need to get per netns sysctl for sysctl_[proto]_rmem and sysctl_[proto]_wmem

This patch series adds the basic infrastructure allowing per proto
conversion, and takes care of TCP.

Eric Dumazet (2):
  net: allow per netns sysctl_rmem and sysctl_wmem for protos
  tcp: Namespace-ify sysctl_tcp_rmem and sysctl_tcp_wmem

 include/net/netns/ipv4.h    |  2 ++
 include/net/sock.h          | 22 ++++++++++++++++++++++
 include/net/tcp.h           |  2 --
 include/trace/events/sock.h |  2 +-
 net/core/sock.c             | 10 ++++++----
 net/ipv4/sysctl_net_ipv4.c  | 32 ++++++++++++++++----------------
 net/ipv4/tcp.c              | 21 ++++++++-------------
 net/ipv4/tcp_input.c        | 14 ++++++++------
 net/ipv4/tcp_ipv4.c         | 13 ++++++++++---
 net/ipv4/tcp_output.c       |  2 +-
 net/ipv6/tcp_ipv6.c         |  4 ++--
 11 files changed, 76 insertions(+), 48 deletions(-)

-- 
2.15.0.403.gc27cc4dac6-goog

[PATCH net-next 1/2] net: allow per netns sysctl_rmem and sysctl_wmem for protos

[Eric Dumazet]

As we want to gradually implement per netns sysctl_rmem and sysctl_wmem
on per protocol basis, add two new fields in struct proto,
and two new helpers : sk_get_wmem0() and sk_get_rmem0()

First user will be TCP. Then UDP and SCTP can be easily converted,
while DECNET probably wont get this support.

Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 include/net/sock.h          | 22 ++++++++++++++++++++++
 include/trace/events/sock.h |  2 +-
 net/core/sock.c             | 10 ++++++----
 3 files changed, 29 insertions(+), 5 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index 6f1be9726e02b350162573117ed6eb6cebc627b4..688a823dccc306bd21f47da167c6922161af5a6a 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -1101,8 +1101,12 @@ struct proto {
 	 */
 	unsigned long		*memory_pressure;
 	long			*sysctl_mem;
+
 	int			*sysctl_wmem;
 	int			*sysctl_rmem;
+	u32			sysctl_wmem_offset;
+	u32			sysctl_rmem_offset;
+
 	int			max_header;
 	bool			no_autobind;
 
@@ -2390,4 +2394,22 @@ extern int sysctl_optmem_max;
 extern __u32 sysctl_wmem_default;
 extern __u32 sysctl_rmem_default;
 
+static inline int sk_get_wmem0(const struct sock *sk, const struct proto *proto)
+{
+	/* Does this proto have per netns sysctl_wmem ? */
+	if (proto->sysctl_wmem_offset)
+		return *(int *)((void *)sock_net(sk) + proto->sysctl_wmem_offset);
+
+	return *proto->sysctl_wmem;
+}
+
+static inline int sk_get_rmem0(const struct sock *sk, const struct proto *proto)
+{
+	/* Does this proto have per netns sysctl_rmem ? */
+	if (proto->sysctl_rmem_offset)
+		return *(int *)((void *)sock_net(sk) + proto->sysctl_rmem_offset);
+
+	return *proto->sysctl_rmem;
+}
+
 #endif	/* _SOCK_H */
diff --git a/include/trace/events/sock.h b/include/trace/events/sock.h
index 6d31c0520ef36f89af825acebedfa5dabcaa6f6a..ec4dade2446607bd124133f22366c63e5caba852 100644
--- a/include/trace/events/sock.h
+++ b/include/trace/events/sock.h
@@ -48,7 +48,7 @@ TRACE_EVENT(sock_exceed_buf_limit,
 		strncpy(__entry->name, prot->name, 32);
 		__entry->sysctl_mem = prot->sysctl_mem;
 		__entry->allocated = allocated;
-		__entry->sysctl_rmem = prot->sysctl_rmem[0];
+		__entry->sysctl_rmem = sk_get_rmem0(sk, prot);
 		__entry->rmem_alloc = atomic_read(&sk->sk_rmem_alloc);
 	),
 
diff --git a/net/core/sock.c b/net/core/sock.c
index 759400053110fee3ca1180536ebbe739e614eab6..c59bcf90d90536fedc7809e397f6bd414781b529 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -2346,16 +2346,18 @@ int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind)
 
