From mboxrd@z Thu Jan 1 00:00:00 1970 Authentication-Results: passt.top; dmarc=pass (p=quarantine dis=none) header.from=redhat.com Authentication-Results: passt.top; dkim=pass (1024-bit key; unprotected) header.d=redhat.com header.i=@redhat.com header.a=rsa-sha256 header.s=mimecast20190719 header.b=USsRba7D; dkim-atps=neutral Received: from us-smtp-delivery-124.mimecast.com (us-smtp-delivery-124.mimecast.com [170.10.133.124]) by passt.top (Postfix) with ESMTPS id 9746F5A026F for ; Wed, 29 Oct 2025 10:00:08 +0100 (CET) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=mimecast20190719; t=1761728407; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references; bh=b1P9qvtAKxXJzWjUgxRlOgVVekPyOO2B4xcZ9DNxERA=; b=USsRba7DLQsb7GP21xbcWU7mGEe2qAYLDoyPFShnxW8SO+kOw8Cl3wXQw2p4utucuq8Mfs Y7Es5/ZVPCzB2qAnlkp7TVmTrU5xlMPuDgC0pCfVCLsB0kT6MFSBUGkXsYC7E3+yxyR0rG /ambBA7D0mBgllK1hCf+5p2oxNgj3zg= Received: from mail-ed1-f72.google.com (mail-ed1-f72.google.com [209.85.208.72]) by relay.mimecast.com with ESMTP with STARTTLS (version=TLSv1.3, cipher=TLS_AES_256_GCM_SHA384) id us-mta-190-OEK8kqT1N7y4DnBDrVOK4A-1; Wed, 29 Oct 2025 05:00:05 -0400 X-MC-Unique: OEK8kqT1N7y4DnBDrVOK4A-1 X-Mimecast-MFC-AGG-ID: OEK8kqT1N7y4DnBDrVOK4A_1761728405 Received: by mail-ed1-f72.google.com with SMTP id 4fb4d7f45d1cf-63c3efeed6aso5760187a12.1 for ; Wed, 29 Oct 2025 02:00:05 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1761728405; x=1762333205; h=content-transfer-encoding:cc:to:subject:message-id:date:from :in-reply-to:references:mime-version:x-gm-message-state:from:to:cc :subject:date:message-id:reply-to; bh=b1P9qvtAKxXJzWjUgxRlOgVVekPyOO2B4xcZ9DNxERA=; b=nnluBNk89RMRs0HNZk9rbjYKKMpVxgZBOlkeSbat12hR+EHWS1yKpnhfSas5VwY6Ft /OHkMQmAXQ+BKPGn8hRSN09kUHoHH+9ElCLawUDxl4IYG+QR95gt7fsJnCrkeEyUKNjN iB0ILZ62sb55wWE+pVv94wSKkAaZFxVZGJXIrszBrl8JW15zdhBe7Le07sI0g8m8qvZD hCfyYAQms+4YtJh4tdUDQBbQb4rWMeogwMzCVpaTfnrrqRNkd4KRSotZcve4ZPSYZB2A XsZ9ZTRG0Swt4JKvTrX0cGGZuVXWu9QkXMHQ0szIbbgwzty8+bMhlovNGtWj5yeMnMVz REbQ== X-Gm-Message-State: AOJu0YyVh3K2t4JC4ZqKmen6ppTpJWFvbTQku+L+WWKLsLEkrpVddPEO dDsH/zBqMYmhJj2mF/nXSqe7XYG6rtfVNgmNXneGuM96Y7evKn36scBHNTAy5BWH/1T9TY78FzB 27P9q7NsnEOlo1L6bhdj4jylmyQ0J+OkrqbyCE2pY5w8XAJlJPIQZtrpiTImbmK958qQHREIHXD DamSh0sDqDku6v4cSTy9g1kRiGAHz/ X-Gm-Gg: ASbGncuq9yk4aFua1hQ9Rj4xiyGar9Jl+9yATR2cDVczcW4Bzo6fMhWQRRz4qfogIQb me1Mk8EmPIVrDU+e45MEmD+zc/dygcfjTvLkMr7DRecNUWsJkXIeAxgfqZfrL3XtSCWupDuklzp h+tD8xi73326BAOuNGQMAmdIi/cZelmHua5gney2GoFNbIT/8YbIoQaegQ X-Received: by 2002:a05:6402:1d4d:b0:634:9121:7a2d with SMTP id 4fb4d7f45d1cf-64044272518mr1513668a12.26.1761728404345; Wed, 29 Oct 2025 02:00:04 -0700 (PDT) X-Google-Smtp-Source: AGHT+IHQANOb53yGH291PeVvovyD3Zy8ijGBT207GX1o2f7CgxSkAktHOIapEtFowUc/9zXcJkEfOjY537YAjlhD8ms= X-Received: by 2002:a05:6402:1d4d:b0:634:9121:7a2d with SMTP id 4fb4d7f45d1cf-64044272518mr1513632a12.26.1761728403738; Wed, 29 Oct 2025 02:00:03 -0700 (PDT) MIME-Version: 1.0 References: <20251017062838.21041-1-yuhuang@redhat.com> <20251017062838.21041-5-yuhuang@redhat.com> <20251024010438.16e19757@elisabeth> <20251028124426.534fd236@elisabeth> <20251029053804.711716aa@elisabeth> <20251029080953.4f671f56@elisabeth> <20251029083929.0dc52bc6@elisabeth> In-Reply-To: <20251029083929.0dc52bc6@elisabeth> From: Yumei Huang Date: Wed, 29 Oct 2025 16:59:51 +0800 X-Gm-Features: AWmQ_bnUlJHjV2_y2JTMYwSetWfLzeD5kDQ_P0HHseoYfTquJDtRar0yj9S3E28 Message-ID: