[PATCH net-next 0/2] tcp: extend access to CC information

[PATCH net-next 0/2] tcp: extend access to CC information

[Eric Dumazet]

Instead of being forced to use netlink, we want to get (optional) CC
information using a getsockopt() on the socket file handle.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>

Eric Dumazet (2):
  tcp: prepare CC get_info() access from getsockopt()
  tcp: add TCP_CC_INFO socket option

 include/net/tcp.h              |  5 ++++-
 include/uapi/linux/inet_diag.h |  4 ++++
 include/uapi/linux/tcp.h       |  1 +
 net/ipv4/inet_diag.c           |  8 +++++---
 net/ipv4/tcp.c                 | 21 +++++++++++++++++++++
 net/ipv4/tcp_dctcp.c           | 20 ++++++++++----------
 net/ipv4/tcp_illinois.c        | 21 +++++++++++----------
 net/ipv4/tcp_vegas.c           | 19 ++++++++++---------
 net/ipv4/tcp_vegas.h           |  3 ++-
 9 files changed, 68 insertions(+), 34 deletions(-)

-- 
2.2.0.rc0.207.ga3a616c

[PATCH net-next 1/2] tcp: prepare CC get_info() access from getsockopt()

[Eric Dumazet]

We would like that optional info provided by Congestion Control
modules using netlink can also be read using getsockopt()

This patch changes get_info() to put this information in a buffer,
instead of skb, like tcp_get_info(), so that following patch
can reuse this common infrastructure.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
---
 include/net/tcp.h              |  5 ++++-
 include/uapi/linux/inet_diag.h |  4 ++++
 net/ipv4/inet_diag.c           |  8 +++++---
 net/ipv4/tcp_dctcp.c           | 20 ++++++++++----------
 net/ipv4/tcp_illinois.c        | 21 +++++++++++----------
 net/ipv4/tcp_vegas.c           | 19 ++++++++++---------
 net/ipv4/tcp_vegas.h           |  3 ++-
 7 files changed, 46 insertions(+), 34 deletions(-)

diff --git a/include/net/tcp.h b/include/net/tcp.h
index 051dc5c2802d..d5b3089317fb 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -804,6 +804,8 @@ enum tcp_ca_ack_event_flags {
 /* Requires ECN/ECT set on all packets */
 #define TCP_CONG_NEEDS_ECN	0x2
 
+union tcp_cc_info;
+
 struct tcp_congestion_ops {
 	struct list_head	list;
 	u32 key;
@@ -829,7 +831,8 @@ struct tcp_congestion_ops {
 	/* hook for packet ack accounting (optional) */
 	void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
 	/* get info for inet_diag (optional) */
-	int (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
+	size_t (*get_info)(struct sock *sk, u32 ext, int *attr,
+			   union tcp_cc_info *info);
 
 	char 		name[TCP_CA_NAME_MAX];
 	struct module 	*owner;
diff --git a/include/uapi/linux/inet_diag.h b/include/uapi/linux/inet_diag.h
index d65c0a09efd3..c7093c75bdd6 100644
--- a/include/uapi/linux/inet_diag.h
+++ b/include/uapi/linux/inet_diag.h
@@ -143,4 +143,8 @@ struct tcp_dctcp_info {
 	__u32	dctcp_ab_tot;
 };
 
+union tcp_cc_info {
+	struct tcpvegas_info	vegas;
+	struct tcp_dctcp_info	dctcp;
+};
 #endif /* _UAPI_INET_DIAG_H_ */
diff --git a/net/ipv4/inet_diag.c b/net/ipv4/inet_diag.c
index bb77ebdae3b3..4d32262c7502 100644
--- a/net/ipv4/inet_diag.c
+++ b/net/ipv4/inet_diag.c
@@ -224,14 +224,16 @@ int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk,
 	handler->idiag_get_info(sk, r, info);
 
 	if (sk->sk_state < TCP_TIME_WAIT) {
-		int err = 0;
+		union tcp_cc_info info;
+		size_t sz = 0;
+		int attr;
 
 		rcu_read_lock();
 		ca_ops = READ_ONCE(icsk->icsk_ca_ops);
 		if (ca_ops && ca_ops->get_info)
-			err = ca_ops->get_info(sk, ext, skb);
+			sz = ca_ops->get_info(sk, ext, &attr, &info);
 		rcu_read_unlock();
-		if (err < 0)
+		if (sz && nla_put(skb, attr, sz, &info) < 0)
 			goto errout;
 	}
 
diff --git a/net/ipv4/tcp_dctcp.c b/net/ipv4/tcp_dctcp.c
index 4376016f7fa5..4c41c1287197 100644
--- a/net/ipv4/tcp_dctcp.c
+++ b/net/ipv4/tcp_dctcp.c
@@ -277,7 +277,8 @@ static void dctcp_cwnd_event(struct sock *sk, enum tcp_ca_event ev)
 	}
 }
 
-static int dctcp_get_info(struct sock *sk, u32 ext, struct sk_buff *skb)
+static size_t dctcp_get_info(struct sock *sk, u32 ext, int *attr,
+			     union tcp_cc_info *info)
 {
 	const struct dctcp *ca = inet_csk_ca(sk);
 