 	/* guarantee minimum buffer size under pressure */
 	if (kind == SK_MEM_RECV) {
-		if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
+		if (atomic_read(&sk->sk_rmem_alloc) < sk_get_rmem0(sk, prot))
 			return 1;
 
 	} else { /* SK_MEM_SEND */
+		int wmem0 = sk_get_wmem0(sk, prot);
+
 		if (sk->sk_type == SOCK_STREAM) {
-			if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
+			if (sk->sk_wmem_queued < wmem0)
 				return 1;
-		} else if (refcount_read(&sk->sk_wmem_alloc) <
-			   prot->sysctl_wmem[0])
+		} else if (refcount_read(&sk->sk_wmem_alloc) < wmem0) {
 				return 1;
+		}
 	}
 
 	if (sk_has_memory_pressure(sk)) {
-- 
2.15.0.403.gc27cc4dac6-goog

[PATCH net-next 2/2] tcp: Namespace-ify sysctl_tcp_rmem and sysctl_tcp_wmem

[Eric Dumazet]

Note that when a new netns is created, it inherits its
sysctl_tcp_rmem and sysctl_tcp_wmem from initial netns.

This change is needed so that we can refine TCP rcvbuf autotuning,
to take RTT into consideration.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Wei Wang <weiwan@google.com>
---
 include/net/netns/ipv4.h   |  2 ++
 include/net/tcp.h          |  2 --
 net/ipv4/sysctl_net_ipv4.c | 32 ++++++++++++++++----------------
 net/ipv4/tcp.c             | 21 ++++++++-------------
 net/ipv4/tcp_input.c       | 14 ++++++++------
 net/ipv4/tcp_ipv4.c        | 13 ++++++++++---
 net/ipv4/tcp_output.c      |  2 +-
 net/ipv6/tcp_ipv6.c        |  4 ++--
 8 files changed, 47 insertions(+), 43 deletions(-)

diff --git a/include/net/netns/ipv4.h b/include/net/netns/ipv4.h
index 379550f8124a1fca9987e644dbe00b7fa4bd9fdd..5e12975fc658b8671ad74678d25df1ada6fab1c8 100644
--- a/include/net/netns/ipv4.h
+++ b/include/net/netns/ipv4.h
@@ -155,6 +155,8 @@ struct netns_ipv4 {
 	int sysctl_tcp_invalid_ratelimit;
 	int sysctl_tcp_pacing_ss_ratio;
 	int sysctl_tcp_pacing_ca_ratio;
+	int sysctl_tcp_wmem[3];
+	int sysctl_tcp_rmem[3];
 	struct inet_timewait_death_row tcp_death_row;
 	int sysctl_max_syn_backlog;
 	int sysctl_tcp_fastopen;
diff --git a/include/net/tcp.h b/include/net/tcp.h
index babfd4da1515841f7cf02506736c2c4bb0006e7d..2f2c69ad31b28d7c479768b52bd1388220086f14 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -242,8 +242,6 @@ void tcp_time_wait(struct sock *sk, int state, int timeo);
 /* sysctl variables for tcp */
 extern int sysctl_tcp_max_orphans;
 extern long sysctl_tcp_mem[3];
-extern int sysctl_tcp_wmem[3];
-extern int sysctl_tcp_rmem[3];
 
