Re: [PROPOSAL] Remove P_ or I_ naming convention for header files

Hi Masaori, The functionality does not change as the file name changes： - The P_ files should contain any types and definitions that are private to the subsystem, - The public interface should be contained in a I_ prefixed header. Confusing private and public definitions will cause the code to h

Re: dev@trafficserver.apache.org

Hi Zhiyong, Do you want to decrypt the SSL traffic in forward proxy with ATS ? - Oknet Sudheer Vinukonda 于2019年7月19日周五 上午3:24写道： > > Hi Zhiyong, > SSL termination for ATS is not any different when ATS is being used as a > forward vs reverse proxy. > This might be a good start to configure SSL

Re: dev@trafficserver.apache.org

ATS doesn't have the SSL Man-In-The-Middle feature to decrypt SSL traffic in forward proxy mode. In order to implement the MITM for ATS in forward proxy mode, you should do below things at least: 1. Generate your self CA key and cert 2. Add your CA cert into all the clients behind the forward pro

Re: null_transform.c

Hi Walt, A1: The `TransformVConnection` is a pipe/chain which consisted by multiple `INKVConnInternal` and a `TransformTerminus` at the end. The Transform Plugin is a Continuation which will be called back by `INKVConnInternal`. At least one transform plugin hooked into a HttpSM, 1. The TXN may

My idea about PR#6910

Hi Susan, 1. Do not disable polling for NetVC. - SSLNetVC read data into rBIO and call SSL_accept(). - We have to fill the rBIO by polling READ event on NetVC. - Add a new `EventIO trigger_ep;` for SSLNetVC. 2. Do not share the EventIO with socket fd and openssl wait fd 3. Handle SSL_ERROR_WANT_

Question about NetHandler::mainNetEvent and PollCont::pollEvent

Hi, All! I'm looking into the iocore source code reading now. I found the source code from PollCont::pollEvent() is a part of NetHandler::mainNetEvent() . And the PollCont is designed to as a abstract class for I/O polling operation. so I have a question: why we duplicate the code in mainNetEve

Re: Question about NetHandler::mainNetEvent and PollCont::pollEvent

e PollCont > entirely. As far as I can tell it exists only to hold the poll descriptor > data which could easily be promoted to NetHandhandler. > > > > On Monday, November 2, 2015 8:14 AM, Chao Xu > wrote: > > > Hi, All! > > I'm looking into the

Re: Question about NetHandler::mainNetEvent and PollCont::pollEvent

g which means all of the > UDP support could be removed as well. There has been some talk of QUIC > support but I suspect that would either be a separate handler (ala SPDY and > HTTP/2) or folded in to NetHandler via the work being done for TS-3612. > > > On Tuesday, Nov

ATS Internals（source reading / ATS源代码阅读） in Chinese

Dear All, I’m writting a document to introduce the event system, io core, proxy sm, etc ... The event system section has been done at last weekend. It can be found at : https://github.com/oknet/atsinternals . Hope it is useful to ats developers. And I'm drawing some interesting flow charts, li

a question about mutex of NetHandler

the NetHandler is init at "void initialize_thread_for_net(EThread *thread)". {code} 260 void 261 initialize_thread_for_net(EThread *thread) 262 { 263 new ((ink_dummy_for_new *)get_NetHandler(thread)) NetHandler(); 264 new ((ink_dummy_for_new *)get_PollCont(thread)) PollCont(thread->mutex, get_

maybe wrong retval on ssl_callback_session_ticket with enc==1

Dear All, I met a ticket reuse problem, and I walking into the ssl_callback_session_ticket in file SSLUtils.cc. from the openssl document: https://www.openssl.org/docs/manmaster/ssl/SSL_CTX_set_tlsext_ticket_key_cb.html The return value of the cb function is used by OpenSSL to determine what fu

Question on wbe_event of write_to_net_io() .