@@ -286,18 +287,17 @@ static int dctcp_get_info(struct sock *sk, u32 ext, struct sk_buff *skb)
 	 */
 	if (ext & (1 << (INET_DIAG_DCTCPINFO - 1)) ||
 	    ext & (1 << (INET_DIAG_VEGASINFO - 1))) {
-		struct tcp_dctcp_info info;
-
-		memset(&info, 0, sizeof(info));
+		memset(info, 0, sizeof(struct tcp_dctcp_info));
 		if (inet_csk(sk)->icsk_ca_ops != &dctcp_reno) {
-			info.dctcp_enabled = 1;
-			info.dctcp_ce_state = (u16) ca->ce_state;
-			info.dctcp_alpha = ca->dctcp_alpha;
-			info.dctcp_ab_ecn = ca->acked_bytes_ecn;
-			info.dctcp_ab_tot = ca->acked_bytes_total;
+			info->dctcp.dctcp_enabled = 1;
+			info->dctcp.dctcp_ce_state = (u16) ca->ce_state;
+			info->dctcp.dctcp_alpha = ca->dctcp_alpha;
+			info->dctcp.dctcp_ab_ecn = ca->acked_bytes_ecn;
+			info->dctcp.dctcp_ab_tot = ca->acked_bytes_total;
 		}
 
-		return nla_put(skb, INET_DIAG_DCTCPINFO, sizeof(info), &info);
+		*attr = INET_DIAG_DCTCPINFO;
+		return sizeof(*info);
 	}
 	return 0;
 }
diff --git a/net/ipv4/tcp_illinois.c b/net/ipv4/tcp_illinois.c
index 67476f085e48..f71002e4db0b 100644
--- a/net/ipv4/tcp_illinois.c
+++ b/net/ipv4/tcp_illinois.c
@@ -300,24 +300,25 @@ static u32 tcp_illinois_ssthresh(struct sock *sk)
 }
 
 /* Extract info for Tcp socket info provided via netlink. */
-static int tcp_illinois_info(struct sock *sk, u32 ext, struct sk_buff *skb)
+static size_t tcp_illinois_info(struct sock *sk, u32 ext, int *attr,
+				union tcp_cc_info *info)
 {
 	const struct illinois *ca = inet_csk_ca(sk);
 
 	if (ext & (1 << (INET_DIAG_VEGASINFO - 1))) {
-		struct tcpvegas_info info = {
-			.tcpv_enabled = 1,
-			.tcpv_rttcnt = ca->cnt_rtt,
-			.tcpv_minrtt = ca->base_rtt,
-		};
+		info->vegas.tcpv_enabled = 1;
+		info->vegas.tcpv_rttcnt = ca->cnt_rtt;
+		info->vegas.tcpv_minrtt = ca->base_rtt;
+		info->vegas.tcpv_rtt = 0;
 
-		if (info.tcpv_rttcnt > 0) {
+		if (info->vegas.tcpv_rttcnt > 0) {
 			u64 t = ca->sum_rtt;
 
-			do_div(t, info.tcpv_rttcnt);
-			info.tcpv_rtt = t;
+			do_div(t, info->vegas.tcpv_rttcnt);
+			info->vegas.tcpv_rtt = t;
 		}
-		return nla_put(skb, INET_DIAG_VEGASINFO, sizeof(info), &info);
+		*attr = INET_DIAG_VEGASINFO;
+		return sizeof(struct tcpvegas_info);
 	}
 	return 0;
 }
diff --git a/net/ipv4/tcp_vegas.c b/net/ipv4/tcp_vegas.c
index c71a1b8f7bde..a6cea1d5e20d 100644
--- a/net/ipv4/tcp_vegas.c
+++ b/net/ipv4/tcp_vegas.c
@@ -286,18 +286,19 @@ static void tcp_vegas_cong_avoid(struct sock *sk, u32 ack, u32 acked)
 }
 
 /* Extract info for Tcp socket info provided via netlink. */
-int tcp_vegas_get_info(struct sock *sk, u32 ext, struct sk_buff *skb)
+size_t tcp_vegas_get_info(struct sock *sk, u32 ext, int *attr,
+			  union tcp_cc_info *info)
 {
 	const struct vegas *ca = inet_csk_ca(sk);
+
 	if (ext & (1 << (INET_DIAG_VEGASINFO - 1))) {
-		struct tcpvegas_info info = {
-			.tcpv_enabled = ca->doing_vegas_now,
-			.tcpv_rttcnt = ca->cntRTT,
-			.tcpv_rtt = ca->baseRTT,
-			.tcpv_minrtt = ca->minRTT,
-		};
-
-		return nla_put(skb, INET_DIAG_VEGASINFO, sizeof(info), &info);
+		info->vegas.tcpv_enabled = ca->doing_vegas_now,
+		info->vegas.tcpv_rttcnt = ca->cntRTT,
+		info->vegas.tcpv_rtt = ca->baseRTT,
+		info->vegas.tcpv_minrtt = ca->minRTT,
+
+		*attr = INET_DIAG_VEGASINFO;
+		return sizeof(struct tcpvegas_info);
 	}
 	return 0;
 }
diff --git a/net/ipv4/tcp_vegas.h b/net/ipv4/tcp_vegas.h
index e8a6b33cc61d..ef9da5306c68 100644
--- a/net/ipv4/tcp_vegas.h
+++ b/net/ipv4/tcp_vegas.h
@@ -19,6 +19,7 @@ void tcp_vegas_init(struct sock *sk);
 void tcp_vegas_state(struct sock *sk, u8 ca_state);
 void tcp_vegas_pkts_acked(struct sock *sk, u32 cnt, s32 rtt_us);
 void tcp_vegas_cwnd_event(struct sock *sk, enum tcp_ca_event event);
-int tcp_vegas_get_info(struct sock *sk, u32 ext, struct sk_buff *skb);
+size_t tcp_vegas_get_info(struct sock *sk, u32 ext, int *attr,
+			  union tcp_cc_info *info);
 
