Add EXPORT_SYMBOL macro for tcp_tso_autosize function

.gitignore

@@ -156,6 +156,9 @@ x509.genkey
 /allrandom.config
 /allyes.config
 
+# tmp build/install directory for /gce-install.sh:
+/gce/
+
 # Kconfig savedefconfig output
 /defconfig
 

0001-ss-output-TCP-BBRv3-diag-information.patch

@@ -0,0 +1,165 @@
+From ca7f11ebc4d4a99ccfd44be8555d505b26996c12 Mon Sep 17 00:00:00 2001
+From: Arjun Roy <arjunroy@google.com>
+Date: Mon, 25 Jul 2022 12:49:35 -0400
+Subject: [PATCH 2/2] ss: output TCP BBRv3 diag information
+
+Add logic for printing diag information for TCP BBRv3 congestion
+control.  This commit leaves in place the support for printing the
+earlier TCP BBRv1 congestion control information.
+
+Both BBRv1 and BBRv3 are using the same enum value.  The BBRv3 struct
+starts with the same data as BBRv1, so it is is backward-compatible
+with BBRv1, to allow lder ss binaries to print basic information for
+BBRv3. We use the size of the returned data and the version field to
+check the version of the data.
+
+Signed-off-by: Arjun Roy <arjunroy@google.com>
+Signed-off-by: Neal Cardwell <ncardwell@google.com>
+Signed-off-by: David Morley <morleyd@google.com>
+---
+ include/uapi/linux/inet_diag.h | 23 ++++++++++++
+ misc/ss.c                      | 66 +++++++++++++++++++++++++++++++++-
+ 2 files changed, 88 insertions(+), 1 deletion(-)
+
+diff --git a/include/uapi/linux/inet_diag.h b/include/uapi/linux/inet_diag.h
+index d81cb69a26a9..dca833ecb783 100644
+--- a/include/uapi/linux/inet_diag.h
++++ b/include/uapi/linux/inet_diag.h
+@@ -229,6 +229,29 @@ struct tcp_bbr_info {
+ 	__u32	bbr_min_rtt;		/* min-filtered RTT in uSec */
+ 	__u32	bbr_pacing_gain;	/* pacing gain shifted left 8 bits */
+ 	__u32	bbr_cwnd_gain;		/* cwnd gain shifted left 8 bits */
++	__u32	bbr_bw_hi_lsb;		/* lower 32 bits of bw_hi */
++	__u32	bbr_bw_hi_msb;		/* upper 32 bits of bw_hi */
++	__u32	bbr_bw_lo_lsb;		/* lower 32 bits of bw_lo */
++	__u32	bbr_bw_lo_msb;		/* upper 32 bits of bw_lo */
++	__u8	bbr_mode;		/* current bbr_mode in state machine */
++	__u8	bbr_phase;		/* current state machine phase */
++	__u8	unused1;		/* alignment padding; not used yet */
++	__u8	bbr_version;		/* BBR algorithm version */
++	__u32	bbr_inflight_lo;	/* lower short-term data volume bound */
++	__u32	bbr_inflight_hi;	/* higher long-term data volume bound */
++	__u32	bbr_extra_acked;	/* max excess packets ACKed in epoch */
++};
++
++/* TCP BBR congestion control bbr_phase as reported in netlink/ss stats. */
++enum tcp_bbr_phase {
++	BBR_PHASE_INVALID		= 0,
++	BBR_PHASE_STARTUP		= 1,
++	BBR_PHASE_DRAIN			= 2,
++	BBR_PHASE_PROBE_RTT		= 3,
++	BBR_PHASE_PROBE_BW_UP		= 4,
++	BBR_PHASE_PROBE_BW_DOWN		= 5,
++	BBR_PHASE_PROBE_BW_CRUISE	= 6,
++	BBR_PHASE_PROBE_BW_REFILL	= 7,
+ };
+
+ union tcp_cc_info {
+diff --git a/misc/ss.c b/misc/ss.c
+index e9d813596b91..5f413118f0dd 100644
+--- a/misc/ss.c
++++ b/misc/ss.c
+@@ -912,6 +912,7 @@ struct tcpstat {
+ 	bool		    app_limited;
+ 	struct dctcpstat    *dctcp;
+ 	struct tcp_bbr_info *bbr_info;
