From owner-dev-commits-src-main@freebsd.org Fri Jun 25 13:32:28 2021 Return-Path: Delivered-To: dev-commits-src-main@mailman.nyi.freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2610:1c1:1:606c::19:1]) by mailman.nyi.freebsd.org (Postfix) with ESMTP id D90716614AC; Fri, 25 Jun 2021 13:32:28 +0000 (UTC) (envelope-from git@FreeBSD.org) Received: from mxrelay.nyi.freebsd.org (mxrelay.nyi.freebsd.org [IPv6:2610:1c1:1:606c::19:3]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange X25519 server-signature RSA-PSS (4096 bits) server-digest SHA256 client-signature RSA-PSS (4096 bits) client-digest SHA256) (Client CN "mxrelay.nyi.freebsd.org", Issuer "R3" (verified OK)) by mx1.freebsd.org (Postfix) with ESMTPS id 4GBHx85qhTz4TBs; Fri, 25 Jun 2021 13:32:28 +0000 (UTC) (envelope-from git@FreeBSD.org) Received: from gitrepo.freebsd.org (gitrepo.freebsd.org [IPv6:2610:1c1:1:6068::e6a:5]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange X25519 server-signature RSA-PSS (4096 bits) server-digest SHA256) (Client did not present a certificate) by mxrelay.nyi.freebsd.org (Postfix) with ESMTPS id B1F15411A; Fri, 25 Jun 2021 13:32:28 +0000 (UTC) (envelope-from git@FreeBSD.org) Received: from gitrepo.freebsd.org ([127.0.1.44]) by gitrepo.freebsd.org (8.16.1/8.16.1) with ESMTP id 15PDWSdG068175; Fri, 25 Jun 2021 13:32:28 GMT (envelope-from git@gitrepo.freebsd.org) Received: (from git@localhost) by gitrepo.freebsd.org (8.16.1/8.16.1/Submit) id 15PDWSrC068174; Fri, 25 Jun 2021 13:32:28 GMT (envelope-from git) Date: Fri, 25 Jun 2021 13:32:28 GMT Message-Id: <202106251332.15PDWSrC068174@gitrepo.freebsd.org> To: src-committers@FreeBSD.org, dev-commits-src-all@FreeBSD.org, dev-commits-src-main@FreeBSD.org From: Randall Stewart Subject: git: 9e4d9e4c4d79 - main - tcp: Preparation for allowing hardware TLS to be able to kick a tcp connection that is retransmitting too much out of hardware and back to software. MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 8bit X-Git-Committer: rrs X-Git-Repository: src X-Git-Refname: refs/heads/main X-Git-Reftype: branch X-Git-Commit: 9e4d9e4c4d79db58740a05d67645351e640aa32c Auto-Submitted: auto-generated X-BeenThere: dev-commits-src-main@freebsd.org X-Mailman-Version: 2.1.34 Precedence: list List-Id: Commit messages for the main branch of the src repository List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 25 Jun 2021 13:32:28 -0000 The branch main has been updated by rrs: URL: https://cgit.FreeBSD.org/src/commit/?id=9e4d9e4c4d79db58740a05d67645351e640aa32c commit 9e4d9e4c4d79db58740a05d67645351e640aa32c Author: Randall Stewart AuthorDate: 2021-06-25 13:30:54 +0000 Commit: Randall Stewart CommitDate: 2021-06-25 13:30:54 +0000 tcp: Preparation for allowing hardware TLS to be able to kick a tcp connection that is retransmitting too much out of hardware and back to software. Hardware TLS is now supported in some interface cards and it works well. Except that when we have connections that retransmit a lot we get into trouble with all the retransmits. This prep step makes way for change that Drew will be making so that we can "kick out" a session from hardware TLS. Reviewed by: mtuexen, gallatin Sponsored by: Netflix Inc Differential Revision: https://reviews.freebsd.org/D30895 --- sys/netinet/tcp_output.c | 2 +- sys/netinet/tcp_stacks/bbr.c | 2 +- sys/netinet/tcp_stacks/rack.c | 68 +++++++++++++++++++++++++++++++-------- sys/netinet/tcp_stacks/tcp_rack.h | 6 ++-- sys/netinet/tcp_var.h | 6 ++-- 5 files changed, 63 insertions(+), 21 deletions(-) diff --git a/sys/netinet/tcp_output.c b/sys/netinet/tcp_output.c index 8d78e639636b..d0b56072e9af 100644 --- a/sys/netinet/tcp_output.c +++ b/sys/netinet/tcp_output.c @@ -1560,7 +1560,7 @@ send: out: if (error == 0) - tcp_account_for_send(tp, len, (tp->snd_nxt != tp->snd_max), 0); + tcp_account_for_send(tp, len, (tp->snd_nxt != tp->snd_max), 0, hw_tls); /* * In transmit state, time the transmission and arrange for * the retransmit. In persist state, just set snd_max. diff --git a/sys/netinet/tcp_stacks/bbr.c b/sys/netinet/tcp_stacks/bbr.c index 57abbbf694b6..f8cd0bec1c95 100644 --- a/sys/netinet/tcp_stacks/bbr.c +++ b/sys/netinet/tcp_stacks/bbr.c @@ -13750,7 +13750,7 @@ out: * retransmit. In persist state, just set snd_max. */ if (error == 0) { - tcp_account_for_send(tp, len, (rsm != NULL), doing_tlp); + tcp_account_for_send(tp, len, (rsm != NULL), doing_tlp, hw_tls); if (TCPS_HAVEESTABLISHED(tp->t_state) && (tp->t_flags & TF_SACK_PERMIT) && tp->rcv_numsacks > 0) diff --git a/sys/netinet/tcp_stacks/rack.c b/sys/netinet/tcp_stacks/rack.c index 521fd6c625f5..6678ca92738b 100644 --- a/sys/netinet/tcp_stacks/rack.c +++ b/sys/netinet/tcp_stacks/rack.c @@ -32,6 +32,7 @@ __FBSDID("$FreeBSD$"); #include "opt_ipsec.h" #include "opt_tcpdebug.h" #include "opt_ratelimit.h" +#include "opt_kern_tls.h" #include #include #include @@ -483,7 +484,7 @@ rack_log_ack(struct tcpcb *tp, struct tcpopt *to, static void rack_log_output(struct tcpcb *tp, struct tcpopt *to, int32_t len, uint32_t seq_out, uint8_t th_flags, int32_t err, uint64_t ts, - struct rack_sendmap *hintrsm, uint16_t add_flags, struct mbuf *s_mb, uint32_t s_moff); + struct rack_sendmap *hintrsm, uint16_t add_flags, struct mbuf *s_mb, uint32_t s_moff, int hw_tls); static void rack_log_sack_passed(struct tcpcb *tp, struct tcp_rack *rack, @@ -6061,7 +6062,8 @@ rack_clone_rsm(struct tcp_rack *rack, struct rack_sendmap *nrsm, /* Push bit must go to the right edge as well */ if (rsm->r_flags & RACK_HAD_PUSH) rsm->r_flags &= ~RACK_HAD_PUSH; - + /* Clone over the state of the hw_tls flag */ + nrsm->r_hw_tls = rsm->r_hw_tls; /* * Now we need to find nrsm's new location in the mbuf chain * we basically calculate a new offset, which is soff + @@ -7191,7 +7193,7 @@ rack_update_entry(struct tcpcb *tp, struct tcp_rack *rack, static void rack_log_output(struct tcpcb *tp, struct tcpopt *to, int32_t len, uint32_t seq_out, uint8_t th_flags, int32_t err, uint64_t cts, - struct rack_sendmap *hintrsm, uint16_t add_flag, struct mbuf *s_mb, uint32_t s_moff) + struct rack_sendmap *hintrsm, uint16_t add_flag, struct mbuf *s_mb, uint32_t s_moff, int hw_tls) { struct tcp_rack *rack; struct rack_sendmap *rsm, *nrsm, *insret, fe; @@ -7295,6 +7297,8 @@ again: } else { rsm->r_flags = add_flag; } + if (hw_tls) + rsm->r_hw_tls = 1; rsm->r_tim_lastsent[0] = cts; rsm->r_rtr_cnt = 1; rsm->r_rtr_bytes = 0; @@ -14875,7 +14879,7 @@ rack_log_fsb(struct tcp_rack *rack, struct tcpcb *tp, struct socket *so, uint32_ static struct mbuf * rack_fo_base_copym(struct mbuf *the_m, uint32_t the_off, int32_t *plen, struct rack_fast_send_blk *fsb, - int32_t seglimit, int32_t segsize) + int32_t seglimit, int32_t segsize, int hw_tls) { #ifdef KERN_TLS struct ktls_session *tls, *ntls; @@ -15059,7 +15063,7 @@ rack_fo_m_copym(struct tcp_rack *rack, int32_t *plen, *s_mb = rack->r_ctl.fsb.m; n = rack_fo_base_copym(m, soff, plen, &rack->r_ctl.fsb, - seglimit, segsize); + seglimit, segsize, rack->r_ctl.fsb.hw_tls); return (n); } @@ -15243,7 +15247,7 @@ rack_fast_rsm_output(struct tcpcb *tp, struct tcp_rack *rack, struct rack_sendma /* Fix up the orig_m_len and possibly the mbuf offset */ rack_adjust_orig_mlen(rsm); } - m->m_next = rack_fo_base_copym(rsm->m, rsm->soff, &len, NULL, if_hw_tsomaxsegcount, if_hw_tsomaxsegsize); + m->m_next = rack_fo_base_copym(rsm->m, rsm->soff, &len, NULL, if_hw_tsomaxsegcount, if_hw_tsomaxsegsize, rsm->r_hw_tls); if (len <= segsiz) { /* * Must have ran out of mbufs for the copy @@ -15405,13 +15409,13 @@ rack_fast_rsm_output(struct tcpcb *tp, struct tcp_rack *rack, struct rack_sendma goto failed; } rack_log_output(tp, &to, len, rsm->r_start, flags, error, rack_to_usec_ts(tv), - rsm, RACK_SENT_FP, rsm->m, rsm->soff); + rsm, RACK_SENT_FP, rsm->m, rsm->soff, rsm->r_hw_tls); if (doing_tlp && (rack->fast_rsm_hack == 0)) { rack->rc_tlp_in_progress = 1; rack->r_ctl.rc_tlp_cnt_out++; } if (error == 0) - tcp_account_for_send(tp, len, 1, doing_tlp); + tcp_account_for_send(tp, len, 1, doing_tlp, rsm->r_hw_tls); tp->t_flags &= ~(TF_ACKNOW | TF_DELACK); rack->forced_ack = 0; /* If we send something zap the FA flag */ if (IN_FASTRECOVERY(tp->t_flags) && rsm) @@ -15688,7 +15692,8 @@ again: goto failed; /* s_mb and s_soff are saved for rack_log_output */ - m->m_next = rack_fo_m_copym(rack, &len, if_hw_tsomaxsegcount, if_hw_tsomaxsegsize, &s_mb, &s_soff); + m->m_next = rack_fo_m_copym(rack, &len, if_hw_tsomaxsegcount, if_hw_tsomaxsegsize, + &s_mb, &s_soff); if (len <= segsiz) { /* * Must have ran out of mbufs for the copy @@ -15883,7 +15888,7 @@ again: goto failed; } rack_log_output(tp, &to, len, tp->snd_max, flags, error, rack_to_usec_ts(tv), - NULL, add_flag, s_mb, s_soff); + NULL, add_flag, s_mb, s_soff, rack->r_ctl.fsb.hw_tls); m = NULL; if (tp->snd_una == tp->snd_max) { rack->r_ctl.rc_tlp_rxt_last_time = cts; @@ -15891,7 +15896,7 @@ again: tp->t_acktime = ticks; } if (error == 0) - tcp_account_for_send(tp, len, 0, 0); + tcp_account_for_send(tp, len, 0, 0, rack->r_ctl.fsb.hw_tls); rack->forced_ack = 0; /* If we send something zap the FA flag */ tot_len += len; @@ -17030,6 +17035,10 @@ just_return_nolock: rack->r_ctl.fsb.o_m_len = rack->r_ctl.fsb.m->m_len; rack->r_ctl.fsb.tcp_flags = flags; rack->r_ctl.fsb.left_to_send = orig_len - len; + if (hw_tls) + rack->r_ctl.fsb.hw_tls = 1; + else + rack->r_ctl.fsb.hw_tls = 0; KASSERT((rack->r_ctl.fsb.left_to_send <= (sbavail(sb) - (tp->snd_max - tp->snd_una))), ("rack:%p left_to_send:%u sbavail:%u out:%u", rack, rack->r_ctl.fsb.left_to_send, sbavail(sb), @@ -18118,7 +18127,7 @@ out: * retransmit. In persist state, just set snd_max. */ if (error == 0) { - tcp_account_for_send(tp, len, (rsm != NULL), doing_tlp); + tcp_account_for_send(tp, len, (rsm != NULL), doing_tlp, hw_tls); rack->forced_ack = 0; /* If we send something zap the FA flag */ if (rsm && (doing_tlp == 0)) { /* Set we retransmitted */ @@ -18166,7 +18175,7 @@ out: } rack_log_output(tp, &to, len, rack_seq, (uint8_t) flags, error, rack_to_usec_ts(&tv), - rsm, add_flag, s_mb, s_moff); + rsm, add_flag, s_mb, s_moff, hw_tls); if ((error == 0) && @@ -18489,6 +18498,10 @@ enobufs: rack->r_ctl.fsb.o_m_len = rack->r_ctl.fsb.m->m_len; rack->r_ctl.fsb.tcp_flags = flags; rack->r_ctl.fsb.left_to_send = orig_len - len; + if (hw_tls) + rack->r_ctl.fsb.hw_tls = 1; + else + rack->r_ctl.fsb.hw_tls = 0; KASSERT((rack->r_ctl.fsb.left_to_send <= (sbavail(sb) - (tp->snd_max - tp->snd_una))), ("rack:%p left_to_send:%u sbavail:%u out:%u", rack, rack->r_ctl.fsb.left_to_send, sbavail(sb), @@ -18535,6 +18548,10 @@ enobufs: rack->r_ctl.fsb.o_m_len = rack->r_ctl.fsb.m->m_len; rack->r_ctl.fsb.tcp_flags = flags; rack->r_ctl.fsb.left_to_send = orig_len - len; + if (hw_tls) + rack->r_ctl.fsb.hw_tls = 1; + else + rack->r_ctl.fsb.hw_tls = 0; KASSERT((rack->r_ctl.fsb.left_to_send <= (sbavail(sb) - (tp->snd_max - tp->snd_una))), ("rack:%p left_to_send:%u sbavail:%u out:%u", rack, rack->r_ctl.fsb.left_to_send, sbavail(sb), @@ -19458,6 +19475,29 @@ rack_apply_deferred_options(struct tcp_rack *rack) } } +static void +rack_hw_tls_change(struct tcpcb *tp, int chg) +{ + /* + * HW tls state has changed.. fix all + * rsm's in flight. + */ + struct tcp_rack *rack; + struct rack_sendmap *rsm; + + rack = (struct tcp_rack *)tp->t_fb_ptr; + RB_FOREACH(rsm, rack_rb_tree_head, &rack->r_ctl.rc_mtree) { + if (chg) + rsm->r_hw_tls = 1; + else + rsm->r_hw_tls = 0; + } + if (chg) + rack->r_ctl.fsb.hw_tls = 1; + else + rack->r_ctl.fsb.hw_tls = 0; +} + static int rack_pru_options(struct tcpcb *tp, int flags) { @@ -19483,7 +19523,7 @@ static struct tcp_function_block __tcp_rack = { .tfb_tcp_handoff_ok = rack_handoff_ok, .tfb_tcp_mtu_chg = rack_mtu_change, .tfb_pru_options = rack_pru_options, - + .tfb_hwtls_change = rack_hw_tls_change, }; /* diff --git a/sys/netinet/tcp_stacks/tcp_rack.h b/sys/netinet/tcp_stacks/tcp_rack.h index 349f6daec2f4..54582d4fd9c5 100644 --- a/sys/netinet/tcp_stacks/tcp_rack.h +++ b/sys/netinet/tcp_stacks/tcp_rack.h @@ -68,7 +68,8 @@ struct rack_sendmap { uint8_t r_just_ret : 1, /* After sending, the next pkt was just returned, i.e. limited */ r_one_out_nr : 1, /* Special case 1 outstanding and not in recovery */ r_no_rtt_allowed : 1, /* No rtt measurement allowed */ - r_avail : 5; + r_hw_tls : 1, + r_avail : 4; uint64_t r_tim_lastsent[RACK_NUM_OF_RETRANS]; uint64_t r_ack_arrival; /* This is the time of ack-arrival (if SACK'd) */ RB_ENTRY(rack_sendmap) r_next; /* RB Tree next */ @@ -330,7 +331,8 @@ struct rack_fast_send_blk { struct mbuf *m; uint32_t o_m_len; uint32_t rfo_apply_push : 1, - unused : 31; + hw_tls : 1, + unused : 30; }; struct rack_control { diff --git a/sys/netinet/tcp_var.h b/sys/netinet/tcp_var.h index 553abaf6b334..dd30f89896d2 100644 --- a/sys/netinet/tcp_var.h +++ b/sys/netinet/tcp_var.h @@ -367,6 +367,7 @@ struct tcp_function_block { int (*tfb_tcp_handoff_ok)(struct tcpcb *); void (*tfb_tcp_mtu_chg)(struct tcpcb *); int (*tfb_pru_options)(struct tcpcb *, int); + void (*tfb_hwtls_change)(struct tcpcb *, int); volatile uint32_t tfb_refcnt; uint32_t tfb_flags; uint8_t tfb_id; @@ -1137,7 +1138,8 @@ tcp_fields_to_net(struct tcphdr *th) } static inline void -tcp_account_for_send(struct tcpcb *tp, uint32_t len, uint8_t is_rxt, uint8_t is_tlp) +tcp_account_for_send(struct tcpcb *tp, uint32_t len, uint8_t is_rxt, + uint8_t is_tlp, int hw_tls __unused) { if (is_tlp) { tp->t_sndtlppack++; @@ -1148,9 +1150,7 @@ tcp_account_for_send(struct tcpcb *tp, uint32_t len, uint8_t is_rxt, uint8_t is_ tp->t_snd_rxt_bytes += len; else tp->t_sndbytes += len; - } - #endif /* _KERNEL */ #endif /* _NETINET_TCP_VAR_H_ */