Subject: Re: [PATCH v6 4/4] tcp: Update data retransmission timeout To: Stefano Brivio X-Mimecast-Spam-Score: 0 X-Mimecast-MFC-PROC-ID: w2N2AN603q3hwgD1o8TtweHXy8t4kqRQgv62FXRmWUw_1761728405 X-Mimecast-Originator: redhat.com Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable Message-ID-Hash: C3ITWJ3YHCPHBOUDCMHEEXBBXYRINZ65 X-Message-ID-Hash: C3ITWJ3YHCPHBOUDCMHEEXBBXYRINZ65 X-MailFrom: yuhuang@redhat.com X-Mailman-Rule-Misses: dmarc-mitigation; no-senders; approved; emergency; loop; banned-address; member-moderation; nonmember-moderation; administrivia; implicit-dest; max-recipients; max-size; news-moderation; no-subject; digests; suspicious-header CC: passt-dev@passt.top, david@gibson.dropbear.id.au X-Mailman-Version: 3.3.8 Precedence: list List-Id: Development discussion and patches for passt Archived-At: Archived-At: List-Archive: List-Archive: List-Help: List-Owner: List-Post: List-Subscribe: List-Unsubscribe: On Wed, Oct 29, 2025 at 3:39=E2=80=AFPM Stefano Brivio = wrote: > > On Wed, 29 Oct 2025 15:32:33 +0800 > Yumei Huang wrote: > > > On Wed, Oct 29, 2025 at 3:10=E2=80=AFPM Stefano Brivio wrote: > > > > > > On Wed, 29 Oct 2025 13:11:48 +0800 > > > Yumei Huang wrote: > > > > > > > On Wed, Oct 29, 2025 at 12:38=E2=80=AFPM Stefano Brivio wrote: > > > > > > > > > > On Wed, 29 Oct 2025 11:06:44 +0800 > > > > > Yumei Huang wrote: > > > > > > > > > > > On Tue, Oct 28, 2025 at 7:44=E2=80=AFPM Stefano Brivio wrote: > > > > > > > > > > > > > > On Tue, 28 Oct 2025 16:09:03 +0800 > > > > > > > Yumei Huang wrote: > > > > > > > > > > > > > > > On Fri, Oct 24, 2025 at 7:04=E2=80=AFAM Stefano Brivio wrote: > > > > > > > > > > > > > > > > > > On Fri, 17 Oct 2025 14:28:38 +0800 > > > > > > > > > Yumei Huang wrote: > > > > > > > > > > > > > > > > > > > Use an exponential backoff timeout for data retransmiss= ion according > > > > > > > > > > to RFC 2988 and RFC 6298. Set the initial RTO to one se= cond as discussed > > > > > > > > > > in Appendix A of RFC 6298. > > > > > > > > > > > > > > > > > > > > Also combine the macros defining the initial RTO for bo= th SYN and ACK. > > > > > > > > > > > > > > > > > > > > Signed-off-by: Yumei Huang > > > > > > > > > > Reviewed-by: David Gibson > > > > > > > > > > --- > > > > > > > > > > tcp.c | 27 ++++++++++++--------------- > > > > > > > > > > 1 file changed, 12 insertions(+), 15 deletions(-) > > > > > > > > > > > > > > > > > > > > diff --git a/tcp.c b/tcp.c > > > > > > > > > > index 9385132..dc0ec6c 100644 > > > > > > > > > > --- a/tcp.c > > > > > > > > > > +++ b/tcp.c > > > > > > > > > > @@ -179,16 +179,14 @@ > > > > > > > > > > * > > > > > > > > > > * Timeouts are implemented by means of timerfd timers= , set based on flags: > > > > > > > > > > * > > > > > > > > > > - * - SYN_TIMEOUT_INIT: if no ACK is received from tap/= guest during handshake > > > > > > > > > > - * (flag ACK_FROM_TAP_DUE without ESTABLISHED event)= within this time, resend > > > > > > > > > > - * SYN. It's the starting timeout for the first SYN = retry. If this persists > > > > > > > > > > - * for more than TCP_MAX_RETRIES or (tcp_syn_retries= + > > > > > > > > > > - * tcp_syn_linear_timeouts) times in a row, reset th= e connection > > > > > > > > > > - * > > > > > > > > > > - * - ACK_TIMEOUT: if no ACK segment was received from = tap/guest, after sending > > > > > > > > > > - * data (flag ACK_FROM_TAP_DUE with ESTABLISHED even= t), re-send data