Dear All, I'm reading the ATS code now, there are 2 variables: "signalled" and "wbe_event", the comment: "signalled means we won't send an event" {code} } else if (signalled && (wbe_event != vc->write_buffer_empty_event)) { // @a signalled means we won't send an event, and the event v

Proposal: NetVC Context

As a proxy, there has two side: client side ( Client <-> Proxy ) and server side ( Proxy <-> Server ). Add a new member 'netvc_context' into class NetVConnection to indicate which side the NetVC working on. And remove member 'server_addr' from UnixNetVConnection: {code} 246 // amc - what is th

Re: Proposal: NetVC Context

ddress: we can call CliNetVC->get_client_addr() if we want get client ip in a server side. we can call SvrNetVC->get_server_addr() if we want get original server ip in a client side. 2016-04-25 4:18 GMT+08:00 James Peach : > > > On Apr 22, 2016, at 7:28 PM, Chao Xu wrote: >

Re: Proposal: NetVC Context

0 James Peach : > > > On Apr 22, 2016, at 7:28 PM, Chao Xu wrote: > > > > As a proxy, there has two side: client side ( Client <-> Proxy ) and > server > > side ( Proxy <-> Server ). > > Each side would be a separate NetVConnection. > > > Add a new

Re: Proposal: NetVC Context

Hi All, I have opened a pull request about the NetVC Context proposal to evaluate. Thanks Oknet 2016-04-23 10:28 GMT+08:00 Chao Xu : > As a proxy, there has two side: client side ( Client <-> Proxy ) and > server side ( Proxy <-> Server ). > > Add a new member &#

Re: SSLNetVConnection IOBufferBlock management issue

it is return to net_read_io() from ssl_read_from_net() when the MIOBuffer is full for write. and net_read_io will callback to HttpSM with EVENT_READ_READY to consume the data in MIOBuffer and reenable ReadVIO. then do readReschedule() after callback return from HttpSM. Only read at most 2 blocks,

Re: TS_VCONN_PRE_ACCEPT_HOOK and TS_SSL_SNI_HOOK hooks

Do you try set action=tunnel in ssl_multicert.config ? # action=[tunnel] # If the tunnel matches this line, traffic server will not participate # in the handshake. But rather it will blind tunnel the SSL connection. # If the connection is identified by server name, an openSSL patch must #

Re: TS_VCONN_PRE_ACCEPT_HOOK and TS_SSL_SNI_HOOK hooks

sorry, I did not try the feature, but I think Susan maybe known it in detail. 2016-07-18 12:44 GMT+08:00 James Peach : > > > On Jul 15, 2016, at 4:20 PM, Chao Xu wrote: > > > > Do you try set action=tunnel in ssl_multicert.config ? > > > > # action=[tunnel] &

Is it correct to include a header file that is named with prefix P_ in proxy directory ?

cc:#include "P_Cache.h" logging/LogBuffer.cc:#include "P_EventSystem.h" logging/LogCollationAccept.cc:#include "P_EventSystem.h" logging/LogCollationAccept.h:#include "P_EventSystem.h" logging/LogCollationAccept.h:#include "P_Net.h" logging/LogCollationClientSM.cc:#include "P_EventSystem.h" logging/LogCollationClientSM.cc:#include "P_Net.h" logging/LogCollationClientSM.h:#include "P_HostDB.h" logging/LogCollationClientSM.h:#include "P_Net.h" logging/LogCollationHostSM.cc:#include "P_EventSystem.h" logging/LogCollationHostSM.cc:#include "P_Net.h" logging/LogCollationHostSM.h:#include "P_EventSystem.h" logging/LogConfig.h:#include "P_RecProcess.h" logging/LogFile.cc:#include "P_EventSystem.h" logging/LogObject.cc:#include "P_EventSystem.h" logging/LogSock.cc:#include "P_EventSystem.h" logging/LogStandalone.cc:#include "P_EventSystem.h" logging/LogStandalone.cc:#include "P_RecProcess.h" logging/LogUtils.cc:#include "P_RecProcess.h" sac.cc:#include "P_Net.h" shared/DiagsConfig.cc:#include "P_RecCore.h" shared/UglyLogStubs.cc:#include "P_Net.h" spdy/SpdyCommon.h:#include "P_Net.h" spdy/SpdySessionAccept.h:#include "P_Net.h" spdy/SpdySessionAccept.h:#include "P_EventSystem.h" spdy/SpdySessionAccept.h:#include "P_UnixNet.h" Is it correct? Thanks! B/R Chao Xu

