From antoniehenning at yahoo.com  Mon Oct  1 21:32:35 2012
From: antoniehenning at yahoo.com (Antonie Henning)
Date: Mon, 1 Oct 2012 18:32:35 -0700 (PDT)
Subject: [Click] ToDevice(ppp0)
In-Reply-To: <1348783120.38385.YahooMailNeo@web125303.mail.ne1.yahoo.com>
References: <1348783120.38385.YahooMailNeo@web125303.mail.ne1.yahoo.com>
Message-ID: <1349141555.73370.YahooMailNeo@web125305.mail.ne1.yahoo.com>

Using RawSocket as a replacement for ToDevice accomplishes this goal. While not exactly?eloquent,?its infinitely faster than libnetfilter_queue


________________________________
 From: Antonie Henning <antoniehenning at yahoo.com>
To: "click at pdos.csail.mit.edu" <click at pdos.csail.mit.edu> 
Sent: Thursday, September 27, 2012 11:58 PM
Subject: [Click] ToDevice(ppp0)
 
Hi,

Has anyone ever been successful in sending packets out a ppp interface with userlevel ToDevice??

ToDevice(ppp0,METHOD LINUX) attempts to send the packets but they are dropped by the kernel. METHOD PCAP fails with "Invalid Argument".

The PPPencap element appears not to address the CRC16 and Flags to complete the PPP encapsulation. I've created elements to add these but the drops persist.

Any advise will be appreciated.

A
_______________________________________________
click mailing list
click at amsterdam.lcs.mit.edu
https://amsterdam.lcs.mit.edu/mailman/listinfo/click

From yuanbo.cn at 163.com  Mon Oct  8 21:56:26 2012
From: yuanbo.cn at 163.com (yuanbo.cn)
Date: Tue, 9 Oct 2012 09:56:26 +0800 (CST)
Subject: [Click] How to debug the element?
Message-ID: <149bafad.360b.13a433cc256.Coremail.yuanbo.cn@163.com>

Hi,   when I  write my own element, I  meet some wrong. How can I  take efficient ways to debug the element.  


May I use the click_chatter(message)  or  set breakpoint ?  and whether I must run  the click scripts to print the message?


Thanks.

From Markku.Savela at vtt.fi  Wed Oct 10 08:06:35 2012
From: Markku.Savela at vtt.fi (Markku Savela)
Date: Wed, 10 Oct 2012 15:06:35 +0300
Subject: [Click] Functions htonq and ntohq in <click/integers.hh>
Message-ID: <507564CB.6040905@vtt.fi>

hi,

integers.hh defines functions ntohq/htonq, but to me they
look identical? To me it looks like, in ntohq, the return

     return (((uint64_t)ntohl(lo)) << 32) | ntohl(hi);

should be replaced by

     return (((uint64_t)ntohl(hi)) << 32) | ntohl(lo);

???

From ekohler at gmail.com  Wed Oct 10 19:45:04 2012
From: ekohler at gmail.com (Eddie Kohler)
Date: Wed, 10 Oct 2012 19:45:04 -0400
Subject: [Click] Functions htonq and ntohq in <click/integers.hh>
In-Reply-To: <507564CB.6040905@vtt.fi>
References: <507564CB.6040905@vtt.fi>
Message-ID: <CAF0C5SJk-kLbWjfxC16hCD9zojJUGpbe22rLkgHPbgu7k50jLQ@mail.gmail.com>

Nope, the current version is correct, as you can see by actually running it.