 #endif	/* __TCP_VEGAS_H */
-- 
2.2.0.rc0.207.ga3a616c

[PATCH net-next 2/2] tcp: add TCP_CC_INFO socket option

[Eric Dumazet]

Some Congestion Control modules can provide per flow information,
but current way to get this information is to use netlink.

Like TCP_INFO, let's add TCP_CC_INFO so that applications can
issue a getsockopt() if they have a socket file descriptor,
instead of playing complex netlink games.

Sample usage would be :

  union tcp_cc_info info;
  socklen_t len = sizeof(info);

  if (getsockopt(fd, SOL_TCP, TCP_CC_INFO, &info, &len) == -1)

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
---
 include/uapi/linux/tcp.h |  1 +
 net/ipv4/tcp.c           | 21 +++++++++++++++++++++
 2 files changed, 22 insertions(+)

diff --git a/include/uapi/linux/tcp.h b/include/uapi/linux/tcp.h
index 3b9718328d8b..937ec387276f 100644
--- a/include/uapi/linux/tcp.h
+++ b/include/uapi/linux/tcp.h
@@ -112,6 +112,7 @@ enum {
 #define TCP_FASTOPEN		23	/* Enable FastOpen on listeners */
 #define TCP_TIMESTAMP		24
 #define TCP_NOTSENT_LOWAT	25	/* limit number of unsent bytes in write queue */
+#define TCP_CC_INFO		26	/* Get Congestion Control (optional) info */
 
 struct tcp_repair_opt {
 	__u32	opt_code;
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 8c5cd9efebbc..1e06c75a6837 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -252,6 +252,7 @@
 #include <linux/types.h>
 #include <linux/fcntl.h>
 #include <linux/poll.h>
+#include <linux/inet_diag.h>
 #include <linux/init.h>
 #include <linux/fs.h>
 #include <linux/skbuff.h>
@@ -2734,6 +2735,26 @@ static int do_tcp_getsockopt(struct sock *sk, int level,
 			return -EFAULT;
 		return 0;
 	}
+	case TCP_CC_INFO: {
+		const struct tcp_congestion_ops *ca_ops;
+		union tcp_cc_info info;
+		size_t sz = 0;
+		int attr;
+
+		if (get_user(len, optlen))
+			return -EFAULT;
+
+		ca_ops = icsk->icsk_ca_ops;
+		if (ca_ops && ca_ops->get_info)
+			sz = ca_ops->get_info(sk, ~0U, &attr, &info);
+
+		len = min_t(unsigned int, len, sz);
+		if (put_user(len, optlen))
+			return -EFAULT;
+		if (copy_to_user(optval, &info, len))
+			return -EFAULT;
+		return 0;
+	}
 	case TCP_QUICKACK:
 		val = !icsk->icsk_ack.pingpong;
 		break;
-- 
2.2.0.rc0.207.ga3a616c

Re: [PATCH net-next 1/2] tcp: prepare CC get_info() access from getsockopt()

[Neal Cardwell]

On Tue, Apr 28, 2015 at 7:23 PM, Eric Dumazet <edumazet@google.com> wrote:
> We would like that optional info provided by Congestion Control
> modules using netlink can also be read using getsockopt()
>
> This patch changes get_info() to put this information in a buffer,
> instead of skb, like tcp_get_info(), so that following patch
> can reuse this common infrastructure.
>
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Cc: Yuchung Cheng <ycheng@google.com>
> Cc: Neal Cardwell <ncardwell@google.com>
> ---

Acked-by: Neal Cardwell <ncardwell@google.com>

Thanks, Eric!

neal

Re: [PATCH net-next 2/2] tcp: add TCP_CC_INFO socket option

[Neal Cardwell]

On Tue, Apr 28, 2015 at 7:23 PM, Eric Dumazet <edumazet@google.com> wrote:
> Some Congestion Control modules can provide per flow information,
> but current way to get this information is to use netlink.
>
> Like TCP_INFO, let's add TCP_CC_INFO so that applications can
> issue a getsockopt() if they have a socket file descriptor,
> instead of playing complex netlink games.
>
> Sample usage would be :
>
>   union tcp_cc_info info;
>   socklen_t len = sizeof(info);
>
>   if (getsockopt(fd, SOL_TCP, TCP_CC_INFO, &info, &len) == -1)
>
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Cc: Yuchung Cheng <ycheng@google.com>
> Cc: Neal Cardwell <ncardwell@google.com>
> ---