++	unsigned int	    bbr_info_len;
+ };
+
+ /* SCTP assocs share the same inode number with their parent endpoint. So if we
+@@ -2585,6 +2586,29 @@ static void sctp_stats_print(struct sctp_info *s)
+ 		out(" fraginl:%d", s->sctpi_s_frag_interleave);
+ }
+
++static const char* bbr_phase_to_str(enum tcp_bbr_phase phase)
++{
++	switch (phase) {
++	case BBR_PHASE_STARTUP:
++		return "STARTUP";
++	case BBR_PHASE_DRAIN:
++		return "DRAIN";
++	case BBR_PHASE_PROBE_RTT:
++		return "PROBE_RTT";
++	case BBR_PHASE_PROBE_BW_UP:
++		return "PROBE_BW_UP";
++	case BBR_PHASE_PROBE_BW_DOWN:
++		return "PROBE_BW_DOWN";
++	case BBR_PHASE_PROBE_BW_CRUISE:
++		return "PROBE_BW_CRUISE";
++	case BBR_PHASE_PROBE_BW_REFILL:
++		return "PROBE_BW_REFILL";
++	case BBR_PHASE_INVALID:
++	default:
++		return "INVALID";
++	}
++}
++
+ static void tcp_stats_print(struct tcpstat *s)
+ {
+ 	char b1[64];
+@@ -2658,7 +2682,14 @@ static void tcp_stats_print(struct tcpstat *s)
+ 	}
+
+ 	if (s->bbr_info) {
+-		__u64 bw;
++		/* All versions of the BBR algorithm use the INET_DIAG_BBRINFO
++		 * enum value. Later versions of the tcp_bbr_info struct are
++		 * backward-compatible with earlier versions, to allow older ss
++		 * binaries to print basic information for newer versions of
++		 * the algorithm. We use the size of the returned tcp_bbr_info
++		 * struct to decide how much to print.
++		 */
++		__u64 bw, bw_hi, bw_lo;
+
+ 		bw = s->bbr_info->bbr_bw_hi;
+ 		bw <<= 32;
+@@ -2673,6 +2704,38 @@ static void tcp_stats_print(struct tcpstat *s)
+ 		if (s->bbr_info->bbr_cwnd_gain)
+ 			out(",cwnd_gain:%g",
+ 			    (double)s->bbr_info->bbr_cwnd_gain / 256.0);
++
++		if (s->bbr_info_len >=
++		    (offsetof(struct tcp_bbr_info, bbr_extra_acked) +
++		     sizeof(__u32))) {
++
++			bw_hi = s->bbr_info->bbr_bw_hi_msb;
++			bw_hi <<= 32;
++			bw_hi |= s->bbr_info->bbr_bw_hi_lsb;
++
++			bw_lo = s->bbr_info->bbr_bw_lo_msb;
++			bw_lo <<= 32;
++			bw_lo |= s->bbr_info->bbr_bw_lo_lsb;
++
++			out(",version:%u", s->bbr_info->bbr_version);
++			if (bw_hi != ~0UL)
++				out(",bw_hi:%sbps", sprint_bw(b1, bw_hi * 8.0));
++			if (bw_lo != ~0UL)
++				out(",bw_lo:%sbps", sprint_bw(b1, bw_lo * 8.0));
++			if (s->bbr_info->bbr_inflight_hi != ~0U)
++				out(",inflight_hi:%u",
++				    s->bbr_info->bbr_inflight_hi);
++			if (s->bbr_info->bbr_inflight_lo != ~0U)
++				out(",inflight_lo:%u",
++				    s->bbr_info->bbr_inflight_lo);
++			out(",extra_acked:%u", s->bbr_info->bbr_extra_acked);
++			out(",mode:%d", (int)s->bbr_info->bbr_mode);
++			out(",phase:%s",
++			    bbr_phase_to_str(
++				    (enum tcp_bbr_phase)
++				    s->bbr_info->bbr_phase));
++		}
++
+ 		out(")");
+ 	}
+
+@@ -3147,6 +3210,7 @@ static void tcp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r,
+ 			s.bbr_info = calloc(1, sizeof(*s.bbr_info));
+ 			if (s.bbr_info && bbr_info)
+ 				memcpy(s.bbr_info, bbr_info, len);
++			s.bbr_info_len = len;
+ 		}
+
+ 		if (rtt > 0 && info->tcpi_snd_mss && info->tcpi_snd_cwnd) {
+-- 
+2.41.0.487.g6d72f3e995-goog
+

