| Message ID | 1358887953-28665-1-git-send-email-avagin@openvz.org |
|---|---|
| State | Changes Requested, archived |
| Delegated to: | David Miller |
| Headers | show |
On 01/22/2013 12:52 PM, Andrey Vagin wrote: > If a TCP socket will get live-migrated from one box to another the > timestamps (which are typically ON) will get screwed up -- the new > kernel will generate TS values that has nothing to do with what they > were on dump. The solution is to yet again fix the kernel and put a > "timestamp offset" on a socket. Is there a chance a connection can be moved more than once within the "lifetime" of a given timestamp value? rick jones -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tue, Jan 22, 2013 at 3:52 PM, Andrey Vagin <avagin@openvz.org> wrote: > If a TCP socket will get live-migrated from one box to another the > timestamps (which are typically ON) will get screwed up -- the new > kernel will generate TS values that has nothing to do with what they > were on dump. The solution is to yet again fix the kernel and put a > "timestamp offset" on a socket. One serious issue with this patch is that outgoing timestamp values will no longer correspond to tcp_time_stamp, so echoed timestamp values will also no longer have a meaningful relationship to tcp_time_stamp. That violates assumptions made in several places in the code, which assumes that we can compare echoed timestamp values to tcp_time_stamp; for example, there are several places where we do things like subtracting: tcp_time_stamp - tp->rx_opt.rcv_tsecr to find the estimated RTT for a segment. neal -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
From: Neal Cardwell <ncardwell@google.com> Date: Tue, 22 Jan 2013 16:18:04 -0500 > On Tue, Jan 22, 2013 at 3:52 PM, Andrey Vagin <avagin@openvz.org> wrote: >> If a TCP socket will get live-migrated from one box to another the >> timestamps (which are typically ON) will get screwed up -- the new >> kernel will generate TS values that has nothing to do with what they >> were on dump. The solution is to yet again fix the kernel and put a >> "timestamp offset" on a socket. > > One serious issue with this patch is that outgoing timestamp values > will no longer correspond to tcp_time_stamp, so echoed timestamp > values will also no longer have a meaningful relationship to > tcp_time_stamp. That violates assumptions made in several places in > the code, which assumes that we can compare echoed timestamp values to > tcp_time_stamp; for example, there are several places where we do > things like subtracting: > tcp_time_stamp - tp->rx_opt.rcv_tsecr > to find the estimated RTT for a segment. Right, this change seems pretty bogus as-is. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/include/linux/tcp.h b/include/linux/tcp.h index 4e1d228..746dad5 100644 --- a/include/linux/tcp.h +++ b/include/linux/tcp.h @@ -162,6 +162,8 @@ struct tcp_sock { u32 rcv_tstamp; /* timestamp of last received ACK (for keepalives) */ u32 lsndtime; /* timestamp of last sent data packet (for restart window) */ + u32 snd_tsval_offset; /* offset for snd_tsval */ + struct list_head tsq_node; /* anchor in tsq_tasklet.head list */ unsigned long tsq_flags; diff --git a/include/uapi/linux/tcp.h b/include/uapi/linux/tcp.h index e962faa..6b1ead0 100644 --- a/include/uapi/linux/tcp.h +++ b/include/uapi/linux/tcp.h @@ -111,6 +111,7 @@ enum { #define TCP_QUEUE_SEQ 21 #define TCP_REPAIR_OPTIONS 22 #define TCP_FASTOPEN 23 /* Enable FastOpen on listeners */ +#define TCP_TIMESTAMP 24 struct tcp_repair_opt { __u32 opt_code; diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 1ca2536..72dee28 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -2704,6 +2704,9 @@ static int do_tcp_setsockopt(struct sock *sk, int level, else err = -EINVAL; break; + case TCP_TIMESTAMP: + tp->snd_tsval_offset = val - tcp_time_stamp; + break; default: err = -ENOPROTOOPT; break; @@ -2952,6 +2955,9 @@ static int do_tcp_getsockopt(struct sock *sk, int level, case TCP_USER_TIMEOUT: val = jiffies_to_msecs(icsk->icsk_user_timeout); break; + case TCP_TIMESTAMP: + val = tcp_time_stamp + tp->snd_tsval_offset; + break; default: return -ENOPROTOOPT; } diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c index 5d45159..9b6d485 100644 --- a/net/ipv4/tcp_output.c +++ b/net/ipv4/tcp_output.c @@ -622,7 +622,7 @@ static unsigned int tcp_syn_options(struct sock *sk, struct sk_buff *skb, if (likely(sysctl_tcp_timestamps && *md5 == NULL)) { opts->options |= OPTION_TS; - opts->tsval = TCP_SKB_CB(skb)->when; + opts->tsval = TCP_SKB_CB(skb)->when + tp->snd_tsval_offset; opts->tsecr = tp->rx_opt.ts_recent; remaining -= TCPOLEN_TSTAMP_ALIGNED; } @@ -705,6 +705,7 @@ static unsigned int tcp_synack_options(struct sock *sk, struct tcp_extend_values *xvp, struct tcp_fastopen_cookie *foc) { + struct tcp_sock *tp = tcp_sk(sk); struct inet_request_sock *ireq = inet_rsk(req); unsigned int remaining = MAX_TCP_OPTION_SPACE; u8 cookie_plus = (xvp != NULL && !xvp->cookie_out_never) ? @@ -739,7 +740,7 @@ static unsigned int tcp_synack_options(struct sock *sk, } if (likely(ireq->tstamp_ok)) { opts->options |= OPTION_TS; - opts->tsval = TCP_SKB_CB(skb)->when; + opts->tsval = TCP_SKB_CB(skb)->when + tp->snd_tsval_offset; opts->tsecr = req->ts_recent; remaining -= TCPOLEN_TSTAMP_ALIGNED; } @@ -806,7 +807,7 @@ static unsigned int tcp_established_options(struct sock *sk, struct sk_buff *skb if (likely(tp->rx_opt.tstamp_ok)) { opts->options |= OPTION_TS; - opts->tsval = tcb ? tcb->when : 0; + opts->tsval = tcb ? tcb->when + tp->snd_tsval_offset : 0; opts->tsecr = tp->rx_opt.ts_recent; size += TCPOLEN_TSTAMP_ALIGNED; }
If a TCP socket will get live-migrated from one box to another the timestamps (which are typically ON) will get screwed up -- the new kernel will generate TS values that has nothing to do with what they were on dump. The solution is to yet again fix the kernel and put a "timestamp offset" on a socket. Cc: "David S. Miller" <davem@davemloft.net> Cc: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> Cc: James Morris <jmorris@namei.org> Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org> Cc: Patrick McHardy <kaber@trash.net> Cc: Eric Dumazet <edumazet@google.com> Cc: Yuchung Cheng <ycheng@google.com> Cc: Neal Cardwell <ncardwell@google.com> Cc: Pavel Emelyanov <xemul@parallels.com> Cc: Dave Jones <davej@redhat.com> Cc: Michael Kerrisk <mtk.manpages@gmail.com> Signed-off-by: Andrey Vagin <avagin@openvz.org> --- include/linux/tcp.h | 2 ++ include/uapi/linux/tcp.h | 1 + net/ipv4/tcp.c | 6 ++++++ net/ipv4/tcp_output.c | 7 ++++--- 4 files changed, 13 insertions(+), 3 deletions(-)