Re: [PROPOSAL] Removing SOCKS support in 7.0.0

The Socks Proxy feature is used to route HTTP request in socks request to HttpSM and Tunnel non-Http request to Socks Server. It is a Socks Proxy and it is NOT the replacement of Socks Server. 2016-08-21 10:04 GMT+08:00 James Peach : > > > On Aug 19, 2016, at 11:33 AM, Bryan Call wrote: > > > >

Re: [PROPOSAL] Removing SOCKS support in 7.0.0

ocks proxy functions. https://drive.google.com/open?id=0B6FSuL4tHtSHb2o2dnNTdXNMN0E https://drive.google.com/open?id=0B6FSuL4tHtSHTDFESl9RWmNjQkU 2016-08-22 22:38 GMT+08:00 James Peach : > > > On Aug 22, 2016, at 3:40 AM, Chao Xu wrote: > > > > The Socks Proxy feature is used

Re: [PROPOSAL] Removing SOCKS support in 7.0.0

e speak up and vote > on keeping it. If you are willing to fix bugs on the feature than that is > even better! > > -Bryan > > > > > > On Aug 23, 2016, at 2:04 AM, Chao Xu wrote: > > > > I have do some analysis on socks proxy code, It is a very simple

Re: RFC: convert from NULL tp nullptr

+1 2016-10-17 1:51 GMT+08:00 James Peach : > > > On Oct 16, 2016, at 9:18 AM, Leif Hedstrom wrote: > > > > I'll try to review next week, but I'm +1 on the concept. I assume we > have places were we use 0 but really should use nullptr? > > The only weird think I found is that we used NULL for an

Update the NetVC route graph

Hi buddy, I have update the diagram how the netvc route in the Net Sub-System. ​ CH02-IOCoreNet-002.png ​ Oknet

Re: Current status of WCCP support

I tried to setup WCCP with ATS 6.0.x but it did not work correctly. I really need it work correctly in Q1, 2017. It will be great if you could share any document or guidelines to support me. BTW, do you known which ATS version the WCCP worked correctly ? Thanks. Oknet Xu 2016-11-29 22:38 GMT+

Re: Current status of WCCP support

ere is no > ongoing development. It is still just WCCP V2 because, AFAIK, WCCP V3 was > never released. I haven't used it for a while but I don't see what would > have broken it. Did you (1) --enable-wccp and (2) install a WCCP > configuration file for ATS? > > >

Re: Current status of WCCP support

004 The traffic_manager and traffic_wccp is different. It seems traffic_manager not working properly. I will try "ip wccp web-cache" with "STANDARD" in wccp.config later. Oknet Xu 2016-11-30 18:46 GMT+08:00 Chao Xu : > Hi Alan, > > 1) ./configure with "-

Re: Current status of WCCP support

1 23:20 GMT+08:00 Alan Carroll < solidwallofc...@yahoo-inc.com.invalid>: > That's very odd, since traffic_wccp and traffic_manager link literally the > same code to perform WCCP operations. Any chance you can get a packet > capture of the two HERE_I_AM packets for comparison? > >

Re: [VOTE] Release Apache Traffic Server 6.2.1 (RC0)