0002-ip-introduce-the-ecn_low-per-route-feature.patch

@@ -0,0 +1,120 @@
+From 537b1b761e1d0036923adba7a80d3655cfff095d Mon Sep 17 00:00:00 2001
+From: David Morley <morleyd@google.com>
+Date: Fri, 21 Jul 2023 09:20:46 +0000
+Subject: [PATCH 1/2] ip: introduce the ecn_low per-route feature
+
+Add a a new "ecn_low" feature that administrators can add to a
+particular ip route. The ecn_low feature indicates that the given
+destination network is a low-latency ECN environment, meaning both
+that ECN CE marks are applied by the network using a low-latency
+marking threshold and also that TCP endpoints provide precise
+per-data-segment ECN feedback in ACKs (where the ACK ECE flag echoes
+the received CE status of all newly-acknowledged data segments). This
+ecn_low feature indication can be used by congestion control
+algorithms to decide how to interpret ECN signals over the given
+destination network.
+
+Signed-off-by: David Morley <morleyd@google.com>
+Signed-off-by: Neal Cardwell <ncardwell@google.com>
+---
+ include/uapi/linux/rtnetlink.h |  4 +++-
+ ip/iproute.c                   |  8 +++++++-
+ man/man8/ip-route.8.in         | 19 ++++++++++++++-----
+ 3 files changed, 24 insertions(+), 7 deletions(-)
+
+diff --git a/include/uapi/linux/rtnetlink.h b/include/uapi/linux/rtnetlink.h
+index 2132e941b93a..66e3237ea915 100644
+--- a/include/uapi/linux/rtnetlink.h
++++ b/include/uapi/linux/rtnetlink.h
+@@ -506,9 +506,11 @@ enum {
+ #define RTAX_FEATURE_SACK	(1 << 1)
+ #define RTAX_FEATURE_TIMESTAMP	(1 << 2)
+ #define RTAX_FEATURE_ALLFRAG	(1 << 3)
++#define RTAX_FEATURE_ECN_LOW	(1 << 4)
+
+ #define RTAX_FEATURE_MASK	(RTAX_FEATURE_ECN | RTAX_FEATURE_SACK | \
+-				 RTAX_FEATURE_TIMESTAMP | RTAX_FEATURE_ALLFRAG)
++				 RTAX_FEATURE_TIMESTAMP | RTAX_FEATURE_ALLFRAG | \
++				 RTAX_FEATURE_ECN_LOW)
+
+ struct rta_session {
+ 	__u8	proto;
+diff --git a/ip/iproute.c b/ip/iproute.c
+index fdf1f9a9dd0a..bd5d783cfdbb 100644
+--- a/ip/iproute.c
++++ b/ip/iproute.c
+@@ -96,7 +96,7 @@ static void usage(void)
+ 		"PREF := [ low | medium | high ]\n"
+ 		"TIME := NUMBER[s|ms]\n"
+ 		"BOOL := [1|0]\n"
+-		"FEATURES := ecn\n"
++		"FEATURES := [ ecn | ecn_low ]\n"
+ 		"ENCAPTYPE := [ mpls | ip | ip6 | seg6 | seg6local | rpl | ioam6 | xfrm ]\n"
+ 		"ENCAPHDR := [ MPLSLABEL | SEG6HDR | SEG6LOCAL | IOAM6HDR | XFRMINFO ]\n"
+ 		"SEG6HDR := [ mode SEGMODE ] segs ADDR1,ADDRi,ADDRn [hmac HMACKEYID] [cleanup]\n"
+@@ -350,6 +350,10 @@ static void print_rtax_features(FILE *fp, unsigned int features)
+ 		print_null(PRINT_ANY, "ecn", "ecn ", NULL);
+ 		features &= ~RTAX_FEATURE_ECN;
+ 	}
++	if (features & RTAX_FEATURE_ECN_LOW) {
++		print_null(PRINT_ANY, "ecn_low", "ecn_low ", NULL);
++		features &= ~RTAX_FEATURE_ECN_LOW;
++	}
+
+ 	if (features)
+ 		print_0xhex(PRINT_ANY,
+@@ -1349,6 +1353,8 @@ static int iproute_modify(int cmd, unsigned int flags, int argc, char **argv)
+
+ 				if (strcmp(*argv, "ecn") == 0)
+ 					features |= RTAX_FEATURE_ECN;
++				else if (strcmp(*argv, "ecn_low") == 0)
++					features |= RTAX_FEATURE_ECN_LOW;
+ 				else
+ 					invarg("\"features\" value not valid\n", *argv);
+ 				break;
+diff --git a/man/man8/ip-route.8.in b/man/man8/ip-route.8.in
+index c2b00833e507..10f35f38e6a2 100644
+--- a/man/man8/ip-route.8.in
++++ b/man/man8/ip-route.8.in
+@@ -181,7 +181,7 @@ throw " | " unreachable " | " prohibit " | " blackhole " | " nat " ]"
+
+ .ti -8
+ .IR FEATURES " := [ "
+-.BR ecn " | ]"
++.BR ecn " | " ecn_low " ] "
+
+ .ti -8
+ .IR PREF " := [ "
+@@ -544,16 +544,25 @@ The default value is zero, meaning to use Slow Start value.
+
+ .TP
+ .BI features " FEATURES " (Linux 3.18+ only)
+-Enable or disable per-route features. Only available feature at this
+-time is
+-.B ecn
+-to enable explicit congestion notification when initiating connections to the
++Enable or disable per-route features. Available features include:
++
++.BI ecn
++- to enable explicit congestion notification when initiating connections to the
+ given destination network.
+ When responding to a connection request from the given network, ecn will
+ also be used even if the
+ .B net.ipv4.tcp_ecn
+ sysctl is set to 0.
+
++.BI ecn_low
++- to indicate that the given destination network is a low-latency ECN
++environment, meaning both that ECN CE marks are applied by the network using a
++low-latency marking threshold and also that TCP endpoints provide precise
++per-data-segment ECN feedback in ACKs (where the ACK ECE flag echoes the
++received CE status of all newly-acknowledged data segments). This ecn_low
++feature indication can be used by congestion control algorithms to decide how
++to interpret ECN signals over the given destination network (Linux 6.7+ only).
++
+ .TP
+ .BI quickack " BOOL " "(Linux 3.11+ only)"
+ Enable or disable quick ack for connections to this destination.
+-- 
+2.41.0.487.g6d72f3e995-goog
+