from the > > > > > > > > > > - * socket and reset sequence to what was acknowledge= d. If this persists for > > > > > > > > > > - * more than TCP_MAX_RETRIES times in a row, reset t= he connection > > > > > > > > > > + * - RTO_INIT: if no ACK segment was received from tap= /guest, either during > > > > > > > > > > + * handshake (flag ACK_FROM_TAP_DUE without ESTABLIS= HED event) or after > > > > > > > > > > + * sending data (flag ACK_FROM_TAP_DUE with ESTABLIS= HED event), re-send data > > > > > > > > > > + * from the socket and reset sequence to what was ac= knowledged. This is the > > > > > > > > > > + * timeout for the first retry, in seconds. If this = persists too many times > > > > > > > > > > + * in a row, reset the connection: TCP_MAX_RETRIES f= or established > > > > > > > > > > + * connections, or (tcp_syn_retries + tcp_syn_linear= _timeouts) during the > > > > > > > > > > + * handshake. > > > > > > > > > > * > > > > > > > > > > * - FIN_TIMEOUT: if a FIN segment was sent to tap/gue= st (flag ACK_FROM_TAP_DUE > > > > > > > > > > * with TAP_FIN_SENT event), and no ACK is received = within this time, reset > > > > > > > > > > @@ -342,8 +340,7 @@ enum { > > > > > > > > > > #define WINDOW_DEFAULT 14600 = /* RFC 6928 */ > > > > > > > > > > > > > > > > > > > > #define ACK_INTERVAL 10 /= * ms */ > > > > > > > > > > -#define SYN_TIMEOUT_INIT 1 /= * s */ > > > > > > > > > > -#define ACK_TIMEOUT 2 > > > > > > > > > > +#define RTO_INIT 1 /= * s, RFC 6298 */ > > > > > > > > > > #define FIN_TIMEOUT 60 > > > > > > > > > > #define ACT_TIMEOUT 7200 > > > > > > > > > > > > > > > > > > > > @@ -588,13 +585,13 @@ static void tcp_timer_ctl(const s= truct ctx *c, struct tcp_tap_conn *conn) > > > > > > > > > > } else if (conn->flags & ACK_FROM_TAP_DUE) { > > > > > > > > > > if (!(conn->events & ESTABLISHED)) { > > > > > > > > > > if (conn->retries < c->tcp.syn_li= near_timeouts) > > > > > > > > > > - it.it_value.tv_sec =3D SY= N_TIMEOUT_INIT; > > > > > > > > > > + it.it_value.tv_sec =3D RT= O_INIT; > > > > > > > > > > else > > > > > > > > > > - it.it_value.tv_sec =3D SY= N_TIMEOUT_INIT << > > > > > > > > > > + it.it_value.tv_sec =3D RT= O_INIT << > > > > > > > > > > (conn->retries - = c->tcp.syn_linear_timeouts); > > > > > > > > > > } > > > > > > > > > > else > > > > > > > > > > - it.it_value.tv_sec =3D ACK_TIMEOU= T; > > > > > > > > > > + it.it_value.tv_sec =3D RTO_INIT <= < conn->retries; > > > > > > > > > > > > > > > > > > Same as on 3/4, but here it's clearly more convenient: ju= st assign > > > > > > > > > RTO_INIT, and multiply as needed in the if / else clauses= . > > > > > > > > > > > > > > > > I guess we can't just assign RTO_INIT. Maybe assign it onl= y when > > > > > > > > retries=3D=3D0, otherwise multiply as it.it_value.tv_sec <<= =3D1. > > > > > > > > > > > > > > Why can't you do that? Say: > > > > > > > > > > > > > > it.it_value.tv_sec =3D RTO_INIT > > > > > > > if (!(conn->events & ESTABLISHED)) { > > > > > > > if (conn->retries >=3D c->tcp.syn_linear_time= outs) > > > > > > > it.it_value.tv_sec <<=3D (conn->retri= es - > > > > > > > c->tcp.syn_li= near_timeouts); > > > > > > > > > > > > > > but anyway, see below. > > > > > > > > > > > > > > > But it seems more complicated. What do you think? > > > > > > > > > > > > > > Or maybe, building on my latest comment to 3/4: > > > > > > > > > > > > > > int factor =3D conn->retries; > > > > > > > > > > > > > > if (!