Re: [vpp-dev] 100G with iperf3 server using VCL library

Hi Vijay, Cool experiment. More inline. > On Sep 11, 2020, at 9:42 AM, Vijay Sampath wrote: > > Hi, > > I am using iperf3 as a client on an Ubuntu 18.04 Linux machine connected to > another server running VPP using 100G NICs. Both servers are Intel Xeon with > 24 cores. May I ask the freq

Re: [vpp-dev] 100G with iperf3 server using VCL library

Hi Vijay, Inline. > On Sep 11, 2020, at 1:08 PM, Vijay Sampath wrote: > > Hi Florin, > > Thanks for the response. Please see inline: > > On Fri, Sep 11, 2020 at 10:42 AM Florin Coras <mailto:fcoras.li...@gmail.com>> wrote: > Hi Vijay, > > Cool exp

Re: [vpp-dev] 100G with iperf3 server using VCL library

Hi Vijay, Quick replies inline. > On Sep 11, 2020, at 7:27 PM, Vijay Sampath wrote: > > Hi Florin, > > Thanks once again for looking at this issue. Please see inline: > > On Fri, Sep 11, 2020 at 2:06 PM Florin Coras <mailto:fcoras.li...@gmail.com>> w

Re: [vpp-dev] 100G with iperf3 server using VCL library

Hi Vijay, > On Sep 12, 2020, at 10:06 AM, Vijay Sampath wrote: > > Hi Florin, > > On Fri, Sep 11, 2020 at 11:23 PM Florin Coras <mailto:fcoras.li...@gmail.com>> wrote: > Hi Vijay, > > Quick replies inline. > >> On Sep 11, 2020, at 7:27 PM,

Re: [vpp-dev] 100G with iperf3 server using VCL library

Hi VIjay, > On Sep 12, 2020, at 12:06 PM, Vijay Sampath wrote: > > Hi Florin, > > On Sat, Sep 12, 2020 at 11:44 AM Florin Coras <mailto:fcoras.li...@gmail.com>> wrote: > Hi Vijay, > > >> On Sep 12, 2020, at 10:06 AM, Vijay Sampath > <mailto

Re: [vpp-dev] 100G with iperf3 server using VCL library

3G > for the same numa core vs different. > > Thanks, > > Vijay > > On Sat, Sep 12, 2020 at 2:40 PM Florin Coras <mailto:fcoras.li...@gmail.com>> wrote: > Hi VIjay, > > >> On Sep 12, 2020, at 12:06 PM, Vijay Sampath > <mailto:vsamp.

Re: [vpp-dev] 100G with iperf3 server using VCL library

py. VPP delivers the data to iperf3 in a shared memory fifo, but iperf3 then wants it copied into its own buffer. Unfortunately, we can’t optimize this without changing the apps. Regards, Florin > > Thanks, > > Vijay > > On Mon, Sep 14, 2020 at 8:30 PM Florin Coras <mailto:fc

Re: [vpp-dev] 100G with iperf3 server using VCL library

ry from test to test. But I was never able to get beyond > 50G with 10connections with iperf on the other numa node. As I mentioned in > the previous email, when I repeat this test with Linux TCP as the server, I > am able to get 100G no matter which cores I start iperf on. > > Thanks

Re: [vpp-dev] 100G with iperf3 server using VCL library

ions. Thanks for all the help. I have a much better idea now. > > Thanks, > > Vijay > > On Tue, Sep 15, 2020 at 11:25 AM Florin Coras <mailto:fcoras.li...@gmail.com>> wrote: > Hi Vijay, > > Yes, that is the case for this iperf3 test. The data is already

Re: [vpp-dev] 100G with iperf3 server using VCL library

ided completely, but the applications have to still reside > as a separate process. > > Thanks, > > Vijay > > On Tue, Sep 15, 2020 at 11:44 AM Florin Coras <mailto:fcoras.li...@gmail.com>> wrote: > Hi Vijay, > > Currently, builtin applications can

Re: [vpp-dev] 100G with iperf3 server using VCL library

not yet present, > but can be added with little effort. Is that correct? > > Thanks, > > Vijay > > On Tue, Sep 15, 2020 at 1:31 PM Florin Coras <mailto:fcoras.li...@gmail.com>> wrote: > Hi Vijay, > > Oh, by no means. Builtin applications, i.e., applicati

Re: [vpp-dev] In the host stack, TCP cannot receive data less than 6 bytes？

Hi, TCP coalesces both when sending and when receiving. If the sender really sends only every second a few bytes, those should be the only ones delivered, assuming the receiver reads the data as soon as it was received. To check that everything behaves as expected, try sending only 2 bytes and