Could you merge PR#1276 TS-5015 into ? It fix a socks bug, was reported by 06chaynes@github ( https://github.com/apache/trafficserver/issues/1175). Oknet Xu 2016-12-28 1:30 GMT+08:00 Phil Sorber : > Hello All, > > I've prepared a release for v6.2.1 (RC0) which is the latest stable release > in

Re: [VOTE] Release Apache Traffic Server 6.2.1 (RC0)

It is not a -1 vote for 6.2.1. But I hope you can merge the socks bug fix early. The socks proxy not working without PR#1276. Oknet 2016-12-28 23:43 GMT+08:00 Phil Sorber : > Are you -1'ing the vote, or is this something that can go into 6.2.2? > > On Wed, Dec 28, 2016 at 7

Re: Status of ATS v7.1.0 (or lack thereof)

I found POST Transform can not working with server vc reconnect or retry. ATS would crash randomly if server vc got failed after POST Transform finished. Can you finger out whether or not the POST Transform is enabled in those above issues ? And the POST Transform could working with chunked pos

Question about TransformTerminus::handle_event()

amc, leif and bcall, I have a question about the code in TransformTerminus::handle_event() while I'm reading the ats code. file: proxy/Transform.cc ``` 141 int 142 TransformTerminus::handle_event(int event, void * /* edata ATS_UNUSED */) 143 { ... 158 if (m_closed != 0 && m_tvc->m_closed !=

Re: Last call: any PRs for 7.1.0 release?

Hi Leif, I think that you forgot to revert this. commit 1450456d609fa37a0b831717dc826c5e759dafe9 Author: Leif Hedstrom Date: Wed Apr 12 15:21:08 2017 -0700 Revert "TS-4895: Initialize default_inactivity_timeout (CID 1021743 from TS-4612)" This reverts commit 9108bc942d1f0d141ec3de008

Re: Last call: any PRs for 7.1.0 release?

Hi Leif, https://github.com/apache/trafficserver/pull/1765 Hope not too late Oknet Xu 2017-05-03 10:24 GMT+08:00 Leif Hedstrom : > > > On May 2, 2017, at 10:28 AM, Bryan Call wrote: > > > > I have one minor cherry pick that would be helpful not to maintain: > > > > commit 80dd6702931e175230cd

Re: Setting fds_limit.

Hi Peter, The `fds_limit` is hard limit of fds. The 192 fds are reserved for ats internal use (epoll fd , EThread signal, etc ...). The throttle is soft limit of fds and (fds_limit - 192) is the max value of throttle. We can not and must not change the hard limit since it is set by operating syst

Re: [API PROPOSAL] TSIpAddrParse

Why we don't expose the ts/lib for plugins ? I think the ts/lib is an alternative for libstdc++. Oknet Xu 2017-09-15 22:29 GMT+08:00 Alan Carroll : > Well, my goal has been to, at some point, make all of ink_inet.h available > to plugins. This seems a reasonable start. > > On Sep 14, 2017 6:09

Fwd: Talk about the QUICNetVConnection

the QUICNetVC, it will be managed by NetHandler just like SSLNetVC. [image: 内嵌图片 3] Oknet Xu -- Forwarded message -- From: Masaori Koshiba Date: 2017-10-13 20:01 GMT+08:00 Subject: Re: Talk about the QUICNetVConnection To: Chao Xu , Masakazu Kitajo , sc...@apache.org Would you

Re: Talk about the QUICNetVConnection

Here is the url for the diagram : http://home.apache.org/~oknet/ QUIC.201710171429.png 2017-10-17 14:33 GMT+08:00 Chao Xu : > Here is the url for the diagram : http://home.apache.org/~oknet/ > QUIC.201710171429.png > > 2017-10-17 14:25 GMT+08:00 Chao Xu : > >> /cc dev &g

Re: Talk about the QUICNetVConnection

(tls-mapping) says > handshake use stream 0. > So handshake needs to be build on frame reordering (stream) and loss > detection. This is why current quic-latest branch implements QUICHandshake > as a QUIC Application. > > - Masaori > > I'm glad to hear that you'll com