Acked-by: Neal Cardwell <ncardwell@google.com>

A very useful facility. Thanks for putting this together, Eric!

neal

Re: [PATCH net-next 1/2] tcp: prepare CC get_info() access from getsockopt()

[Daniel Borkmann]

On 04/29/2015 01:23 AM, Eric Dumazet wrote:
> We would like that optional info provided by Congestion Control
> modules using netlink can also be read using getsockopt()
>
> This patch changes get_info() to put this information in a buffer,
> instead of skb, like tcp_get_info(), so that following patch
> can reuse this common infrastructure.
>
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Cc: Yuchung Cheng <ycheng@google.com>
> Cc: Neal Cardwell <ncardwell@google.com>

Looks good to me, thanks!

Acked-by: Daniel Borkmann <daniel@iogearbox.net>

Re: [PATCH net-next 2/2] tcp: add TCP_CC_INFO socket option

[Daniel Borkmann]

On 04/29/2015 01:23 AM, Eric Dumazet wrote:
> Some Congestion Control modules can provide per flow information,
> but current way to get this information is to use netlink.
>
> Like TCP_INFO, let's add TCP_CC_INFO so that applications can
> issue a getsockopt() if they have a socket file descriptor,
> instead of playing complex netlink games.
>
> Sample usage would be :
>
>    union tcp_cc_info info;
>    socklen_t len = sizeof(info);
>
>    if (getsockopt(fd, SOL_TCP, TCP_CC_INFO, &info, &len) == -1)
>
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Cc: Yuchung Cheng <ycheng@google.com>
> Cc: Neal Cardwell <ncardwell@google.com>

Presuming other cc algorithms would in future also export
internal information through this interface, would it make
sense to put tcp_cc_info into a container structure so we
don't miss out attr (vegas, dctcp, ...), like:

   struct tcp_cc_exp {
     u32               kind;
     union tcp_cc_info info;
   };

Otherwise looks good:

Acked-by: Daniel Borkmann <daniel@iogearbox.net>

Re: [PATCH net-next 2/2] tcp: add TCP_CC_INFO socket option

[Eric Dumazet]

On Wed, 2015-04-29 at 10:17 +0200, Daniel Borkmann wrote:
> On 04/29/2015 01:23 AM, Eric Dumazet wrote:
> > Some Congestion Control modules can provide per flow information,
> > but current way to get this information is to use netlink.
> >
> > Like TCP_INFO, let's add TCP_CC_INFO so that applications can
> > issue a getsockopt() if they have a socket file descriptor,
> > instead of playing complex netlink games.
> >
> > Sample usage would be :
> >
> >    union tcp_cc_info info;
> >    socklen_t len = sizeof(info);
> >
> >    if (getsockopt(fd, SOL_TCP, TCP_CC_INFO, &info, &len) == -1)
> >
> > Signed-off-by: Eric Dumazet <edumazet@google.com>
> > Cc: Yuchung Cheng <ycheng@google.com>
> > Cc: Neal Cardwell <ncardwell@google.com>
> 
> Presuming other cc algorithms would in future also export
> internal information through this interface, would it make
> sense to put tcp_cc_info into a container structure so we
> don't miss out attr (vegas, dctcp, ...), like:
> 
>    struct tcp_cc_exp {
>      u32               kind;
>      union tcp_cc_info info;
>    };
> 
> Otherwise looks good:
> 
> Acked-by: Daniel Borkmann <daniel@iogearbox.net>

I thought of this, but I really believe this is not needed, as the
application can already fetch CC name (if really its does not know yet)

And I also wanted to get same layout for info provided by netlink and
getsockopt()

Thanks !

Re: [PATCH net-next 2/2] tcp: add TCP_CC_INFO socket option

[Daniel Borkmann]

On 04/29/2015 01:07 PM, Eric Dumazet wrote:
> On Wed, 2015-04-29 at 10:17 +0200, Daniel Borkmann wrote:
>> On 04/29/2015 01:23 AM, Eric Dumazet wrote:
>>> Some Congestion Control modules can provide per flow information,
>>> but current way to get this information is to use netlink.
>>>
>>> Like TCP_INFO, let's add TCP_CC_INFO so that applications can
>>> issue a getsockopt() if they have a socket file descriptor,
>>> instead of playing complex netlink games.
>>>
>>> Sample usage would be :
>>>
>>>     union tcp_cc_info info;
>>>     socklen_t len = sizeof(info);
>>>
>>>     if (getsockopt(fd, SOL_TCP, TCP_CC_INFO, &info, &len) == -1)
>>>
>>> Signed-off-by: Eric Dumazet <edumazet@google.com>
>>> Cc: Yuchung Cheng <ycheng@google.com>
>>> Cc: Neal Cardwell <ncardwell@google.com>
>>
>> Presuming other cc algorithms would in future also export
>> internal information through this interface, would it make
>> sense to put tcp_cc_info into a container structure so we
>> don't miss out attr (vegas, dctcp, ...), like:
>>
>>     struct tcp_cc_exp {
>>       u32               kind;
>>       union tcp_cc_info info;
>>     };
>>
>> Otherwise looks good:
>>
>> Acked-by: Daniel Borkmann <daniel@iogearbox.net>
>
> I thought of this, but I really believe this is not needed, as the
> application can already fetch CC name (if really its does not know yet)
>
> And I also wanted to get same layout for info provided by netlink and
> getsockopt()

Ok, I'm fine with that. Presumably, applications making use of
this facility would most likely set the cc op themselves from
the application, so they know exactly what to expect here. Or,
might have fetched it via the cc name albeit a bit inconvenient,
but not impossible.