Re: [vpp-dev] Distributing VPPCOM TX sessions to different worker thread #vnet #vppcom #TxBuff

Hi Ravi, I assume you’re using vppcom_session_connect to open those udp sessions? If the listeners receive incoming data, make sure you also set the VPPCOM_ATTR_SET_CONNECTED on the listener with vppcom_session_attr() before calling listen. By default udp sessions are opened on thread 1 and s

Re: [vpp-dev] Distributing VPPCOM TX sessions to different worker thread #vnet #vppcom #TxBuff

Hi Ravi, At this time, there’s no option to pin connected udp sessions to workers, instead we default to worker 1. You could try changing the logic here [1] to select a random thread for each connection. We’re not doing that because whenever we receive return traffic we pin the session to the

Re: [vpp-dev] VPP committers: VPP PTL vote

+1 Florin > On Sep 25, 2020, at 12:13 PM, Dave Barach via lists.fd.io > wrote: > > Folks, > > The self-nomination period closed yesterday. We had one self-nomination, from > Damjan Marion. At this point, we can proceed with a vote. > > I’m sure that Damjan will do a great job, so let me start

Re: [vpp-committers] [vpp-dev] VPP committers: VPP PTL vote

Congrats, Damjan!! Those are some huge “shoes" you’ll have to fill but I’m sure they’ll fit ;-) Dave, a few sentences won’t do all your contributions as PTL any justice! Nonetheless, thank you for starting the project, for diligently working towards growing it and for entrusting us, as a commun

Re: [vpp-dev] clearing sessions when application exits #vppcom #vpp-hoststack

Hi Tahir, If a vcl app/worker is not closed cleanly, it won’t detach from vpp. In vpp 20.05, the binary api infra will detect that vcl's binary api client crashed, and cleanup the stale app state, but only after about 30s. In 21.01 (master latest) after this [1] vcl can be configured to use th

Re: [vpp-dev] fifo tuning callback issue

Hi Bill, Welcome to the project and thanks for the report! Agreed, that is a bug in non-default scenarios like yours where custom fifo tuning is required. Would you be willing to push a patch or should I do it? Regards, Florin > On Nov 11, 2020, at 10:33 AM, Bill Vaughan wrote: > > Hi all,

Re: [vpp-dev] fifo tuning callback issue

ed, Nov 11, 2020 at 2:09 PM Florin Coras <mailto:fcoras.li...@gmail.com>> wrote: > Hi Bill, > > Welcome to the project and thanks for the report! Agreed, that is a bug in > non-default scenarios like yours where custom fifo tuning is required. Would > you be will

Re: [vpp-dev] #vpp #vpp-dev

Hi Nikhil, If your client and server are attached to the same vpp instance, data/segments never make it into vpp. Instead, during the connect, after vpp detects that the two endpoints are both attached locally it sets up the session buffers (fifos) in a piece of shared memory that’s mapped by

Re: [vpp-dev] #vpp #vppcom #vpp-dev #vnet

Hi Nikhil, Just use a default sshd_config file and change the 3 lines mentioned in the wiki. Note that this has been tested only with “through the stack” tcp connections, i.e., server running on top of vpp on one host and the client (running on linux, not using vpp) on a different host. You

Re: [vpp-dev] TCP debug macros

Hi Bill, Inline. > On Nov 22, 2020, at 6:17 AM, Bill Vaughan wrote: > > I should have stated that I am running 2009. > > On Sun, Nov 22, 2020 at 8:48 AM Bill Vaughan via lists.fd.io > > wrote: > Hi all, > > I am working on a project wher

Re: [vpp-dev] Stale sessions are getting cleared up and clear session from VPPCTL doesn't work

Hi Rushi, If the application crashes/restarts without detaching, it will take some time for vpp to detect that. In particular, the binary api needs to first detect that the binary api client went away (about 30s) and then everything should be cleaned up. To solve that, you could try updating

Re: [vpp-dev] TCP debug macros

onn 138 disp error state TIME_WAIT >> flags 0x12 SYN ACK > > > You mentioned that this could be a threading issue. I have been running only > with the main thread for all of this testing (no workers). I plan to add > those, but I want to test for correctness first then

Re: [vpp-dev] [VCL] Memory access error for different size of mutex with different glibc versions in VPP and VCL app

lso because libpthread is dynamically linked > into VCL and VPP. > This is really a big limitation. > > Regards, > Hanlin > > > wanghanlin > > wanghan...@corp.netease.com <mailto:wanghan...@corp.netease.com> > 签名由 网易邮箱大师 <https://mail.163.com/dashi

Re: [vpp-dev] [VCL] Memory access error for different size of mutex with different glibc versions in VPP and VCL app

