Re: [vpp-dev] some questions about LACP(link bonding mode 4)

[Zhiyong Yang]

I mean,  Is there no limit on the number of active-slaves as well?

From: vpp-dev@lists.fd.io [mailto:vpp-dev@lists.fd.io] On Behalf Of Zhiyong Yang
Sent: Thursday, June 13, 2019 1:30 PM
To: Steven Luong (sluong) ; vpp-dev@lists.fd.io; Carter, 
Thomas N 
Cc: Kinsella, Ray 
Subject: Re: [vpp-dev] some questions about LACP(link bonding mode 4)

Thanks a lot, Steven.

Does it mean that all linkings(active-slaves in the same bonding group) of 
negotiating successfully can join loading balance for TX in VPP , right?

Thanks
Zhiyong
From: Steven Luong (sluong) [mailto:slu...@cisco.com]
Sent: Thursday, June 13, 2019 12:10 PM
To: Yang, Zhiyong mailto:zhiyong.y...@intel.com>>; 
vpp-dev@lists.fd.io; Carter, Thomas N 
mailto:thomas.car...@charter.com>>
Cc: Kinsella, Ray mailto:ray.kinse...@intel.com>>
Subject: Re: some questions about LACP(link bonding mode 4)

There is no limit on the number of slaves in a bonding group in VPP’s 
implementation. I don’t know/remember how to select one port over another from 
the spec without reading it carefully again.

Steven

From: "Yang, Zhiyong" mailto:zhiyong.y...@intel.com>>
Date: Tuesday, June 11, 2019 at 11:09 PM
To: "vpp-dev@lists.fd.io" 
mailto:vpp-dev@lists.fd.io>>, "Steven Luong (sluong)" 
mailto:slu...@cisco.com>>, "Carter, Thomas N" 
mailto:thomas.car...@charter.com>>
Cc: "Kinsella, Ray" mailto:ray.kinse...@intel.com>>
Subject: some questions about LACP(link bonding mode 4)

Hi Steven and VPP guys,

I’m studying the lacp implementation. and want to know if it is 
possible that Numa is considered in LACP active port selection. As we all know, 
if  slave with local numa can be preferred to help improve throughput.
One question is that current code seems no linking number limit for 
active-slave, right ?
Does it mean if we can add any number of linkings to link aggregation group? If 
two sides (actor and partner) are negotiated  well for linking?  I also don’t 
see that how to selection policy in group. What do I miss?
Port_priority is set to 0xff by default and don’t change in any case.
If numa is considered, do we use it when negotiation happens, and make slave 
with local numa selected in priority?

Thanks
Zhiyong



-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13274): https://lists.fd.io/g/vpp-dev/message/13274
Mute This Topic: https://lists.fd.io/mt/32038398/21656
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-

Re: [vpp-dev] some questions about LACP(link bonding mode 4)

[Zhiyong Yang]

Thanks a lot, Steven.

Does it mean that all linkings(active-slaves in the same bonding group) of 
negotiating successfully can join loading balance for TX in VPP , right?

Thanks
Zhiyong
From: Steven Luong (sluong) [mailto:slu...@cisco.com]
Sent: Thursday, June 13, 2019 12:10 PM
To: Yang, Zhiyong ; vpp-dev@lists.fd.io; Carter, Thomas 
N 
Cc: Kinsella, Ray 
Subject: Re: some questions about LACP(link bonding mode 4)

There is no limit on the number of slaves in a bonding group in VPP’s 
implementation. I don’t know/remember how to select one port over another from 
the spec without reading it carefully again.

Steven

From: "Yang, Zhiyong" mailto:zhiyong.y...@intel.com>>
Date: Tuesday, June 11, 2019 at 11:09 PM
To: "vpp-dev@lists.fd.io" 
mailto:vpp-dev@lists.fd.io>>, "Steven Luong (sluong)" 
mailto:slu...@cisco.com>>, "Carter, Thomas N" 
mailto:thomas.car...@charter.com>>
Cc: "Kinsella, Ray" mailto:ray.kinse...@intel.com>>
Subject: some questions about LACP(link bonding mode 4)

Hi Steven and VPP guys,

I’m studying the lacp implementation. and want to know if it is 
possible that Numa is considered in LACP active port selection. As we all know, 
if  slave with local numa can be preferred to help improve throughput.
One question is that current code seems no linking number limit for 
active-slave, right ?
Does it mean if we can add any number of linkings to link aggregation group? If 
two sides (actor and partner) are negotiated  well for linking?  I also don’t 
see that how to selection policy in group. What do I miss?
Port_priority is set to 0xff by default and don’t change in any case.
If numa is considered, do we use it when negotiation happens, and make slave 
with local numa selected in priority?