 #define TCP_RACK_LOSS_DETECTION  0x1 /* Use RACK to detect losses */
 #define TCP_RACK_STATIC_REO_WND  0x2 /* Use static RACK reo wnd */
diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c
index a82b4403830816351f070fa47e2a7338bcc6abf5..ef0ff3357a44757d59f658614f3858687ef2ef47 100644
--- a/net/ipv4/sysctl_net_ipv4.c
+++ b/net/ipv4/sysctl_net_ipv4.c
@@ -440,22 +440,6 @@ static struct ctl_table ipv4_table[] = {
 		.mode		= 0644,
 		.proc_handler	= proc_doulongvec_minmax,
 	},
-	{
-		.procname	= "tcp_wmem",
-		.data		= &sysctl_tcp_wmem,
-		.maxlen		= sizeof(sysctl_tcp_wmem),
-		.mode		= 0644,
-		.proc_handler	= proc_dointvec_minmax,
-		.extra1		= &one,
-	},
-	{
-		.procname	= "tcp_rmem",
-		.data		= &sysctl_tcp_rmem,
-		.maxlen		= sizeof(sysctl_tcp_rmem),
-		.mode		= 0644,
-		.proc_handler	= proc_dointvec_minmax,
-		.extra1		= &one,
-	},
 	{
 		.procname	= "tcp_low_latency",
 		.data		= &sysctl_tcp_low_latency,
@@ -1164,6 +1148,22 @@ static struct ctl_table ipv4_net_table[] = {
 		.extra1		= &zero,
 		.extra2		= &thousand,
 	},
+	{
+		.procname	= "tcp_wmem",
+		.data		= &init_net.ipv4.sysctl_tcp_wmem,
+		.maxlen		= sizeof(init_net.ipv4.sysctl_tcp_wmem),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= &one,
+	},
+	{
+		.procname	= "tcp_rmem",
+		.data		= &init_net.ipv4.sysctl_tcp_rmem,
+		.maxlen		= sizeof(init_net.ipv4.sysctl_tcp_rmem),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= &one,
+	},
 	{ }
 };
 
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index c4cb19ed4628180ce627f9731def66feb6140baf..bc71a27d5ad98208a9b5f744b04d9abf9b0b9a6a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -289,12 +289,7 @@ struct percpu_counter tcp_orphan_count;
 EXPORT_SYMBOL_GPL(tcp_orphan_count);
 
 long sysctl_tcp_mem[3] __read_mostly;
-int sysctl_tcp_wmem[3] __read_mostly;
-int sysctl_tcp_rmem[3] __read_mostly;
-
 EXPORT_SYMBOL(sysctl_tcp_mem);
-EXPORT_SYMBOL(sysctl_tcp_rmem);
-EXPORT_SYMBOL(sysctl_tcp_wmem);
 
 atomic_long_t tcp_memory_allocated;	/* Current allocated memory. */
 EXPORT_SYMBOL(tcp_memory_allocated);
@@ -456,8 +451,8 @@ void tcp_init_sock(struct sock *sk)
 
 	icsk->icsk_sync_mss = tcp_sync_mss;
 
-	sk->sk_sndbuf = sysctl_tcp_wmem[1];
-	sk->sk_rcvbuf = sysctl_tcp_rmem[1];
+	sk->sk_sndbuf = sock_net(sk)->ipv4.sysctl_tcp_wmem[1];
+	sk->sk_rcvbuf = sock_net(sk)->ipv4.sysctl_tcp_rmem[1];
 
 	sk_sockets_allocated_inc(sk);
 }
@@ -3636,13 +3631,13 @@ void __init tcp_init(void)
 	max_wshare = min(4UL*1024*1024, limit);
 	max_rshare = min(6UL*1024*1024, limit);
 
-	sysctl_tcp_wmem[0] = SK_MEM_QUANTUM;
-	sysctl_tcp_wmem[1] = 16*1024;
-	sysctl_tcp_wmem[2] = max(64*1024, max_wshare);
+	init_net.ipv4.sysctl_tcp_wmem[0] = SK_MEM_QUANTUM;
+	init_net.ipv4.sysctl_tcp_wmem[1] = 16*1024;
+	init_net.ipv4.sysctl_tcp_wmem[2] = max(64*1024, max_wshare);
 