Hi Hanlin, And here’s the patch [1]. Regards, Florin [1] https://gerrit.fd.io/r/c/vpp/+/30185 > On Nov 27, 2020, at 9:07 AM, Florin Coras via lists.fd.io > wrote: > > Hi Hanlin, > > Good point! I actually have that on my todo list to see how/if it affects > perfo

Re: [vpp-dev] vppcom and binary api #binapi #vapi #vppcom

Hi Ravi, By default, vcl relies on the binary api to attach to the session layer in vpp. It is possible to initialize another binary api connection from your app, apart from the one transparently created by vcl, but it requires some careful non-default initialization of a series of data struct

Re: [vpp-dev] vppcom and binary api #binapi #vapi #vppcom

Hi Ravi, VCL is not part of the vpp app/process, it’s a library that applications can link against to be able to interact with the session layer in a more posix-like manner. So if your app needs a binary api connection to vpp, it needs to set it up independent of vcl. With regard to the “vpe

Re: [vpp-dev] vppcom and binary api #binapi #vapi #vppcom

madhav Josyula wrote: > > Hi Florin, > > This is nothing to do with the subject being discussed. > > Can you please direct me to the following > i) example(s) of VCL ? > ii) Any tutorial explaining the working of VCL ? > > Thanks & Regards, > Venu > &

Re: [vpp-dev] [VCL] Memory access error for different size of mutex with different glibc versions in VPP and VCL app

b7caff1cb54.png> > 签名由 网易邮箱大师 <https://mail.163.com/dashi/dlpro.html?from=mail81> 定制 > On 12/1/2020 00:10，Florin Coras > <mailto:fcoras.li...@gmail.com> wrote： > Hi Hanlin, > > And here’s the patch [1]. > > Regards, > Florin > > [1] https://gerrit

Re: [vpp-dev] vppcom and binary api #binapi #vapi #vppcom

/src/plugins/hs_apps/vcl/sock_test_server.c > These files use normal socket(s) right now VCL > > Thanks, > Regards, > Venu > > On Tue, 8 Dec 2020 at 01:07, Florin Coras <mailto:fcoras.li...@gmail.com>> wrote: > Hi Venu, > > You’ll find documentation regarding the h

Re: [vpp-dev] vppcom and binary api #binapi #vapi #vppcom

een ldp and direct vcl ? I mean in terms of use cases > > Thanks and Regards > Venu > > On Tue, 8 Dec, 2020, 8:47 pm Florin Coras, <mailto:fcoras.li...@gmail.com>> wrote: > Hi Venu, > > Those are used for ldp tests. That is, they use normal sockets which go > t

Re: [vpp-dev] New perfmon plugin

+1 Really cool! Thanks! Cheers, Florin > On Dec 11, 2020, at 9:33 AM, Dave Barach wrote: > > Looks really cool! Feel free to move the old one to deprecated... 😉... Dave > > -Original Message- > From: vpp-dev@lists.fd.io > On Be

Re: [vpp-dev] move to clang-format

+1 Florin > On Dec 16, 2020, at 6:12 AM, Damjan Marion via lists.fd.io > wrote: > > > Any feedback? > > Any good reason not to do the switch now when we have stable/2101 created? > > Thanks, > > Damjan > > >> On 14.12.2020., at 09:32, Benoit Ganne (bganne) wrote: >> >> Sounds good to m

Re: [vpp-dev] Message posting policy change

+1 Florin > On Dec 16, 2020, at 4:47 AM, Dave Barach wrote: > > Folks, > > At the moment, we hold non-list-member posts in the vpp-dev moderation queue. > Unfortunately, spam comprises almost all of the messages in the queue. I’ve > been deleting 10-20 junk emails per day – “want to buy a lis

Re: [vpp-dev] support of raw socket like support in vpp

Hi Venu, We don’t support raw sockets in the host stack and implicitly in vcl. You might be able to achieve what you want with memif interfaces. Regards, Florin > On Dec 22, 2020, at 12:43 AM, Venumadhav Josyula wrote: > > Hi All. > > There is any raw socket like support in vpp. Our stack

Re: [vpp-dev] support of raw socket like support in vpp

nks, > Regards > Venu > > On Tue, 22 Dec, 2020, 8:52 pm Florin Coras, <mailto:fcoras.li...@gmail.com>> wrote: > Hi Venu, > > We don’t support raw sockets in the host stack and implicitly in vcl. You > might be able to achieve what you want with memif interfaces. >

Re: [vpp-dev] Problem running vcl-test-server inside a docker

Hi Venu, That last error "vcl_bapi_init:604: vcl<682:0>: ERROR couldn't connect to VPP!” pretty much explains everything. VCL is trying to connect over the binary api to vpp and it can’t find the socket. Make sure that vpp exposes the binary api’s socket where you want (/var/run as far as I ca