Thanks
Zhiyong



-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13273): https://lists.fd.io/g/vpp-dev/message/13273
Mute This Topic: https://lists.fd.io/mt/32038398/21656
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-

Re: [vpp-dev] some questions about LACP(link bonding mode 4)

[steven luong via Lists.Fd.Io]

There is no limit on the number of slaves in a bonding group in VPP’s 
implementation. I don’t know/remember how to select one port over another from 
the spec without reading it carefully again.

Steven

From: "Yang, Zhiyong" 
Date: Tuesday, June 11, 2019 at 11:09 PM
To: "vpp-dev@lists.fd.io" , "Steven Luong (sluong)" 
, "Carter, Thomas N" 
Cc: "Kinsella, Ray" 
Subject: some questions about LACP(link bonding mode 4)

Hi Steven and VPP guys,

I’m studying the lacp implementation. and want to know if it is 
possible that Numa is considered in LACP active port selection. As we all know, 
if  slave with local numa can be preferred to help improve throughput.
One question is that current code seems no linking number limit for 
active-slave, right ?
Does it mean if we can add any number of linkings to link aggregation group? If 
two sides (actor and partner) are negotiated  well for linking?  I also don’t 
see that how to selection policy in group. What do I miss?
Port_priority is set to 0xff by default and don’t change in any case.
If numa is considered, do we use it when negotiation happens, and make slave 
with local numa selected in priority?

Thanks
Zhiyong



-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13272): https://lists.fd.io/g/vpp-dev/message/13272
Mute This Topic: https://lists.fd.io/mt/32038398/21656
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-

Re: [vpp-dev] setsockopt with TCP_NODELAY

[Florin Coras]

Unfortunately, we don’t support any of those options at this time. In 
particular, the stack won’t delay segments and tcp doesn’t do keepalives. Do 
you really need those?

If you need to constrain tcp mss, you can indirectly do it by configuring the 
tcp mtu as a vpp startup parameter. You can’t do it per connection at this 
time.  

Florin

> On Jun 12, 2019, at 12:01 PM, Nataraj Batchu  wrote:
> 
> Hi Florin,
> 
> Thanks for reply. My question is setsockopt() for not yet established 
> sessions i.e before connect(). I was trying to clearing TCP_NODELAY, setting 
> TCP_MAXSEG(to much lower than interface MTU) and TCP_KEEPINTVL. These options 
> need to be communicating to the vnet right?  I was looking for that.
> 
> Thanks,
> -Nataraj
> 
> On Wed, Jun 12, 2019 at 11:14 AM Florin Coras  > wrote:
> Hi Nataraj, 
> 
> TCP defaults to TCP_NODELAY behavior, unless you need Nagle’s algorithm on, 
> you should be fine. 
> 
> As for the socket options, no. At this time, ldp cannot change properties for 
> established tcp connections. What sort of things are you missing?
> 
> Thanks,
> Florin 
> 
> > On Jun 12, 2019, at 11:04 AM, Nataraj Batchu  > > wrote:
> > 
> > Hi,
> > 
> > I was trying to use TCP_NODELAY socket option in my application. But I see 
> > that relevant code is "VPP-TBD" in vppcom.c. I see same comment for most of 
> > socket options.  Is there any other branch where code is available? If not, 
> > are there any plans to support setsockopt that involve communicating VPP 
> > TCP core.  
> > 
> > case VPPCOM_ATTR_SET_TCP_NODELAY:
> >   if (buffer && buflen && (*buflen == sizeof (int)))
> >  {
> >/* VPP-TBD */
> >if (*(int *) buffer)
> >  VCL_SESS_ATTR_SET (session->attr, VCL_SESS_ATTR_TCP_NODELAY);
> >else
> >  VCL_SESS_ATTR_CLR (session->attr, VCL_SESS_ATTR_TCP_NODELAY);
> >  
> >VDBG (2, "VPPCOM_ATTR_SET_TCP_NODELAY: %d, buflen %d, #VPP-TBD#",
> >  VCL_SESS_ATTR_TEST (session->attr, VCL_SESS_ATTR_TCP_NODELAY),
> >  *buflen);
> >  }
> >  
> > -=-=-=-=-=-=-=-=-=-=-=-
> > Links: You receive all messages sent to this group.
> > 
> > View/Reply Online (#13268): https://lists.fd.io/g/vpp-dev/message/13268 
> > 
> > Mute This Topic: https://lists.fd.io/mt/32043473/675152 
> > 
> > Group Owner: vpp-dev+ow...@lists.fd.io 
> > Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub 
> >   [fcoras.li...@gmail.com 
> > ]
> > -=-=-=-=-=-=-=-=-=-=-=-
> 
> 
> 
> -- 
> Thanks,
> -Nataraj