-	sysctl_tcp_rmem[0] = SK_MEM_QUANTUM;
-	sysctl_tcp_rmem[1] = 87380;
-	sysctl_tcp_rmem[2] = max(87380, max_rshare);
+	init_net.ipv4.sysctl_tcp_rmem[0] = SK_MEM_QUANTUM;
+	init_net.ipv4.sysctl_tcp_rmem[1] = 87380;
+	init_net.ipv4.sysctl_tcp_rmem[2] = max(87380, max_rshare);
 
 	pr_info("Hash tables configured (established %u bind %u)\n",
 		tcp_hashinfo.ehash_mask + 1, tcp_hashinfo.bhash_size);
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index 0ada8bfc2ebdf56b785d419edbe87a2ac3156389..e41123d854797910890dbd84900b5a7393089394 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -320,7 +320,7 @@ static void tcp_sndbuf_expand(struct sock *sk)
 	sndmem *= nr_segs * per_mss;
 
 	if (sk->sk_sndbuf < sndmem)
-		sk->sk_sndbuf = min(sndmem, sysctl_tcp_wmem[2]);
+		sk->sk_sndbuf = min(sndmem, sock_net(sk)->ipv4.sysctl_tcp_wmem[2]);
 }
 
 /* 2. Tuning advertised window (window_clamp, rcv_ssthresh)
@@ -354,7 +354,7 @@ static int __tcp_grow_window(const struct sock *sk, const struct sk_buff *skb)
 	struct tcp_sock *tp = tcp_sk(sk);
 	/* Optimize this! */
 	int truesize = tcp_win_from_space(sk, skb->truesize) >> 1;
-	int window = tcp_win_from_space(sk, sysctl_tcp_rmem[2]) >> 1;
+	int window = tcp_win_from_space(sk, sock_net(sk)->ipv4.sysctl_tcp_rmem[2]) >> 1;
 
 	while (tp->rcv_ssthresh <= window) {
 		if (truesize <= skb->len)
@@ -409,7 +409,7 @@ static void tcp_fixup_rcvbuf(struct sock *sk)
 		rcvmem <<= 2;
 
 	if (sk->sk_rcvbuf < rcvmem)
-		sk->sk_rcvbuf = min(rcvmem, sysctl_tcp_rmem[2]);
+		sk->sk_rcvbuf = min(rcvmem, sock_net(sk)->ipv4.sysctl_tcp_rmem[2]);
 }
 
 /* 4. Try to fixup all. It is made immediately after connection enters
@@ -457,15 +457,16 @@ static void tcp_clamp_window(struct sock *sk)
 {
 	struct tcp_sock *tp = tcp_sk(sk);
 	struct inet_connection_sock *icsk = inet_csk(sk);
+	struct net *net = sock_net(sk);
 
 	icsk->icsk_ack.quick = 0;
 
-	if (sk->sk_rcvbuf < sysctl_tcp_rmem[2] &&
+	if (sk->sk_rcvbuf < net->ipv4.sysctl_tcp_rmem[2] &&
 	    !(sk->sk_userlocks & SOCK_RCVBUF_LOCK) &&
 	    !tcp_under_memory_pressure(sk) &&
 	    sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)) {
 		sk->sk_rcvbuf = min(atomic_read(&sk->sk_rmem_alloc),
-				    sysctl_tcp_rmem[2]);
+				    net->ipv4.sysctl_tcp_rmem[2]);
 	}
 	if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf)
 		tp->rcv_ssthresh = min(tp->window_clamp, 2U * tp->advmss);
@@ -623,7 +624,8 @@ void tcp_rcv_space_adjust(struct sock *sk)
 		while (tcp_win_from_space(sk, rcvmem) < tp->advmss)
 			rcvmem += 128;
 
-		rcvbuf = min(rcvwin / tp->advmss * rcvmem, sysctl_tcp_rmem[2]);
+		rcvbuf = min(rcvwin / tp->advmss * rcvmem,
+			     sock_net(sk)->ipv4.sysctl_tcp_rmem[2]);
 		if (rcvbuf > sk->sk_rcvbuf) {
 			sk->sk_rcvbuf = rcvbuf;
 