Thanks,
Daniel

Re: [PATCH net-next 1/2] tcp: prepare CC get_info() access from getsockopt()

[Yuchung Cheng]

On Tue, Apr 28, 2015 at 4:23 PM, Eric Dumazet <edumazet@google.com> wrote:
> We would like that optional info provided by Congestion Control
> modules using netlink can also be read using getsockopt()
>
> This patch changes get_info() to put this information in a buffer,
> instead of skb, like tcp_get_info(), so that following patch
> can reuse this common infrastructure.
>
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Cc: Yuchung Cheng <ycheng@google.com>
> Cc: Neal Cardwell <ncardwell@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>

> ---
>  include/net/tcp.h              |  5 ++++-
>  include/uapi/linux/inet_diag.h |  4 ++++
>  net/ipv4/inet_diag.c           |  8 +++++---
>  net/ipv4/tcp_dctcp.c           | 20 ++++++++++----------
>  net/ipv4/tcp_illinois.c        | 21 +++++++++++----------
>  net/ipv4/tcp_vegas.c           | 19 ++++++++++---------
>  net/ipv4/tcp_vegas.h           |  3 ++-
>  7 files changed, 46 insertions(+), 34 deletions(-)
>
> diff --git a/include/net/tcp.h b/include/net/tcp.h
> index 051dc5c2802d..d5b3089317fb 100644
> --- a/include/net/tcp.h
> +++ b/include/net/tcp.h
> @@ -804,6 +804,8 @@ enum tcp_ca_ack_event_flags {
>  /* Requires ECN/ECT set on all packets */
>  #define TCP_CONG_NEEDS_ECN     0x2
>
> +union tcp_cc_info;
> +
>  struct tcp_congestion_ops {
>         struct list_head        list;
>         u32 key;
> @@ -829,7 +831,8 @@ struct tcp_congestion_ops {
>         /* hook for packet ack accounting (optional) */
>         void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
>         /* get info for inet_diag (optional) */
> -       int (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
> +       size_t (*get_info)(struct sock *sk, u32 ext, int *attr,
> +                          union tcp_cc_info *info);
>
>         char            name[TCP_CA_NAME_MAX];
>         struct module   *owner;
> diff --git a/include/uapi/linux/inet_diag.h b/include/uapi/linux/inet_diag.h
> index d65c0a09efd3..c7093c75bdd6 100644
> --- a/include/uapi/linux/inet_diag.h
> +++ b/include/uapi/linux/inet_diag.h
> @@ -143,4 +143,8 @@ struct tcp_dctcp_info {
>         __u32   dctcp_ab_tot;
>  };
>
> +union tcp_cc_info {
> +       struct tcpvegas_info    vegas;
> +       struct tcp_dctcp_info   dctcp;
> +};
>  #endif /* _UAPI_INET_DIAG_H_ */
> diff --git a/net/ipv4/inet_diag.c b/net/ipv4/inet_diag.c
> index bb77ebdae3b3..4d32262c7502 100644
> --- a/net/ipv4/inet_diag.c
> +++ b/net/ipv4/inet_diag.c
> @@ -224,14 +224,16 @@ int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk,
>         handler->idiag_get_info(sk, r, info);
>
>         if (sk->sk_state < TCP_TIME_WAIT) {
> -               int err = 0;
> +               union tcp_cc_info info;
> +               size_t sz = 0;
> +               int attr;
>
>                 rcu_read_lock();
>                 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
>                 if (ca_ops && ca_ops->get_info)
> -                       err = ca_ops->get_info(sk, ext, skb);
> +                       sz = ca_ops->get_info(sk, ext, &attr, &info);
>                 rcu_read_unlock();
> -               if (err < 0)
> +               if (sz && nla_put(skb, attr, sz, &info) < 0)
>                         goto errout;
>         }
>
> diff --git a/net/ipv4/tcp_dctcp.c b/net/ipv4/tcp_dctcp.c
> index 4376016f7fa5..4c41c1287197 100644
> --- a/net/ipv4/tcp_dctcp.c
> +++ b/net/ipv4/tcp_dctcp.c
> @@ -277,7 +277,8 @@ static void dctcp_cwnd_event(struct sock *sk, enum tcp_ca_event ev)
>         }
>  }
>
> -static int dctcp_get_info(struct sock *sk, u32 ext, struct sk_buff *skb)
> +static size_t dctcp_get_info(struct sock *sk, u32 ext, int *attr,
> +                            union tcp_cc_info *info)
>  {
>         const struct dctcp *ca = inet_csk_ca(sk);
>
> @@ -286,18 +287,17 @@ static int dctcp_get_info(struct sock *sk, u32 ext, struct sk_buff *skb)
>          */
>         if (ext & (1 << (INET_DIAG_DCTCPINFO - 1)) ||
>             ext & (1 << (INET_DIAG_VEGASINFO - 1))) {
> -               struct tcp_dctcp_info info;
> -
> -               memset(&info, 0, sizeof(info));
> +               memset(info, 0, sizeof(struct tcp_dctcp_info));
>                 if (inet_csk(sk)->icsk_ca_ops != &dctcp_reno) {
> -                       info.dctcp_enabled = 1;
> -                       info.dctcp_ce_state = (u16) ca->ce_state;