Re: [vpp-dev] Problem running vcl-test-server inside a docker

ps://wiki.fd.io/images/9/9c/Vpp-hoststack-kc-eu19.pdf <https://wiki.fd.io/images/9/9c/Vpp-hoststack-kc-eu19.pdf> > > Thanks, > Regards, > Venu > > > On Thu, 24 Dec 2020 at 23:29, Florin Coras <mailto:fcoras.li...@gmail.com>> wrote: > Hi Venu, > >

Re: [vpp-dev] Problem running vcl-test-server inside a docker

cut-through, my understanding is that l2 | l3 | l4 stack comes into play > w/o bypassing them ? > My understanding is that cut-through session(s) do not go through vpp stack > it is more session layer to session layer communication. > Thanks, > Regards, > Venu > > &

Re: [vpp-dev] support of raw socket like support in vpp

> Thanks, > Regards, > Venu > > > On Tue, 22 Dec 2020 at 20:52, Florin Coras <mailto:fcoras.li...@gmail.com>> wrote: > Hi Venu, > > We don’t support raw sockets in the host stack and implicitly in vcl. You > might be able to achieve what you want w

[vpp-dev] session layer/vcl no longer rely on fixed va shm segment mappings

Hi folks, Happy New Year! This is an FYI for those of you who might’ve hit issues with the session layer shared memory infra's reliance on shm segments mapped at fixed virtual addresses. Over the past weeks, I’ve refactored the svm infra, i.e., fifo segments, fifos and the message queues, a

Re: [vpp-dev] VCL API failure

Hi Vijay, VCL is part of our CI infra (see src/vcl/test/test_vcl.py), i.e., it’s constantly tested, so the issue is most probably configuration related. It looks like your vcl app (the server) cannot attach to vpp. Probably that's because of a mismatch of socket files. So, if you use the bina

Re: [vpp-dev] VCL API failure

ference. > > Pls let me know if we are missing something or is something wrong > > Regards, > Vijay Kumar N > > On Sat, Jan 9, 2021 at 1:06 AM Florin Coras <mailto:fcoras.li...@gmail.com>> wrote: > Hi Vijay, > > VCL is part of our CI infra (see src/vcl/t

Re: [vpp-dev] VCL API failure

nect (fd 3, '/tmp/vpp-api.sock'): Connection refused > vppcom_connect_to_vpp:468: vcl<1212:0>: ERROR app (ldp-1212-app) socket > connect failed! > vppcom_app_create:1200: vcl<1212:0>: ERROR couldn't connect to VPP! > > LDP<1212>: ERROR: ldp_constructor: failed! > &qu

Re: [vpp-dev] #vpp-hoststack BBR support in host-stack

Hi PK, We don’t currently support BBR but we should have most of the components needed for it, e.g., rate estimator, in place. Let’s sync up off list if you decide to pick this up. Regards, Florin > On Jan 11, 2021, at 6:35 PM, PRANAB DAS wrote: > > > > Hi > > I am wondering about BBR

Re: [vpp-dev] Performance of VPP TCP Stack #hoststack #vnet #vpp-dev #vpp_1804 #vppcapi #vppcom #vpp-hoststack #vppfortelecomm

Hi Anurag, Unfortunately, we don’t currently track tcp performance in csit but you should be able to do some throughput measurements yourself using something like this [1]. For reference, on Intel Xeon Gold 6146 and XL710 nics, one connection should be able to saturate the nic (~37Gbps). If yo

Re: [vpp-dev] Assertion failure on half-open connection pending cleanup

Hi Ivan, You’re right that the assert wrongly assumes the connection must’ve been established. I’m guessing you’re only sporadically hitting it because typically the peer should not retransmit the syn after our reset (but probably the packet is lost). Want to push a patch to remove it or shou

Re: [vpp-dev] Assertion failure on half-open connection pending cleanup

Thanks! Cheers, Florin > On Jan 15, 2021, at 4:49 PM, Ivan Shvedunov wrote: > > https://gerrit.fd.io/r/c/vpp/+/30791 <https://gerrit.fd.io/r/c/vpp/+/30791> > > On Sat, Jan 16, 2021 at 3:22 AM Florin Coras <mailto:fcoras.li...@gmail.com>> wrote: > Hi Ivan,

Re: [vpp-dev] Assertion failure on half-open connection pending cleanup

ilures in unrelated test cases ... > https://jenkins.fd.io/job/vpp-debug-verify-master-ubuntu2004-x86_64/696/console > > <https://jenkins.fd.io/job/vpp-debug-verify-master-ubuntu2004-x86_64/696/console> > > On Sat, Jan 16, 2021 at 3:58 AM Florin Coras <mailto:fcoras.