(conn->events & ESTABLISHED)) > > > > > > > factor -=3D c->tcp.syn_linear_timeout= s; > > > > > > > > > > > > > > it.it_value.tv_sec =3D RTO_INIT << MAX(factor= , 0); > > > > > > > > > > > > > > ? > > > > > > > > > > > > Yeah, I understand this part now. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > } else if (CONN_HAS(conn, SOCK_FIN_SENT | TAP_FIN= _ACKED)) { > > > > > > > > > > it.it_value.tv_sec =3D FIN_TIMEOUT; > > > > > > > > > > } else { > > > > > > > > > > > > > > > > > > The rest of the series looks good to me. > > > > > > > > > > > > > > > > > > It might be slightly more practical to factor in directly= the RTO > > > > > > > > > clamp, and I don't think it's complicated now that you ha= ve the helper > > > > > > > > > from 2/4, but it's not a strong preference from my side, = as the series > > > > > > > > > makes sense in any case. > > > > > > > > > > > > > > > > Reading tcp_rto_max_ms can be easy with the helper. My conc= ern is > > > > > > > > about the way we get the total time for retries. > > > > > > > > > > > > > > > > I used to do it like this in v2, > > > > > > > > https://archives.passt.top/passt-dev/20251010074700.22177-4= -yuhuang@redhat.com/: > > > > > > > > > > > > > > > > +#define RETRY_ELAPSED(timeout_init, retries) \ > > > > > > > > + ((timeout_init) * ((1 << ((retries) + 1)) - 2)) > > > > > > > > > > > > > > > > Though the formula is not quite right, we could refine it a= s below: > > > > > > > > > > > > > > > > #define RETRY_ELAPSED(retries) ((RTO_INIT) * ((1 << ((retri= es) + 1)) - 1)) > > > > > > > > > > > > > > > > Does it make sense to get the time this way? > > > > > > > > > > > > > > Well, it also depends on c->tcp.syn_linear_timeouts, right? > > > > > > > > > > > > Not really, it's only used for data retransmission, so > > > > > > syn_linear_timeouts is not relevant. > > > > > > > > > > Hmm, no, why? RFC 6298 covers SYN retries as well, and that's the= one > > > > > stating: > > > > > > > > I meant RETRY_ELAPSED was only used for data retransmission, which > > > > uses exponential backoff timeout directly, so syn_linear_timeouts w= as > > > > not relevant. > > > > > > Ah, okay. > > > > > > > > > > > > > (2.5) A maximum value MAY be placed on RTO provided it is at l= east 60 > > > > > seconds. > > > > > > > > For SYN retries, as we used linear backoff + exponential backoff, a= nd > > > > also limited by TCP_MAX_RETRIES, the possible max RTO is far less t= han > > > > 60s. So we didn't clamp it. Do you think we need to clamp it as wel= l? > > > > > > If syn_linear_timeouts is 0 and tcp_syn_retries is 7, I guess we'll > > > reach 2^0 + 2^1 + 2^2 + 2^3 + 2^4 + 2^5 + 2^6 + 2^7 =3D 247 seconds? = Or > > > just up to ... + 2^6, that is, 119 seconds? > > > > > > In any case, what is the difference compared to data retransmissions? > > > > You are right. I assumed wrongly that syn_linear_timeouts is always 4. > > When it's 0, it's the same as data retransmissions. > > > > > > Don't we have 3 bits to store the retry count as well, so we're limit= ed > > > by TCP_MAX_RETRIES anyway? Looking at patch 1/4 I'd say it's the same > > > counter. > > > > Yes, it's the same counter. I guess you mean we should clamp it as well= . > > I didn't check this part, I thought we already did, but if we don't, we > should do that, yes. > > > > > > ...the only thing that I don't see implemented in this version of= the > > > > > patch is paragraph 5.7: > > > > > > > > > > (5.7) If the timer expires awaiting the ACK of a SYN segment a= nd the > > > > > TCP implementation is using an RTO less than 3 seconds, = the RTO > > > > > MUST be re-initialized to 3 seconds when data transmissi= on > > > > > begins (i.e., after the three-way handshake completes). > > > > > > > > > > I missed that while reviewing earlier versions. I guess we need t= o use > > > > > a MAX(x, 3) clamp if (c->conn->events & ESTABLISHED). I think it'= s > > > > > simpler than re-introducing separate starting values (one second = and > > > > > three seconds). Sorry I'd like to confirm one more thing. By "use MAX(x, 3) clamp if (c->conn->events & ESTABLISHED)", I guess you mean: it.it_value.tv_sec =3D MAX(it.it_value.tv_sec, 3); Then the code would be: it.it_value.tv_sec =3D RTO_INIT; if (!(conn->events & ESTABLISHED)) { int exp =3D conn->retries - c->tcp.syn_linear_timeouts; it.it_value.tv_sec <<=3D MAX(exp, 0); } else { it.it_value.tv_sec =3D MAX(it.it_value.tv_sec, 3); it.it_value.tv_sec <<=3D conn->retries; } it.it_value.tv_sec =3D MIN( it.it_value.tv_sec, c->tcp.tcp_rto_max); we basically set the starting value to 3 for data retransmissions no matter if we have retried SYN. And the max total timeout would be 3+6+12+24+48+96+120+120 =3D 429s. Is it what we want? Besides, I guess I need to define a macro for "3" as well, like "ACK_TIMEOUT_INIT"? > > > > > > > > I'm not sure I understand here. If the timer expires, didn't we res= et > > > > the connection directly? We would never get to the data transmissio= n > > > > phase? > > > > > > In the RFC, "the timer expires" indicates one iteration of the timeou= t > > > algorithm, that is, it's the same as our timer (tcp_timer_handler()) > > > triggering. > > > > I see. > > > > If you agree, I can add a new patch in this series to address the clamp= ing. > > I don't really have a preference, I guess it could be directly in 4/4 or > in a separate patch, neither option should complicate things too much. > > > > The paragraph says "after the three-way handshake completes", so it > > > looks like the connection wasn't reset. > > > > > > > > > Probably I should name it more clearly. > > > > > > > > > > > > > > > > > > > > But in any case, do you really need to calculate that explici= tly? I was > > > > > > > thinking that you can just clamp the value when you use it in > > > > > > > tcp_timer_ctl(). > > > > > > > > > > > > > > If c->tcp.rto_max is DIV_ROUND_CLOSEST(x, 1000), where 'x' is= the value > > > > > > > you read from the kernel, then I guess it's just: > > > > > > > > > > > > > > --- > > > > > > > it.it_value.tv_sec =3D MIN(it.it_value.tv_sec= , c->tcp.rto_max); > > > > > > > > > > > > After reading the comments in v3 when tcp_rto_max_ms was first > > > > > > mentioned again, I realized I got something wrong again. I thou= ght it > > > > > > was for the total timeout for all retries, so I need to calcula= te that > > > > > > and decide to reset the connection as in v2. > > > > > > > > > > I think it actually applies to all the retries. > > > > > > > > > > > Anyway, you are right. We don't need to do that. Thanks for you= r patience. > > > > > > > > > > > > > } ... > > > > > > > > > > > > > > if (timerfd_settime(conn->timer, 0, &it, NULL)) > > > > > > > flow_perror(conn, "failed to set timer"); > > > > > > > --- > > -- > Stefano > --=20 Thanks, Yumei Huang