About the code review for UDP Core

Hi dev, I found some issues and will push PRs to resolve them. Pushed 3 PRs first, please review: https://github.com/apache/trafficserver/pull/2716 https://github.com/apache/trafficserver/pull/2718 https://github.com/apache/trafficserver/pull/2719 If you have some issues which is UDP related, p

Re: Talk about the QUICNetVConnection

With PR #2719 : http://people.apache.org/~oknet/quic_and_udp.2017103118.png 2017-10-21 10:23 GMT+08:00 Chao Xu : > Here is a diagram for UDP core : http://people.apache.org/~ > oknet/quic_and_udp.2017102020.png > > - Oknet > > 2017-10-17 15:33 GMT+08:00 Masaori Koshiba :

Re: Talk about the QUICNetVConnection

Hi Leif, May I create a PR for the diagrams ? I prefer the name: iocore/net/quic_and_udp.png. - Oknet 2017-11-02 5:11 GMT+08:00 Leif Hedstrom : > > > > On Oct 31, 2017, at 4:38 AM, Chao Xu wrote: > > > > With PR #2719 : http://people.apache.org/~oknet/quic_a

Is it a mistake of openssl docs ?

Hi dev, https://www.openssl.org/docs/man1.1.0/crypto/EVP_CIPHER_CTX_ctrl.html EVP_CIPHER_CTX_ctrl(ctx, EVP_CTRL_AEAD_SET_TAG, taglen, tag); Sets the expected tag to taglen bytes from tag. This call is only legal when decrypting data and must be made before any data is processed (e.g. before any

Re: Is it a mistake of openssl docs ?

question to https://github.com/openssl/openssl/issues/1624 > Also it looks like the latest docs has been changed. > > https://www.openssl.org/docs/manmaster/man3/EVP_ > EncryptInit.html#GCM-and-OCB-Modes > > - Masaori > > 2017年11月3日(金) 15:24 Chao Xu : > > > Hi dev, &g

Max QUIC Packet length and MSS

Hi Masaori , Maskit , Scw00 Is it a RFC mistake ? The max size of QUIC Packet is 1280 but the default MSS for loose detection is 1460. According to https://quicwg.github.io/base-drafts/draft-ietf-quic-transport.html#rfc.section.9 ``` 9.

Re: Change to HttpSM history

Hi AMC, It could get confused if circle buffer overwrote. Therefore, in my opinion, linear buffer is better than circular buffer. Oknet 2017-11-10 7:05 GMT+08:00 scw00 <616955...@qq.com>: > Hi AMC > > > Yes I’ve changed history in the linear buffer because the bug is not > always related to the

Re: [API Proposal] - TSVConnArgs

Hi AMC, " We should rename TS_VCONN_PRE_ACCEPT_HOOK to TS_VCONN_START_HOOK. " IMO, TS_VCONN_OPENED_HOOK when the OS connection is established. TS_VCONN_ACCEPTED_HOOK as a instead for TS_VCONN_PRE_ACCEPT_HOOK. - Oknet 2017-11-14 23:48 GMT+08:00 Dk Jack : > I concur with the idea that connection

Re: [API Proposal] - TSVConnArgs

, TS_VCONN_SSL_LAST_HOOK = TS_VCONN_OPENED_HOOK } TSVConnHookID; ``` - Oknet 2017-11-15 22:58 GMT+08:00 Chao Xu : > Hi AMC, > > " We should rename TS_VCONN_PRE_ACCEPT_HOOK to TS_VCONN_START_HOOK. " > > IMO, TS_VCONN_OPENED_HOOK when the OS connection is established. > TS_VCONN_AC