Re: [vpp-dev] git push to gerrit fails?

https://wiki.fd.io/view/DEV/Setting_up_Gerrit Regards, Florin > On Jan 19, 2021, at 10:02 AM, hemant via lists.fd.io > wrote: > > I cloned the gerrit VPP repo, created a new classify-val branch using git, > made code changes, and used commit.

Re: [vpp-dev] VPP Host Stack #hoststack #vpp-hoststack #vppfortelecomm #vpp-dev #vppcom #vppwithoutdpdk

Hi Anurag, Inline. > On Jan 19, 2021, at 11:54 PM, anurag.bhumc...@hotmail.com wrote: > > Hi All, > > I have few queries related to VPP Host TCP stack: > 1. Can we use only VPP TCP stack(L4 layer) only and disable L2/L3. If not > what is its advantage of TLDK or similar stacks. Not sure I fo

Re: [vpp-dev] #vppcom #vpp-hoststack

Hi Srini, We do not support tcp keep-alives. If you own the two ends of a communication, it should be relatively easy to implement application level keep-alives. Regards, Florin > On Jan 27, 2021, at 2:14 AM, srinimurth...@gmail.com wrote: > > Hi All, >I'm looking to use TCP keep-a

Re: [vpp-dev] [csit-dev] VPP release 21.01 release is available on packagecloud.io/fdio/release !

Awesome!! Congrats! Cheers, Florin > On Jan 27, 2021, at 1:07 PM, Andrew Yourtchenko wrote: > > Hi all, > > VPP release 21.01 is complete and is available from the usual > packagecloud.io/fdio/release location! > > I have verified using the scripts [0] that the new release installs > and runs

Re: [vpp-dev] Jenkins/Gerrit Failure Cause Notifications

Cool! Florin > On Feb 2, 2021, at 6:03 PM, Dave Barach wrote: > > Thanks very much! > > Dave > > -Original Message- > From: vpp-dev@lists.fd.io > On Behalf Of Dave Wallace > Sent: Tuesday, February 2, 2021 8:28 PM > To: vpp-dev

Re: [vpp-dev] vnet_connect() from workers

Hi Bill, First of all, thanks for sharing the details of the project. The scale you're looking, at least in terms of total number of active sessions should be relatively easily achievable with the right hardware. In case you haven’t done it yet, I’ll only recommend that you preallocate the numb

Re: [vpp-dev] vnet_connect() from workers

or now. > > I'll take a look at session_mq_connect_handler(). > > Much appreciated, > -Bill > > > > > > > > > > > On Fri, Feb 5, 2021 at 11:23 AM Florin Coras <mailto:fcoras.li...@gmail.com>> wrote: > Hi Bill, > >

Re: [vpp-dev] vppcom_app_create() is failed.

Hi Nikhil, That’s a vcl api, it can only fail if it’s called twice or if the attachment to vpp fails. Enable vcl debug by exporting VCL_DEBUG=2 and check what failed. Regards, Florin > On Feb 10, 2021, at 6:40 AM, nikhil subhedar wrote: > > Hi All, > > my vppcom_app_create() is failing .

Re: [vpp-dev] RFC: Interface Mirroring for Linux Network Stackintegration

Hi Neale, You should be able to fake a binary api connection from within vpp with vl_api_memclnt_create_internal (there be dragons). The mutex is used only to synchronize producers and consumer (typically should be only one) so it shouldn’t block unless you run out of space in the queue. Simil

Re: [vpp-dev] Facing Failed: vppctl show cpu: clib_socket_init: connect (fd 3, '/run/vpp/cli.sock'): Connection refused.

Hi Nikhil, There are a couple of things conflated here. First, the cli is vpp’s cli which can be connected to through a socket. That’s the “/run/vpp/cli.sock” and it’s created by vpp at startup so remove the one you’ve created. The binary api also exposes a socket and you’ve configured it to u

Re: [vpp-dev] A newbie question: "VPP worker threads" vs "vcl worker threads" #vppcom #hoststack

Hi Hyong, VCL runs as part of the app process so yes, your intuition was right, vcl and vpp workers are different. Note however that the session layer also accepts attachments from vpp “builtin” applications. Those end up consuming vpp worker cpu cycles as all the signaling will happen on the

Re: [vpp-dev] Newbie question: best practice for getting data from plugins? #vpp #vppcapi #vapi

Hi Hyong, It ultimately depends on what “function” does and how you plan to solve worker contention for whatever resources need to be shared. Assuming that 1) “function” must be executed asynchronously and 2) the node forwards the packets after it copies the interesting data out of them, some