-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13271): https://lists.fd.io/g/vpp-dev/message/13271
Mute This Topic: https://lists.fd.io/mt/32043473/21656
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-

Re: [vpp-dev] setsockopt with TCP_NODELAY

[Nataraj Batchu]

Hi Florin,

Thanks for reply. My question is setsockopt() for not yet established
sessions i.e before connect(). I was trying to clearing TCP_NODELAY,
setting TCP_MAXSEG(to much lower than interface MTU) and TCP_KEEPINTVL.
These options need to be communicating to the vnet right?  I was looking
for that.

Thanks,
-Nataraj

On Wed, Jun 12, 2019 at 11:14 AM Florin Coras 
wrote:

> Hi Nataraj,
>
> TCP defaults to TCP_NODELAY behavior, unless you need Nagle’s algorithm
> on, you should be fine.
>
> As for the socket options, no. At this time, ldp cannot change properties
> for established tcp connections. What sort of things are you missing?
>
> Thanks,
> Florin
>
> > On Jun 12, 2019, at 11:04 AM, Nataraj Batchu 
> wrote:
> >
> > Hi,
> >
> > I was trying to use TCP_NODELAY socket option in my application. But I
> see that relevant code is "VPP-TBD" in vppcom.c. I see same comment for
> most of socket options.  Is there any other branch where code is available?
> If not, are there any plans to support setsockopt that involve
> communicating VPP TCP core.
> >
> > case VPPCOM_ATTR_SET_TCP_NODELAY:
> >   if (buffer && buflen && (*buflen == sizeof (int)))
> >  {
> >/* VPP-TBD */
> >if (*(int *) buffer)
> >  VCL_SESS_ATTR_SET (session->attr, VCL_SESS_ATTR_TCP_NODELAY);
> >else
> >  VCL_SESS_ATTR_CLR (session->attr, VCL_SESS_ATTR_TCP_NODELAY);
> >
> >VDBG (2, "VPPCOM_ATTR_SET_TCP_NODELAY: %d, buflen %d, #VPP-TBD#",
> >  VCL_SESS_ATTR_TEST (session->attr, VCL_SESS_ATTR_TCP_NODELAY),
> >  *buflen);
> >  }
> >
> > -=-=-=-=-=-=-=-=-=-=-=-
> > Links: You receive all messages sent to this group.
> >
> > View/Reply Online (#13268): https://lists.fd.io/g/vpp-dev/message/13268
> > Mute This Topic: https://lists.fd.io/mt/32043473/675152
> > Group Owner: vpp-dev+ow...@lists.fd.io
> > Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [
> fcoras.li...@gmail.com]
> > -=-=-=-=-=-=-=-=-=-=-=-
>
>

-- 
Thanks,
-Nataraj
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13270): https://lists.fd.io/g/vpp-dev/message/13270
Mute This Topic: https://lists.fd.io/mt/32043473/21656
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-

Re: [vpp-dev] setsockopt with TCP_NODELAY

[Florin Coras]

Hi Nataraj, 

TCP defaults to TCP_NODELAY behavior, unless you need Nagle’s algorithm on, you 
should be fine. 

As for the socket options, no. At this time, ldp cannot change properties for 
established tcp connections. What sort of things are you missing?