> -                       info.dctcp_alpha = ca->dctcp_alpha;
> -                       info.dctcp_ab_ecn = ca->acked_bytes_ecn;
> -                       info.dctcp_ab_tot = ca->acked_bytes_total;
> +                       info->dctcp.dctcp_enabled = 1;
> +                       info->dctcp.dctcp_ce_state = (u16) ca->ce_state;
> +                       info->dctcp.dctcp_alpha = ca->dctcp_alpha;
> +                       info->dctcp.dctcp_ab_ecn = ca->acked_bytes_ecn;
> +                       info->dctcp.dctcp_ab_tot = ca->acked_bytes_total;
>                 }
>
> -               return nla_put(skb, INET_DIAG_DCTCPINFO, sizeof(info), &info);
> +               *attr = INET_DIAG_DCTCPINFO;
> +               return sizeof(*info);
>         }
>         return 0;
>  }
> diff --git a/net/ipv4/tcp_illinois.c b/net/ipv4/tcp_illinois.c
> index 67476f085e48..f71002e4db0b 100644
> --- a/net/ipv4/tcp_illinois.c
> +++ b/net/ipv4/tcp_illinois.c
> @@ -300,24 +300,25 @@ static u32 tcp_illinois_ssthresh(struct sock *sk)
>  }
>
>  /* Extract info for Tcp socket info provided via netlink. */
> -static int tcp_illinois_info(struct sock *sk, u32 ext, struct sk_buff *skb)
> +static size_t tcp_illinois_info(struct sock *sk, u32 ext, int *attr,
> +                               union tcp_cc_info *info)
>  {
>         const struct illinois *ca = inet_csk_ca(sk);
>
>         if (ext & (1 << (INET_DIAG_VEGASINFO - 1))) {
> -               struct tcpvegas_info info = {
> -                       .tcpv_enabled = 1,
> -                       .tcpv_rttcnt = ca->cnt_rtt,
> -                       .tcpv_minrtt = ca->base_rtt,
> -               };
> +               info->vegas.tcpv_enabled = 1;
> +               info->vegas.tcpv_rttcnt = ca->cnt_rtt;
> +               info->vegas.tcpv_minrtt = ca->base_rtt;
> +               info->vegas.tcpv_rtt = 0;
>
> -               if (info.tcpv_rttcnt > 0) {
> +               if (info->vegas.tcpv_rttcnt > 0) {
>                         u64 t = ca->sum_rtt;
>
> -                       do_div(t, info.tcpv_rttcnt);
> -                       info.tcpv_rtt = t;
> +                       do_div(t, info->vegas.tcpv_rttcnt);
> +                       info->vegas.tcpv_rtt = t;
>                 }
> -               return nla_put(skb, INET_DIAG_VEGASINFO, sizeof(info), &info);
> +               *attr = INET_DIAG_VEGASINFO;
> +               return sizeof(struct tcpvegas_info);
>         }
>         return 0;
>  }
> diff --git a/net/ipv4/tcp_vegas.c b/net/ipv4/tcp_vegas.c
> index c71a1b8f7bde..a6cea1d5e20d 100644
> --- a/net/ipv4/tcp_vegas.c
> +++ b/net/ipv4/tcp_vegas.c
> @@ -286,18 +286,19 @@ static void tcp_vegas_cong_avoid(struct sock *sk, u32 ack, u32 acked)
>  }
>
>  /* Extract info for Tcp socket info provided via netlink. */
> -int tcp_vegas_get_info(struct sock *sk, u32 ext, struct sk_buff *skb)
> +size_t tcp_vegas_get_info(struct sock *sk, u32 ext, int *attr,
> +                         union tcp_cc_info *info)
>  {
>         const struct vegas *ca = inet_csk_ca(sk);
> +
>         if (ext & (1 << (INET_DIAG_VEGASINFO - 1))) {
> -               struct tcpvegas_info info = {
> -                       .tcpv_enabled = ca->doing_vegas_now,
> -                       .tcpv_rttcnt = ca->cntRTT,
> -                       .tcpv_rtt = ca->baseRTT,
> -                       .tcpv_minrtt = ca->minRTT,
> -               };
> -
> -               return nla_put(skb, INET_DIAG_VEGASINFO, sizeof(info), &info);
> +               info->vegas.tcpv_enabled = ca->doing_vegas_now,
> +               info->vegas.tcpv_rttcnt = ca->cntRTT,
> +               info->vegas.tcpv_rtt = ca->baseRTT,
> +               info->vegas.tcpv_minrtt = ca->minRTT,
> +
> +               *attr = INET_DIAG_VEGASINFO;
> +               return sizeof(struct tcpvegas_info);
>         }
>         return 0;
>  }
> diff --git a/net/ipv4/tcp_vegas.h b/net/ipv4/tcp_vegas.h
> index e8a6b33cc61d..ef9da5306c68 100644
> --- a/net/ipv4/tcp_vegas.h
> +++ b/net/ipv4/tcp_vegas.h
> @@ -19,6 +19,7 @@ void tcp_vegas_init(struct sock *sk);
>  void tcp_vegas_state(struct sock *sk, u8 ca_state);
>  void tcp_vegas_pkts_acked(struct sock *sk, u32 cnt, s32 rtt_us);
>  void tcp_vegas_cwnd_event(struct sock *sk, enum tcp_ca_event event);
> -int tcp_vegas_get_info(struct sock *sk, u32 ext, struct sk_buff *skb);
> +size_t tcp_vegas_get_info(struct sock *sk, u32 ext, int *attr,
> +                         union tcp_cc_info *info);
>
>  #endif /* __TCP_VEGAS_H */
> --
> 2.2.0.rc0.207.ga3a616c
>