Re: [vpp-dev] fd problems when use cut-through connections

Hi Yacan, Unfortunately what you found is a known scale (and somewhat performance) limitation for cut-through sessions. There is a plan to alleviate that by allocating per app pair segments (as opposed to per connection) but no ETA for it. If anybody else is interested in tackling this, let me

Re: [vpp-dev] fd problems when use cut-through connections

are there any obstacles to allocate > multiple fifo pairs in one segment like non-cutthrough connection ? > > Best Regard, > Yacan > > On 2/27/2021 03:48，Florin Coras > <mailto:fcoras.li...@gmail.com> wrote： > Hi Yacan, > > Unfortunately what you found i

Re: [vpp-dev] TCP connection using VPP

Hi Chetan, That is a pretty old version of vpp, so I would first recommend updating to something newer. Regarding the specific issue you are hitting, it’s hard to tell what’s causing it from your description. Maybe also try iperf3 like here [1] and see if that works. If that’s still not workin

Re: [vpp-dev] VPP Host Stack - TCP Proxy without FIFO #vpp-hoststack

Hi, Inline. > On Mar 2, 2021, at 8:19 AM, theoliviadun...@gmail.com wrote: > > Dear All, > > The example application hs_apps/proxy.c > seems > to achieve the functionality of TCP proxy using VPP host stack. FC: Yes, but k

Re: [vpp-dev] TCP connection using VPP

> > On Tuesday, March 2, 2021, 09:38:34 PM GMT+5:30, Florin Coras > wrote: > > > Hi Chetan, > > That is a pretty old version of vpp, so I would first recommend updating to > something newer. Regarding the specific issue you are hitting, it’s hard to > tell wh

Re: [vpp-dev] VPP Host Stack - TCP Proxy without FIFO #vpp-hoststack

Hi Olivia, > On Mar 2, 2021, at 8:14 PM, Olivia Dunham wrote: > > Thanks Florin! That really helped. Glad it did! > Yes, but keep in mind that’s an explicit proxy (client connects to the > proxy’s ip). > Yes, you're right. How about a transparent one by listening on 0/0 all ports > and by

Re: [vpp-dev] vcl epoll timeout fix not working on vpp 21.01

Hi Ashish, Hard to say without additional details. What is the exact behavior you’re seeing and are you using by chance eventds for notifications? In that case you’ll probably need at least [1]. Regards, Florin [1] https://gerrit.fd.io/r/c/vpp/+/30959 > On Mar 4, 2021, at 6:37 AM, ashish.sa

Re: [vpp-dev] Performance problem in multithreaded vcl application #vpp

Hi, Is it safe to assume this is vpp master? More inline. > On Mar 9, 2021, at 5:16 AM, sokolov.3v...@gmail.com wrote: > > Hello! > > I am learning VPP and I need to transfer more than 10G using UDP protocol in > both directions (tx and rx). > I tried to write an UDP server. There are three

Re: [vpp-dev] RFC: vlib_global_main_t

+1 Florin > On Mar 11, 2021, at 7:18 AM, Dave Barach wrote: > > Looks like a win to me... > > -Original Message- > From: vpp-dev@lists.fd.io > On Behalf Of Damjan Marion via lists.fd.io > > Sent: Thursday,

Re: [vpp-dev] UDP data corruption and MTU issues

Hi Ivan, Inline. > On Mar 15, 2021, at 9:08 AM, Ivan Shvedunov wrote: > > Hello, > > I've noticed a problem in the VPP session layer and posted a patch to fix > it [1]. > Namely, UDP datagrams with length > 1908 that require buffer chaining get > corrupted, because there's a bug in sess

Re: [vpp-dev] UDP data corruption and MTU issues

Hi Ivan, > On Mar 16, 2021, at 6:40 AM, Ivan Shvedunov wrote: > > Hi Florin, > > (inline) > > On Mon, Mar 15, 2021 at 10:27 PM Florin Coras <mailto:fcoras.li...@gmail.com>> wrote: >> On Mar 15, 2021, at 9:08 AM, Ivan Shvedunov > <mailto

Re: [vpp-dev] UDP data corruption and MTU issues

-vpp > <https://github.com/travelping/upg-vpp> > > > On Tue, Mar 16, 2021 at 6:19 PM Florin Coras <mailto:fcoras.li...@gmail.com>> wrote: > Hi Ivan, > > >> On Mar 16, 2021, at 6:40 AM, Ivan Shvedunov > <mailto:ivan...@gmail.com>> wrote: >

Re: [vpp-dev] UDP data corruption and MTU issues

pass there. So indeed must be something related to > the test environment > > On Tue, Mar 16, 2021 at 7:02 PM Florin Coras <mailto:fcoras.li...@gmail.com>> wrote: > Hi Ivan, > > >> On Mar 16, 2021, at 8:51 AM, Ivan Shvedunov > <mailto:ivan...@g