0003-ss-display-ecn_low-if-tcp_info-tcpi_options-TCPI_OPT.patch

@@ -0,0 +1,59 @@
+From 107339d7f48c95ae8a7461150e143fc53b08fea9 Mon Sep 17 00:00:00 2001
+From: Neal Cardwell <ncardwell@google.com>
+Date: Sun, 23 Jul 2023 23:33:21 -0400
+Subject: [PATCH] ss: display "ecn_low" if tcp_info tcpi_options
+ TCPI_OPT_ECN_LOW bit is set
+
+Display "ecn_low" if the TCPI_OPT_ECN_LOW bit is set in the
+tcpi_options field in tcp_info.
+
+Signed-off-by: Neal Cardwell <ncardwell@google.com>
+---
+ include/uapi/linux/tcp.h | 1 +
+ misc/ss.c                | 4 ++++
+ 2 files changed, 5 insertions(+)
+
+diff --git a/include/uapi/linux/tcp.h b/include/uapi/linux/tcp.h
+index 4ee5d7721c6f..238e95d7693c 100644
+--- a/include/uapi/linux/tcp.h
++++ b/include/uapi/linux/tcp.h
+@@ -170,6 +170,7 @@ enum tcp_fastopen_client_fail {
+ #define TCPI_OPT_ECN		8 /* ECN was negociated at TCP session init */
+ #define TCPI_OPT_ECN_SEEN	16 /* we received at least one packet with ECT */
+ #define TCPI_OPT_SYN_DATA	32 /* SYN-ACK acked data in SYN sent or rcvd */
++#define TCPI_OPT_ECN_LOW	64 /* Low-latency ECN configured at init */
+
+ /*
+  * Sender's congestion state indicating normal or abnormal situations
+diff --git a/misc/ss.c b/misc/ss.c
+index 5f413118f0dd..634285adb7d3 100644
+--- a/misc/ss.c
++++ b/misc/ss.c
+@@ -907,6 +907,7 @@ struct tcpstat {
+ 	bool		    has_sack_opt;
+ 	bool		    has_ecn_opt;
+ 	bool		    has_ecnseen_opt;
++	bool		    has_ecn_low_opt;
+ 	bool		    has_fastopen_opt;
+ 	bool		    has_wscale_opt;
+ 	bool		    app_limited;
+@@ -2621,6 +2622,8 @@ static void tcp_stats_print(struct tcpstat *s)
+ 		out(" ecn");
+ 	if (s->has_ecnseen_opt)
+ 		out(" ecnseen");
++	if (s->has_ecn_low_opt)
++		out(" ecn_low");
+ 	if (s->has_fastopen_opt)
+ 		out(" fastopen");
+ 	if (s->cong_alg[0])
+@@ -3134,6 +3137,7 @@ static void tcp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r,
+ 			s.has_sack_opt	   = TCPI_HAS_OPT(info, TCPI_OPT_SACK);
+ 			s.has_ecn_opt	   = TCPI_HAS_OPT(info, TCPI_OPT_ECN);
+ 			s.has_ecnseen_opt  = TCPI_HAS_OPT(info, TCPI_OPT_ECN_SEEN);
++			s.has_ecn_low_opt  = TCPI_HAS_OPT(info, TCPI_OPT_ECN_LOW);
+ 			s.has_fastopen_opt = TCPI_HAS_OPT(info, TCPI_OPT_SYN_DATA);
+ 		}
+
+-- 
+2.41.0.487.g6d72f3e995-goog
+