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 0162c577bb9cf85160ed833e3e069daf66a2cd15..1eac84b8044e70fe99c6d3c4d8c5319392141b40 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -2409,8 +2409,8 @@ struct proto tcp_prot = {
 	.memory_allocated	= &tcp_memory_allocated,
 	.memory_pressure	= &tcp_memory_pressure,
 	.sysctl_mem		= sysctl_tcp_mem,
-	.sysctl_wmem		= sysctl_tcp_wmem,
-	.sysctl_rmem		= sysctl_tcp_rmem,
+	.sysctl_wmem_offset	= offsetof(struct net, ipv4.sysctl_tcp_wmem),
+	.sysctl_rmem_offset	= offsetof(struct net, ipv4.sysctl_tcp_rmem),
 	.max_header		= MAX_TCP_HEADER,
 	.obj_size		= sizeof(struct tcp_sock),
 	.slab_flags		= SLAB_TYPESAFE_BY_RCU,
@@ -2509,7 +2509,14 @@ static int __net_init tcp_sk_init(struct net *net)
 	net->ipv4.sysctl_tcp_invalid_ratelimit = HZ/2;
 	net->ipv4.sysctl_tcp_pacing_ss_ratio = 200;
 	net->ipv4.sysctl_tcp_pacing_ca_ratio = 120;
-
+	if (net != &init_net) {
+		memcpy(net->ipv4.sysctl_tcp_rmem,
+		       init_net.ipv4.sysctl_tcp_rmem,
+		       sizeof(init_net.ipv4.sysctl_tcp_rmem));
+		memcpy(net->ipv4.sysctl_tcp_wmem,
+		       init_net.ipv4.sysctl_tcp_wmem,
+		       sizeof(init_net.ipv4.sysctl_tcp_wmem));
+	}
 	net->ipv4.sysctl_tcp_fastopen = TFO_CLIENT_ENABLE;
 	spin_lock_init(&net->ipv4.tcp_fastopen_ctx_lock);
 	net->ipv4.sysctl_tcp_fastopen_blackhole_timeout = 60 * 60;
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index a9d917e4dad5c6fec1dc81e96c16b78223786d4c..9b98d35aa0d8d0a829e4a41985d805d4e2895a8e 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -220,7 +220,7 @@ void tcp_select_initial_window(const struct sock *sk, int __space, __u32 mss,
 	(*rcv_wscale) = 0;
 	if (wscale_ok) {
 		/* Set window scaling on max possible window */
-		space = max_t(u32, space, sysctl_tcp_rmem[2]);
+		space = max_t(u32, space, sock_net(sk)->ipv4.sysctl_tcp_rmem[2]);
 		space = max_t(u32, space, sysctl_rmem_max);
 		space = min_t(u32, space, *window_clamp);
 		while (space > U16_MAX && (*rcv_wscale) < TCP_MAX_WSCALE) {
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index 0e2529958b52c422ac9ca66186dae17d69628f0b..6bb98c93edfe2ed2f16fe5229605f8108cfc7f9a 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -1940,8 +1940,8 @@ struct proto tcpv6_prot = {
 	.memory_pressure	= &tcp_memory_pressure,
 	.orphan_count		= &tcp_orphan_count,
 	.sysctl_mem		= sysctl_tcp_mem,
-	.sysctl_wmem		= sysctl_tcp_wmem,
-	.sysctl_rmem		= sysctl_tcp_rmem,
+	.sysctl_wmem_offset	= offsetof(struct net, ipv4.sysctl_tcp_wmem),
+	.sysctl_rmem_offset	= offsetof(struct net, ipv4.sysctl_tcp_rmem),
 	.max_header		= MAX_TCP_HEADER,
 	.obj_size		= sizeof(struct tcp6_sock),
 	.slab_flags		= SLAB_TYPESAFE_BY_RCU,
-- 
2.15.0.403.gc27cc4dac6-goog

Re: [PATCH net-next 0/2] net: Namespace-ify sysctl_tcp_rmem and sysctl_tcp_wmem

[David Miller]

From: Eric Dumazet <edumazet@google.com>
Date: Tue,  7 Nov 2017 00:29:26 -0800

> We need to get per netns sysctl for sysctl_[proto]_rmem and
> sysctl_[proto]_wmem
> 
> This patch series adds the basic infrastructure allowing per proto
> conversion, and takes care of TCP.

Series applied, thanks Eric.