Re: [vpp-dev] Issue in vcl socket in VPP release 21.01

Hi Ashish, Interesting. What does your vcl app do after vpp restarts? Does it call vppcom_app_destroy() and then reattaches or are you trying to use the existing attachment context? If the app doesn’t detect vpp’s restart, it’s the latter because vcl currently doesn’t detect vpp restarts. Re

Re: [vpp-dev] VPP / Quic perfomances

Awesome!! Cheers, Florin > On Nov 30, 2020, at 2:57 AM, Mathias Raoul -X (mraoul - LIANEO at Cisco) via > lists.fd.io wrote: > > Hello, > > Following the work already done this year, we are currently updating our quic > plugin in vpp to benefits from last features and optimizations introduc

Re: [vpp-dev] vppcom epoll half-closed tcp session

Hi Vadim, When peer initiates the closing of the connection (sends fin), session layer notifies vcl about the event and consequently vcl generates an EPOLLHUP | EPOLLRDHUP event towards the app, assuming app uses epoll. Outstanding data can still be read and new data can be written. Are you lo

Re: [vpp-dev] TCP delay ACK

Hi Yacan, We don’t support delayed acks on purpose, the reasoning being that by construction we only send an ack per burst of packets received. Moreover, delayed acks tend to interfere with congestion control algorithms and rtt estimators. Out of curiosity, why are you looking into that? Re

Re: [vpp-dev] keeping tests outside of src/

Not a strong opinion but, at least for vnet tests, I agree with Damjan. Florin > On Mar 26, 2021, at 9:37 AM, Damjan Marion via lists.fd.io > wrote: > > > >> On 25.03.2021., at 21:14, Dave Wallace > > wrote: >> >> Hi Damjan, >> >> This initiative originated wi

Re: [vpp-dev] vppcom epoll half-closed tcp session

n peer FIN vcl should generate EPOLLRDHUP event towards the App > On full session close vcl should generate EPOLLRDHUP | EPOLLHUP event towards > the App > > Regards, > Vadim > > > From: Florin Coras > Sent: Tuesday, March 23, 2021 5:59:49 PM > To: Vadim Platonov >

Re: [vpp-dev] VPP host stack thread ownership #hoststack

Hi Olivia, Inline. > On Mar 29, 2021, at 8:19 AM, Olivia Dunham wrote: > > Dear All, > > I have been reading VPP host stack and saw the example TCP proxy in > (src/plugins/hs_apps/proxy.c). > I understand that the main thread handles the CLI command and the callbacks > are registered for TC

Re: [vpp-dev] vppcom epoll half-closed tcp session

we want to support > half-closed tcp sessions scenario. But as I can suggest by vpp source code > there is no support for half-closed tcp sessions. Is there any plans for > supporting half-closed tcp sessions? > > Regards, > Vadim > From: Florin Coras > Sent: Monday, Marc

Re: [vpp-dev] New Committer Proposal

+1 Florin > On Mar 31, 2021, at 10:58 AM, Damjan Marion via lists.fd.io > wrote: > > > Dear VPP Committers, > > I would like to propose Roy Fan Zhang from Intel as a new VPP committer. > Fan made significant contributions to the VPP including the async crypto > infrastructure and crypto sche

Re: [vpp-dev] Is the tcp connection timestamp updated too early

Hi Yacan, We only update tc->tsval_recent if the segment is fine and tc->tsval_recent is less than rcv_opts.tsval, i.e., the new tsval moves forward. Even if the ack is dropped, we only use tsval_recent in paws check. On the other hand rcv_opts.tsecr is not used unless the ack is valid. Did

Re: [vpp-dev] Is the tcp connection timestamp updated too early

ted. > > BR, > yacan > On 4/1/2021 11:29，Florin Coras > <mailto:fcoras.li...@gmail.com> wrote： > Hi Yacan, > > We only update tc->tsval_recent if the segment is fine and tc->tsval_recent > is less than rcv_opts.tsval, i.e., the new tsval moves forward

Re: [vpp-dev] Is the tcp connection timestamp updated too early

.. > Em.It seems that even if the ACK bit is off, TS.Recent will still be > updated. > Well, I will stop here. > > BR， > Yacan > On 4/1/2021 12:38，Florin Coras > <mailto:fcoras.li...@gmail.com> wrote： > Hi Yacan, > > Ah, I’m assuming subsequent

Re: [vpp-dev] Can tcp4 packet bypass ip4-lookup node ?

Hi Yacan, At this point, only ipv6 link local packets end up in ip-rewrite. Because of the v4/v6 symmetry we have arcs for both address families. With respect to the two questions. First, tcp doesn’t cache the dpo for the remote ip because that would require tracking the dpo (in case it chang

Re: [vpp-dev] gdb vpp with make test?