Re: [PATCH net-next 2/2] tcp: add TCP_CC_INFO socket option

[Yuchung Cheng]

On Tue, Apr 28, 2015 at 4:23 PM, Eric Dumazet <edumazet@google.com> wrote:
> Some Congestion Control modules can provide per flow information,
> but current way to get this information is to use netlink.
>
> Like TCP_INFO, let's add TCP_CC_INFO so that applications can
> issue a getsockopt() if they have a socket file descriptor,
> instead of playing complex netlink games.
>
> Sample usage would be :
>
>   union tcp_cc_info info;
>   socklen_t len = sizeof(info);
>
>   if (getsockopt(fd, SOL_TCP, TCP_CC_INFO, &info, &len) == -1)
>
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Cc: Yuchung Cheng <ycheng@google.com>
> Cc: Neal Cardwell <ncardwell@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>

Nice extension!

> ---
>  include/uapi/linux/tcp.h |  1 +
>  net/ipv4/tcp.c           | 21 +++++++++++++++++++++
>  2 files changed, 22 insertions(+)
>
> diff --git a/include/uapi/linux/tcp.h b/include/uapi/linux/tcp.h
> index 3b9718328d8b..937ec387276f 100644
> --- a/include/uapi/linux/tcp.h
> +++ b/include/uapi/linux/tcp.h
> @@ -112,6 +112,7 @@ enum {
>  #define TCP_FASTOPEN           23      /* Enable FastOpen on listeners */
>  #define TCP_TIMESTAMP          24
>  #define TCP_NOTSENT_LOWAT      25      /* limit number of unsent bytes in write queue */
> +#define TCP_CC_INFO            26      /* Get Congestion Control (optional) info */
>
>  struct tcp_repair_opt {
>         __u32   opt_code;
> diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
> index 8c5cd9efebbc..1e06c75a6837 100644
> --- a/net/ipv4/tcp.c
> +++ b/net/ipv4/tcp.c
> @@ -252,6 +252,7 @@
>  #include <linux/types.h>
>  #include <linux/fcntl.h>
>  #include <linux/poll.h>
> +#include <linux/inet_diag.h>
>  #include <linux/init.h>
>  #include <linux/fs.h>
>  #include <linux/skbuff.h>
> @@ -2734,6 +2735,26 @@ static int do_tcp_getsockopt(struct sock *sk, int level,
>                         return -EFAULT;
>                 return 0;
>         }
> +       case TCP_CC_INFO: {
> +               const struct tcp_congestion_ops *ca_ops;
> +               union tcp_cc_info info;
> +               size_t sz = 0;
> +               int attr;
> +
> +               if (get_user(len, optlen))
> +                       return -EFAULT;
> +
> +               ca_ops = icsk->icsk_ca_ops;
> +               if (ca_ops && ca_ops->get_info)
> +                       sz = ca_ops->get_info(sk, ~0U, &attr, &info);
> +
> +               len = min_t(unsigned int, len, sz);
> +               if (put_user(len, optlen))
> +                       return -EFAULT;
> +               if (copy_to_user(optval, &info, len))
> +                       return -EFAULT;
> +               return 0;
> +       }
>         case TCP_QUICKACK:
>                 val = !icsk->icsk_ack.pingpong;
>                 break;
> --
> 2.2.0.rc0.207.ga3a616c
>

Re: [PATCH net-next 1/2] tcp: prepare CC get_info() access from getsockopt()

[David Miller]

From: Eric Dumazet <edumazet@google.com>
Date: Tue, 28 Apr 2015 16:23:48 -0700

> We would like that optional info provided by Congestion Control
> modules using netlink can also be read using getsockopt()
> 
> This patch changes get_info() to put this information in a buffer,
> instead of skb, like tcp_get_info(), so that following patch
> can reuse this common infrastructure.
> 
> Signed-off-by: Eric Dumazet <edumazet@google.com>

Applied.