Thanks,
Florin 

> On Jun 12, 2019, at 11:04 AM, Nataraj Batchu  wrote:
> 
> Hi,
> 
> I was trying to use TCP_NODELAY socket option in my application. But I see 
> that relevant code is "VPP-TBD" in vppcom.c. I see same comment for most of 
> socket options.  Is there any other branch where code is available? If not, 
> are there any plans to support setsockopt that involve communicating VPP TCP 
> core.  
> 
> case VPPCOM_ATTR_SET_TCP_NODELAY:
>   if (buffer && buflen && (*buflen == sizeof (int)))
>  {
>/* VPP-TBD */
>if (*(int *) buffer)
>  VCL_SESS_ATTR_SET (session->attr, VCL_SESS_ATTR_TCP_NODELAY);
>else
>  VCL_SESS_ATTR_CLR (session->attr, VCL_SESS_ATTR_TCP_NODELAY);
>  
>VDBG (2, "VPPCOM_ATTR_SET_TCP_NODELAY: %d, buflen %d, #VPP-TBD#",
>  VCL_SESS_ATTR_TEST (session->attr, VCL_SESS_ATTR_TCP_NODELAY),
>  *buflen);
>  }
>  
> -=-=-=-=-=-=-=-=-=-=-=-
> Links: You receive all messages sent to this group.
> 
> View/Reply Online (#13268): https://lists.fd.io/g/vpp-dev/message/13268
> Mute This Topic: https://lists.fd.io/mt/32043473/675152
> Group Owner: vpp-dev+ow...@lists.fd.io
> Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [fcoras.li...@gmail.com]
> -=-=-=-=-=-=-=-=-=-=-=-

-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13269): https://lists.fd.io/g/vpp-dev/message/13269
Mute This Topic: https://lists.fd.io/mt/32043473/21656
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-

[vpp-dev] setsockopt with TCP_NODELAY

[Nataraj Batchu]

Hi,

I was trying to use TCP_NODELAY socket option in my application. But I see that 
relevant code is "VPP-TBD" in vppcom.c. I see same comment for most of socket 
options.  Is there any other branch where code is available? If not, are there 
any plans to support setsockopt that involve communicating VPP TCP core.

case VPPCOM_ATTR_SET_TCP_NODELAY:
if (buffer && buflen && (*buflen == sizeof (int)))
{
/* VPP-TBD */
if (*(int *) buffer)
VCL_SESS_ATTR_SET (session->attr, VCL_SESS_ATTR_TCP_NODELAY);
else
VCL_SESS_ATTR_CLR (session->attr, VCL_SESS_ATTR_TCP_NODELAY);

VDBG (2, "VPPCOM_ATTR_SET_TCP_NODELAY: %d, buflen %d, #VPP-TBD#",
VCL_SESS_ATTR_TEST (session->attr, VCL_SESS_ATTR_TCP_NODELAY),
*buflen);
}
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13268): https://lists.fd.io/g/vpp-dev/message/13268
Mute This Topic: https://lists.fd.io/mt/32043473/21656
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-

[vpp-dev] Active close case with outstanding data case issue of version 19.04 #vnet

[guangwei]

>From the code, when active close in tcp_connection_close, the TCP state will 
>change to FIN_WAIT_1 and if there is TX data meanwhile, it will set TCP flag 
>TCP_CONN_FINPNDG also,
but there is no method in the stack for this TCP to step into the FIN_WAIT_2 
state, the only thing this TCP can happen only wait waitclose timer expire and 
break up the connection.

This issue is by design ? or something I missed ?
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13267): https://lists.fd.io/g/vpp-dev/message/13267
Mute This Topic: https://lists.fd.io/mt/32042003/21656
Mute #vnet: https://lists.fd.io/mk?hashtag=vnet&subid=1480452
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-

Re: [vpp-dev] timer wheel usage queries

[Andreas Schultz]

Additionally to Dave's answer, you also have to be very careful with
the expected precision of the timers.

When you're inserting a timer with a interval on N into a timer with a
period of P, you timeout is not guaranteed to be N * P, it is more
something like (N-1)*P < timeout < N*P.
The reason for that is that the wheel works on a an internal tick and
that tick does no necessarily equal NOW. You might be near the end of
current tick when inserting and that will lead to a shorter that
expected timeout.

