[net-next,v2,2/2] tcp: allow to turn tcp timestamp randomization off

Eric says: "By looking at tcpdump, and TS val of xmit packets of multiple
flows, we can deduct the relative qdisc delays (think of fq pacing).
This should work even if we have one flow per remote peer."

Having random per flow (or host) offsets doesn't allow that anymore so add
a way to turn this off.

Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
---
 change since v2: do check in secure_tcpv4/6_sequence_number so outgoing
 syn packets won't have a random offset either in if randomization is off.

 Tested:
 sysctl_tcp_timestamps==1, tcpdump on lo, both ends have same values.

 Documentation/networking/ip-sysctl.txt | 9 +++++++--
 net/core/secure_seq.c                  | 5 +++--
 net/ipv4/tcp_input.c                   | 3 ++-
 3 files changed, 12 insertions(+), 5 deletions(-)

On Wed, Nov 30, 2016 at 4:28 AM, Florian Westphal <fw@strlen.de> wrote:
> Eric says: "By looking at tcpdump, and TS val of xmit packets of multiple
> flows, we can deduct the relative qdisc delays (think of fq pacing).
> This should work even if we have one flow per remote peer."
>
> Having random per flow (or host) offsets doesn't allow that anymore so add
> a way to turn this off.
>
> Suggested-by: Eric Dumazet <edumazet@google.com>
> Signed-off-by: Florian Westphal <fw@strlen.de>
> ---
>  change since v2: do check in secure_tcpv4/6_sequence_number so outgoing
>  syn packets won't have a random offset either in if randomization is off.
>
>  Tested:
>  sysctl_tcp_timestamps==1, tcpdump on lo, both ends have same values.
>
>  Documentation/networking/ip-sysctl.txt | 9 +++++++--
>  net/core/secure_seq.c                  | 5 +++--
>  net/ipv4/tcp_input.c                   | 3 ++-
>  3 files changed, 12 insertions(+), 5 deletions(-)
>
> diff --git a/Documentation/networking/ip-sysctl.txt b/Documentation/networking/ip-sysctl.txt
> index 5af48dd7c5fc..de2448313799 100644
> --- a/Documentation/networking/ip-sysctl.txt
> +++ b/Documentation/networking/ip-sysctl.txt
> @@ -610,8 +610,13 @@ tcp_syn_retries - INTEGER
>         with the current initial RTO of 1second. With this the final timeout
>         for an active TCP connection attempt will happen after 127seconds.
>
> -tcp_timestamps - BOOLEAN
> -       Enable timestamps as defined in RFC1323.
> +tcp_timestamps - INTEGER
> +Enable timestamps as defined in RFC1323.
> +       0: Disabled.
> +       1: Enable timestamps as defined in RFC1323.
> +       2: Like 1, but also use a random offset for each connection
> +       rather than only using the current time.
> +       Default: 2
Small suggestion: I suspect host/server configs manually set the knob
to 1. Perhaps swap 1 and 2 to maximize the coverage of this new
feature?

>
>  tcp_min_tso_segs - INTEGER
>         Minimal number of segments per TSO frame.
> diff --git a/net/core/secure_seq.c b/net/core/secure_seq.c
> index a8d6062cbb4a..36addd3d9633 100644
> --- a/net/core/secure_seq.c
> +++ b/net/core/secure_seq.c
> @@ -12,6 +12,7 @@
>  #include <net/secure_seq.h>
>
>  #if IS_ENABLED(CONFIG_IPV6) || IS_ENABLED(CONFIG_INET)
> +#include <net/tcp.h>
>  #define NET_SECRET_SIZE (MD5_MESSAGE_BYTES / 4)
>
>  static u32 net_secret[NET_SECRET_SIZE] ____cacheline_aligned;
> @@ -58,7 +59,7 @@ u32 secure_tcpv6_sequence_number(const __be32 *saddr, const __be32 *daddr,
>
>         md5_transform(hash, secret);
>
> -       *tsoff = hash[1];
> +       *tsoff = sysctl_tcp_timestamps == 2 ? hash[1] : 0;
>         return seq_scale(hash[0]);
>  }
>  EXPORT_SYMBOL(secure_tcpv6_sequence_number);
> @@ -100,7 +101,7 @@ u32 secure_tcp_sequence_number(__be32 saddr, __be32 daddr,
>
>         md5_transform(hash, net_secret);
>
> -       *tsoff = hash[1];
> +       *tsoff = sysctl_tcp_timestamps == 2 ? hash[1] : 0;
>         return seq_scale(hash[0]);
>  }
>
> diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
> index 1b1921c71f7c..5f6d4efd2551 100644
> --- a/net/ipv4/tcp_input.c
> +++ b/net/ipv4/tcp_input.c
> @@ -76,7 +76,7 @@
>  #include <asm/unaligned.h>
>  #include <linux/errqueue.h>
>
> -int sysctl_tcp_timestamps __read_mostly = 1;
> +int sysctl_tcp_timestamps __read_mostly = 2;
>  int sysctl_tcp_window_scaling __read_mostly = 1;
>  int sysctl_tcp_sack __read_mostly = 1;
>  int sysctl_tcp_fack __read_mostly = 1;
> @@ -85,6 +85,7 @@ int sysctl_tcp_dsack __read_mostly = 1;
>  int sysctl_tcp_app_win __read_mostly = 31;
>  int sysctl_tcp_adv_win_scale __read_mostly = 1;
>  EXPORT_SYMBOL(sysctl_tcp_adv_win_scale);
> +EXPORT_SYMBOL(sysctl_tcp_timestamps);
>
>  /* rfc5961 challenge ack rate limiting */
>  int sysctl_tcp_challenge_ack_limit = 1000;
> --
> 2.7.3
>

Yuchung Cheng <ycheng@google.com> wrote:
> > +tcp_timestamps - INTEGER
> > +Enable timestamps as defined in RFC1323.
> > +       0: Disabled.
> > +       1: Enable timestamps as defined in RFC1323.
> > +       2: Like 1, but also use a random offset for each connection
> > +       rather than only using the current time.
> > +       Default: 2
> Small suggestion: I suspect host/server configs manually set the knob
> to 1. Perhaps swap 1 and 2 to maximize the coverage of this new
> feature?

You mean:
1 (default): randomize
2: don't randomize?

I think its good idea, will send v3.  Thanks!