Re: [API Proposal] - TSVConnArgs

which is the event that signals that action. > > On Wed, Nov 15, 2017 at 8:58 AM, Chao Xu wrote: > > > Hi AMC, > > > > " We should rename TS_VCONN_PRE_ACCEPT_HOOK to TS_VCONN_START_HOOK. " > > > > IMO, TS_VCONN_OPENED_HOOK when the OS connection i

Re: [API Proposal] - TSVConnArgs

Yes, Origin Server. :-) TS_VCONN_CONNECTED_HOOK maybe more accurate than TS_VCONN_OPENED_HOOK for "once a connection is established" - Oknet 2017-11-15 23:08 GMT+08:00 Alan Carroll : > Ah, by "OS" you mean "Origin Server", not "Operating System". &

A diagram of QUIC Handshake.

Hi All To help you understand the handshake of QUIC, I draw a diagram which is similar to TCP 3-ways handshake. http://people.apache.org/~oknet/quic_handshake.2017112315.png Please email me if there are any questions/mistakes. - Oknet Xu

Re: A diagram of QUIC Handshake.

Client Address Validation" is optional during QUIC Handshake progress. - Oknet Xu 2017-11-23 15:25 GMT+08:00 Chao Xu : > Hi All > > To help you understand the handshake of QUIC, I draw a diagram which is > similar to TCP 3-ways handshake. > > http://people.apache.org/~okne

Re: [VOTE] Release Apache Traffic Server 7.1.2 (RC1)

+1 2017-12-20 14:04 GMT+08:00 Leif Hedstrom : > +1 from me, tested on 10 prod boxes with no issues. > > — Leif > > > On Dec 19, 2017, at 10:03 PM, Leif Hedstrom wrote: > > > > I've prepared a release for 7.1.2 (RC1), which is a bug fix release on > the previous v7.1.1. The release notes for 7.1.

Re: Crash - looks like 1765 -

Hi salil, please try to apply https://github.com/apache/trafficserver/commit/49c903b0c13a2fd2c085325f6942326c4f9c926a . - Oknet 2018-05-14 22:03 GMT+08:00 salil GK : > Hello > >I am using ATS 6.2.2 and found a crash happening in > TransformVConnection::do_io_close:467. This looks like bug 1

Re: [DISCUSS] Time to EOL 6.2.x?

Hi Leif, I still working on 6.0.x branch and pushed a PR#3539 for 6.2.x recently. The PR#3539 fix a bug for hostdb from 2.0.x ~ 6.2.x. :-D - Oknet 2018-05-30 7:40 GMT+08:00 Leif Hedstrom : > > > > On May 29, 2018, at 17:10, salil GK wrote: > > > > We too are in 6.2.2 and plans are in place t

The cache rules with dest_ip has no effect

Hi dev The dest_ip is filled after DNS lookup. But the cache rules performed before the DNS lookup. It makes the dest_ip has no effect. Is it a bug ? or by design ? -- - Oknet Xu

Re: TSHttpHookAdd() thread safety

The plugins are loaded and initialized within Main.cc by calling plugin_init() one by one. The global hooks list is initialized before any http port / service is ready to serve. There is no thread issue because all the threads are blocked for incoming events. - oknet Walt Karas 于2018年12月13日周四

Re: Mistake for delete some branches

I have a recently copy. bogon:oknet-ats oknet$ git branch -a | grep remotes/official remotes/official/2.0.x remotes/official/3.0.x remotes/official/3.2.x remotes/official/3.3.x remotes/official/4.0.x remotes/official/4.1.x remotes/official/4.2.1.x remotes/official/4.2.x remotes/o

Re: Mistake for delete some branches

make_hwloc_required quic-12 quic-13 quic-15 - oknet Chao Xu 于2019年2月15日周五 下午3:37写道： > > I have a recently copy. > > bogon:oknet-ats oknet$ git branch -a | grep remotes/official > remotes/official/2.0.x > remotes/official/3.0.x > remotes/official/3.2.x >