For timers where the timeout is magnitudes larger then the interval,
that does not matter. But when you have to compare the expected time
of the timeout to NOW you might get into trouble.

Especially since NOW is a float, calculating NOW+(N*period) will not
result in the same value as adding period N times to NOW
(NOW+period+period+period+period) in a loop.
Rounding differences lead to a shift that can break a simple
earlier/later comparison on those results.

Regards
Andreas

Am Mi., 12. Juni 2019 um 13:22 Uhr schrieb Dave Barach via Lists.Fd.Io
:
>
> You didn't mention calling tw_timer_expire_timers_xxx(tw, now), which is the 
> only force of physics which will actually cause a timer to expire.
>
> See also .../src/vppinfra/test_tw_timer.c, src/vlib/main.c or 
> src/vnet/tcp/tcp.c for usage examples. The timer wheel code has been tested 
> to N=60 million concurrent timers. It's not trivially broken.
>
> D
>
> 
> From: vpp-dev@lists.fd.io  on behalf of hari_akkin via 
> Lists.Fd.Io 
> Sent: Wednesday, June 12, 2019 7:00 AM
> To: vpp-dev@lists.fd.io
> Cc: vpp-dev@lists.fd.io
> Subject: [vpp-dev] timer wheel usage queries
>
> Hello
> I am trying to use timer_wheel framework to start a timer. I followed the 
> examples given in fdio, but the expiry_callback function never being called 
> back.
> I am not sure if timer failed to expire or my usage has problems. I used the 
> below functions:
>
> Timer wheel Init:
> {
> tw_timer_wheel_1t_1w_2048sl_t* tw = &example_main->timer_wheels;
> tw_timer_wheel_init_1t_1w_2048sl (tw, example_timer_expired_callback, 
> 1.0 /* timer period 1s */ , 1024);
> tw->last_run_time = vlib_time_now (vm);
> }
>
> Timer Start:
> handle = tw_timer_start_1t_1w_2048sl(&example->timer_wheels,1,0,5);
>
> The expiry functiion example_timer_expired_callback never been called.
>
> Is there any issue in the way I am using or is there any knwon issue in TW 
> framework?
>
> -=-=-=-=-=-=-=-=-=-=-=-
> Links: You receive all messages sent to this group.
>
> View/Reply Online (#13262): https://lists.fd.io/g/vpp-dev/message/13262
> Mute This Topic: https://lists.fd.io/mt/32039603/675601
> Group Owner: vpp-dev+ow...@lists.fd.io
> Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  
> [andreas.schu...@travelping.com]
> -=-=-=-=-=-=-=-=-=-=-=-



-- 
-- 
Dipl.-Inform. Andreas Schultz

--- enabling your networks --
Travelping GmbH Phone:  +49-391-81 90 99 0
Roentgenstr. 13 Fax:+49-391-81 90 99 299
39108 Magdeburg Email:  i...@travelping.com
GERMANY Web:http://www.travelping.com

Company Registration: Amtsgericht StendalReg No.:   HRB 10578
Geschaeftsfuehrer: Holger Winkelmann  VAT ID No.: DE236673780
-
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13266): https://lists.fd.io/g/vpp-dev/message/13266
Mute This Topic: https://lists.fd.io/mt/32039603/21656
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-

[vpp-dev] MAINTAINERS update

[Damjan Marion via Lists.Fd.Io]

Dear all,

After the yesterday discussion, I submitted update for MAINTAINERS file.

https://gerrit.fd.io/r/20089 


I kindly ask all folks who maintain parts of the code to take a look and add 
missing/update existing entries…

Thanks,

Damjan-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13265): https://lists.fd.io/g/vpp-dev/message/13265
Mute This Topic: https://lists.fd.io/mt/32040415/21656
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-

Re: [vpp-dev] Re-connect to vpp will make client CPU high. (I have solved it).

[Jin Zhihua]

Hi,
   Sorry, I made a mistake in last mail, if “vac_disconnect” called 
“unset_timeout” instead of “set_timeout", which would make client never ever 
connect to vpp. 
I  found the correct solution and tested, in “vac_connect” function, it should 
call “timeout_in_progress=false” to unset the global variable which was set 
after “vac_disconnect” called.   