Re: [PATCH net-next 2/2] tcp: add TCP_CC_INFO socket option

[David Miller]

From: Eric Dumazet <edumazet@google.com>
Date: Tue, 28 Apr 2015 16:23:49 -0700

> Some Congestion Control modules can provide per flow information,
> but current way to get this information is to use netlink.
> 
> Like TCP_INFO, let's add TCP_CC_INFO so that applications can
> issue a getsockopt() if they have a socket file descriptor,
> instead of playing complex netlink games.
> 
> Sample usage would be :
> 
>   union tcp_cc_info info;
>   socklen_t len = sizeof(info);
> 
>   if (getsockopt(fd, SOL_TCP, TCP_CC_INFO, &info, &len) == -1)
> 
> Signed-off-by: Eric Dumazet <edumazet@google.com>

Applied.

Re: [PATCH net-next 1/2] tcp: prepare CC get_info() access from getsockopt()

[Eric Dumazet]

On Wed, 2015-04-29 at 18:12 -0400, David Miller wrote:
> From: Eric Dumazet <edumazet@google.com>
> Date: Tue, 28 Apr 2015 16:23:48 -0700
> 
> > We would like that optional info provided by Congestion Control
> > modules using netlink can also be read using getsockopt()
> > 
> > This patch changes get_info() to put this information in a buffer,
> > instead of skb, like tcp_get_info(), so that following patch
> > can reuse this common infrastructure.
> > 
> > Signed-off-by: Eric Dumazet <edumazet@google.com>
> 
> Applied.

Thanks David, I'll send an update to tcp_westwood, kbuild test robot
gave me a notice.

[PATCH net-next] tcp_westwood: fix tcp_westwood_info()

[Eric Dumazet]

From: Eric Dumazet <edumazet@google.com>

I forgot to update tcp_westwood when changing get_info() behavior,
this patch should fix this.

Fixes: 64f40ff5bbdb ("tcp: prepare CC get_info() access from getsockopt()")
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 net/ipv4/tcp_westwood.c |   15 ++++++++-------
 1 file changed, 8 insertions(+), 7 deletions(-)

diff --git a/net/ipv4/tcp_westwood.c b/net/ipv4/tcp_westwood.c
index b3c57cceb9907fe9d79f33f33369a96b647b5f26..c10732e39837872c724b801700f627a7fb1c9390 100644
--- a/net/ipv4/tcp_westwood.c
+++ b/net/ipv4/tcp_westwood.c
@@ -256,18 +256,19 @@ static void tcp_westwood_event(struct sock *sk, enum tcp_ca_event event)
 }
 
 /* Extract info for Tcp socket info provided via netlink. */
-static int tcp_westwood_info(struct sock *sk, u32 ext, struct sk_buff *skb)
+static size_t tcp_westwood_info(struct sock *sk, u32 ext, int *attr,
+				union tcp_cc_info *info)
 {
 	const struct westwood *ca = inet_csk_ca(sk);
 
 	if (ext & (1 << (INET_DIAG_VEGASINFO - 1))) {
-		struct tcpvegas_info info = {
-			.tcpv_enabled = 1,
-			.tcpv_rtt = jiffies_to_usecs(ca->rtt),
-			.tcpv_minrtt = jiffies_to_usecs(ca->rtt_min),
-		};
+		info->vegas.tcpv_enabled = 1;
+		info->vegas.tcpv_rttcnt	= 0;
+		info->vegas.tcpv_rtt	= jiffies_to_usecs(ca->rtt),
+		info->vegas.tcpv_minrtt	= jiffies_to_usecs(ca->rtt_min),
 
-		return nla_put(skb, INET_DIAG_VEGASINFO, sizeof(info), &info);
+		*attr = INET_DIAG_VEGASINFO;
+		return sizeof(struct tcpvegas_info);
 	}
 	return 0;
 }

Re: [PATCH net-next] tcp_westwood: fix tcp_westwood_info()

[Neal Cardwell]

On Wed, Apr 29, 2015 at 7:20 PM, Eric Dumazet <eric.dumazet@gmail.com> wrote:
> From: Eric Dumazet <edumazet@google.com>
>
> I forgot to update tcp_westwood when changing get_info() behavior,
> this patch should fix this.
>
> Fixes: 64f40ff5bbdb ("tcp: prepare CC get_info() access from getsockopt()")
> Reported-by: kbuild test robot <fengguang.wu@intel.com>
> Signed-off-by: Eric Dumazet <edumazet@google.com>

Acked-by: Neal Cardwell <ncardwell@google.com>

neal

Re: [PATCH net-next] tcp_westwood: fix tcp_westwood_info()

[David Miller]

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Wed, 29 Apr 2015 16:20:58 -0700

> From: Eric Dumazet <edumazet@google.com>
> 
> I forgot to update tcp_westwood when changing get_info() behavior,
> this patch should fix this.
> 
> Fixes: 64f40ff5bbdb ("tcp: prepare CC get_info() access from getsockopt()")
> Reported-by: kbuild test robot <fengguang.wu@intel.com>
> Signed-off-by: Eric Dumazet <edumazet@google.com>

Applied, thanks Eric.