Hi Hemant, Probably you’re looking for: make test TEST=test_foo DEBUG=gdb Regards, Florin > On Apr 19, 2021, at 4:00 PM, hemant via lists.fd.io > wrote: > > Hi folks, > > I know how to use gdb with vpp when a vpp shell is logged into using “sudo > vppctl -s /run/vpp/cli.sock”. > > Howeve

Re: [vpp-dev] VPP TCP hoststack performance data #vpp-hoststack #vnet

Hi Natraj, Host stack tests are not currently part of trending but we do have a per release report [1] which we are currently working on expanding/improving. Keep in mind that tcp perf depends a lot on cpu frequency, so if you want to know precisely how fast tcp is in your environment, I’d re

Re: [vpp-dev] vcl_test_client is failing with Unsupported application config (-108) #vppcom

Hi, Is there anything configured on vpp side for session layer? Is this vpp 21.06rc0 or something older? The error number seems to suggest an older release. One option would be to just comment out use-mq-eventfd and see if that fixes the issue. Message queue eventfds should work with the bin

Re: [vpp-dev] vcl_test_client is failing with Unsupported application config (-108) #vppcom

i.e application_verify_cfg (seg_type) > seg_type = SSVM_SEGMENT_MEMFD; > > gdb: > > (gdb) p seg_type > $22 = SSVM_SEGMENT_MEMFD > (gdb) n > 492 options = a->options; > (gdb) p *a > $23 = {api_client_index = 0, name = 0x7fffce651780 "vcl_test_c

Re: [vpp-dev] vcl_test_client is failing with Unsupported application config (-108) #vppcom

Hi Sastry, Can’t do much without a backtrace. Also, not entirely sure how vcl got a reply if vpp crashed. Session layer is the main component of the host stack that keeps track of external apps and facilitates the exchange of data and notifications between apps (like vcl) and transports. So

Re: [vpp-dev] vcl_test_client is failing with Unsupported application config (-108) #vppcom

Try running vpp from gdb or, as previously recommended, try upgrading to latest version. Regards, Florin > On May 5, 2021, at 8:41 AM, Sastry Sista wrote: > > Hi Florin, > I am not seeing even backtrace. bt is just empty now. > Also I have just added session { } with em

Re: [vpp-dev] vcl_test_client is failing with Unsupported application config (-108) #vppcom

[vpp-dev moved to bcc because this is becoming quite noisy for the list.] session_main_init and session_manager_main_enable. Regards, Florin > On May 5, 2021, at 8:49 AM, Sastry Sista wrote: > > Hi Florin, > I am running in gdb only. > > last few lines in gdb: > > fdio

Re: [vpp-dev] vcl_test_client is failing with Unsupported application config (-108) #vppcom

Hi Sastry, That is pretty weird. Re-check your extended config to make sure it doesn’t have any errors in it. VCL is part of our CI so most probably the issue is some misconfiguration. For older releases you do need to explicitly ask that the event queues be allocated in a memfd segment, lik

Re: [vpp-dev] VCL app with multiple TCP sessions handling in single worker

Hi Sastry, Inline. > On May 21, 2021, at 11:22 AM, Sastry Sista wrote: > > Hi Florin, > Gone through your old mails on eventfd of mq and given me decent > inputs. > vppcom_mq_epoll_fd() is returning mq_epfd which I can use for my app dispatch > linux epoll. But, My concern is, Can I

Re: [vpp-dev] VCL app requires source address/interface control due to loopback

Hi Sastry, As you can see from the code lower, vcl in 20.05 doesn’t. It is supported however since 21.01 [1]. Regards, Florin [1] https://git.fd.io/vpp/tree/src/vcl/vppcom.c?h=v21.01#n206 > On May 23, 2021, at 11:33 PM, Sastry Sista wrote: > > Hi, >lcl_ip and lcl_port looks sou

Re: [vpp-dev] VCL app with multiple TCP sessions handling in single worker

Hi Sastry, Inline. > On May 25, 2021, at 7:44 AM, Sastry Sista wrote: > > Hi Florin, > Thank you. > > Let me be very specific and focused topic: > > I have just single worker but having multiple TCP sessions. So, I will poll > on vppcom_mq_epoll_fd() for all notifications. S

Re: [vpp-dev] VCL app requires source address/interface control due to loopback

Hi Sastry, So the source address should be configurable, but for if the app requires a certain vrf, you should consider the option of using session layer namespaces. See for instance test/test_vcl.py and how we configure those. Note however that will restrict the app to use only one vrf. Reg

<    1   2   3   4   5   6   7   8   >