   I will try gerrit. Thanks for reading.



> 在 2019年6月12日，上午1:17，Ole Troan  写道：
> 
> Hi Jin,
> 
> Thank you for the describtion.
> Great if you can push your patch to gerrit and I will review and merge.
> 
> Best regards,
> Ole
> 
>> On 4 Jun 2019, at 08:29, Jin Zhihua  wrote:
>> 
>> Dear all:
>>I am developing a vpp client built with govpp by using “AsyncConnect” 
>> function. The first time I run the client before or after vpp, everything is 
>> ok. After I restart vpp while keeping the client running, the CPU usage of 
>> the client becomes high to around 70%. 
>>   I have found the problem and solved it. I didn’t find any Jira ticket 
>> entry, so I posted here and I hope the developers would solve it in next 
>> release. Thanks.
>> 
>>  I used “top -H” and gdb to find the high cpu usage thread which is 
>> “vac_timeout_thread_fn” in file “src/vpp-api/client/client.c”. The first 
>> time it blocked in “pthread_cond_wait”, which won’t take cpu. 
>> but in second time, it runs the loop all the time because 
>> “timeout_in_progress” becomes true and “read_timeout” is still zero, which 
>> make “pthread_cond_timedwait” always returns. 
>> 
>> I found the problem is in “vac_disconnect” , it calls “set_timeout(0)”, 
>> which will make the issue above happend. 
>>  In fact, “vac_disconnect” should call “unset_timeout()” to make thread 
>> quit gracefully and keep the global variables connect.
>> 
>> 
>> -=-=-=-=-=-=-=-=-=-=-=-
>> Links: You receive all messages sent to this group.
>> 
>> View/Reply Online (#13253): https://lists.fd.io/g/vpp-dev/message/13253
>> Mute This Topic: https://lists.fd.io/mt/32015501/675193
>> Group Owner: vpp-dev+ow...@lists.fd.io
>> Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [otr...@employees.org]
>> -=-=-=-=-=-=-=-=-=-=-=-
> 

-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13264): https://lists.fd.io/g/vpp-dev/message/13264
Mute This Topic: https://lists.fd.io/mt/32015501/21656
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-

Re: [vpp-dev] IPv4 packet is dropped

[Dave Barach via Lists.Fd.Io]

Identify the input interface – “show interface” – identify the corresponding 
input node – “show run” - and then 
s/dpdk-input/actual-input-node-sourcing-traffic/ in the packet tracer “trace 
add” command.

Please take the time to study available tools.

D.

From: Esin Erenoğlu 
Sent: Wednesday, June 12, 2019 7:25 AM
To: Dave Barach (dbarach) 
Subject: Re: [vpp-dev] IPv4 packet is dropped

First I try "show trace ", then "trace add dpdk input 100" command. After 
running '' tcpreplay - i vpp1out Facebook.pcap" command, I looked buffer with 
show trace command. There is no packet in buffer and still ip4 packets are 
dropping.

12 Haz 2019 Çar 14:12 tarihinde Dave Barach (dbarach) 
mailto:dbar...@cisco.com>> şunu yazdı:
If I had to bet: the FIB is not programmed correctly for the traffic that 
you're sending.

"show error" or "trace add dpdk-input 100" - start pcap replay - "show trace".

Please take a look at 
https://fdio-vpp.readthedocs.io/en/latest/troubleshooting/reportingissues/reportingissues.html.


From: vpp-dev@lists.fd.io 
mailto:vpp-dev@lists.fd.io>> on behalf of Esin Erenoğlu 
mailto:erenoglue...@gmail.com>>
Sent: Wednesday, June 12, 2019 6:51 AM
To: vpp-dev@lists.fd.io
Subject: [vpp-dev] IPv4 packet is dropped

Hi everyone,

I try to send pcap file to host-interface in vpp with tcpreplay . When I try to 
send pcap file, ipv4 packets dropped. When I try to ping host-interface, It 
doens't drop. Why ipv4 packets drop in vpp? How can I fix this?
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13263): https://lists.fd.io/g/vpp-dev/message/13263
Mute This Topic: https://lists.fd.io/mt/32039541/21656
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-

Re: [vpp-dev] timer wheel usage queries

[Dave Barach via Lists.Fd.Io]

You didn't mention calling tw_timer_expire_timers_xxx(tw, now), which is the 
only force of physics which will actually cause a timer to expire.

See also .../src/vppinfra/test_tw_timer.c, src/vlib/main.c or 
src/vnet/tcp/tcp.c for usage examples. The timer wheel code has been tested to 
N=60 million concurrent timers. It's not trivially broken.

D


From: vpp-dev@lists.fd.io  on behalf of hari_akkin via 
Lists.Fd.Io 
Sent: Wednesday, June 12, 2019 7:00 AM
To: vpp-dev@lists.fd.io
Cc: vpp-dev@lists.fd.io
Subject: [vpp-dev] timer wheel usage queries

Hello
I am trying to use timer_wheel framework to start a timer. I followed the 
examples given in fdio, but the expiry_callback function never being called 
back.
I am not sure if timer failed to expire or my usage has problems. I used the 
below functions:

Timer wheel Init:
{
tw_timer_wheel_1t_1w_2048sl_t* tw = &example_main->timer_wheels;
tw_timer_wheel_init_1t_1w_2048sl (tw, example_timer_expired_callback, 
1.0 /* timer period 1s */ , 1024);
tw->last_run_time = vlib_time_now (vm);
}

Timer Start:
handle = tw_timer_start_1t_1w_2048sl(&example->timer_wheels,1,0,5);

The expiry functiion example_timer_expired_callback never been called.

Is there any issue in the way I am using or is there any knwon issue in TW 
framework?

-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13262): https://lists.fd.io/g/vpp-dev/message/13262
Mute This Topic: https://lists.fd.io/mt/32039603/21656
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-

Re: [vpp-dev] IPv4 packet is dropped

[Dave Barach via Lists.Fd.Io]

If I had to bet: the FIB is not programmed correctly for the traffic that 
you're sending.

"show error" or "trace add dpdk-input 100" - start pcap replay - "show trace".

Please take a look at 
https://fdio-vpp.readthedocs.io/en/latest/troubleshooting/reportingissues/reportingissues.html.


From: vpp-dev@lists.fd.io  on behalf of Esin Erenoğlu 

Sent: Wednesday, June 12, 2019 6:51 AM
To: vpp-dev@lists.fd.io
Subject: [vpp-dev] IPv4 packet is dropped

Hi everyone,

I try to send pcap file to host-interface in vpp with tcpreplay . When I try to 
send pcap file, ipv4 packets dropped. When I try to ping host-interface, It 
doens't drop. Why ipv4 packets drop in vpp? How can I fix this?
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13261): https://lists.fd.io/g/vpp-dev/message/13261
Mute This Topic: https://lists.fd.io/mt/32039541/21656
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-

[vpp-dev] timer wheel usage queries

[hari_akkin via Lists.Fd.Io]

Hello
I am trying to use timer_wheel framework to start a timer. I followed the 
examples given in fdio, but the expiry_callback function never being called 
back.
I am not sure if timer failed to expire or my usage has problems. I used the 
below functions:

Timer wheel Init:
{
tw_timer_wheel_1t_1w_2048sl_t* tw = &example_main->timer_wheels;
tw_timer_wheel_init_1t_1w_2048sl (tw, example_timer_expired_callback, 1.0 /* 
timer period 1s */ , 1024);
tw->last_run_time = vlib_time_now (vm);
}

Timer Start:
handle = tw_timer_start_1t_1w_2048sl(&example->timer_wheels,1,0,5);

The expiry functiion example_timer_expired_callback never been called.

Is there any issue in the way I am using or is there any knwon issue in TW 
framework?
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13260): https://lists.fd.io/g/vpp-dev/message/13260
Mute This Topic: https://lists.fd.io/mt/32039603/21656
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-

[vpp-dev] IPv4 packet is dropped

[Esin Erenoğlu]

Hi everyone,

I try to send pcap file to host-interface in vpp with tcpreplay . When I
try to send pcap file, ipv4 packets dropped. When I try to ping
host-interface, It doens't drop. Why ipv4 packets drop in vpp? How can I
fix this?
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13259): https://lists.fd.io/g/vpp-dev/message/13259
Mute This Topic: https://lists.fd.io/mt/32039541/21656
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-