Eddie


On Wed, Oct 10, 2012 at 8:06 AM, Markku Savela <Markku.Savela at vtt.fi> wrote:
> hi,
>
> integers.hh defines functions ntohq/htonq, but to me they
> look identical? To me it looks like, in ntohq, the return
>
>      return (((uint64_t)ntohl(lo)) << 32) | ntohl(hi);
>
> should be replaced by
>
>      return (((uint64_t)ntohl(hi)) << 32) | ntohl(lo);
>
> ???
> _______________________________________________
> click mailing list
> click at amsterdam.lcs.mit.edu
> https://amsterdam.lcs.mit.edu/mailman/listinfo/click

From Markku.Savela at vtt.fi  Thu Oct 11 02:37:36 2012
From: Markku.Savela at vtt.fi (Markku Savela)
Date: Thu, 11 Oct 2012 09:37:36 +0300
Subject: [Click] Functions htonq and ntohq in <click/integers.hh>
In-Reply-To: <CAF0C5SJk-kLbWjfxC16hCD9zojJUGpbe22rLkgHPbgu7k50jLQ@mail.gmail.com>
References: <507564CB.6040905@vtt.fi>
	<CAF0C5SJk-kLbWjfxC16hCD9zojJUGpbe22rLkgHPbgu7k50jLQ@mail.gmail.com>
Message-ID: <50766930.7050006@vtt.fi>

On 10/11/2012 02:45 AM, Eddie Kohler wrote:
> Nope, the current version is correct, as you can see by actually running it.

Well, what I was doing was

- using on one machine (64 bit MISP) htonq (click)

- decoding on another (x86) using ntonq (Qt Application)

Didn't get the right results until I changed the
ntonq return on x86 machine. Oh well, must have some
error in my thought process then... will explore more.




From Markku.Savela at vtt.fi  Thu Oct 11 02:42:04 2012
From: Markku.Savela at vtt.fi (Markku Savela)
Date: Thu, 11 Oct 2012 09:42:04 +0300
Subject: [Click] Functions htonq and ntohq in <click/integers.hh>
In-Reply-To: <50766930.7050006@vtt.fi>
References: <507564CB.6040905@vtt.fi>
	<CAF0C5SJk-kLbWjfxC16hCD9zojJUGpbe22rLkgHPbgu7k50jLQ@mail.gmail.com>
	<50766930.7050006@vtt.fi>
Message-ID: <50766A3C.70203@vtt.fi>

.... sigh, I've some dyslexia apparently...

On 10/11/2012 09:37 AM, Markku Savela wrote:
> On 10/11/2012 02:45 AM, Eddie Kohler wrote:
> - using on one machine (64 bit MISP) htonq (click)

MIPS

> - decoding on another (x86) using ntonq (Qt Application)

ntohq !



From Markku.Savela at vtt.fi  Thu Oct 11 03:08:38 2012
From: Markku.Savela at vtt.fi (Markku Savela)
Date: Thu, 11 Oct 2012 10:08:38 +0300
Subject: [Click] Functions htonq and ntohq in <click/integers.hh>
In-Reply-To: <50766930.7050006@vtt.fi>
References: <507564CB.6040905@vtt.fi>
	<CAF0C5SJk-kLbWjfxC16hCD9zojJUGpbe22rLkgHPbgu7k50jLQ@mail.gmail.com>
	<50766930.7050006@vtt.fi>
Message-ID: <50767076.5070502@vtt.fi>

On 10/11/2012 09:37 AM, Markku Savela wrote:
> On 10/11/2012 02:45 AM, Eddie Kohler wrote

> Oh well, must have some
> error in my thought process then... will explore more.

Ok, something fishy on the MIPS (Octeon) side of my code

I've in packet payload...

     uint64_t in_packets;

and before sending the packet out, I do something like

    x.in_packets = htonq(x.in_packets)

when the in_packets is 7, the wireshark shows it as

... 00 00 00 07 00 00 00 00 ...

Apparently MIPS in big endian mode and calling htonq is not
correct. I was using htonq in same sense as "htonl", which
always works, whether host is big or little endian.




From Markku.Savela at vtt.fi  Thu Oct 11 04:02:06 2012
From: Markku.Savela at vtt.fi (Markku Savela)
Date: Thu, 11 Oct 2012 11:02:06 +0300
Subject: [Click] Functions htonq and ntohq in <click/integers.hh>
In-Reply-To: <50767076.5070502@vtt.fi>
References: <507564CB.6040905@vtt.fi>
	<CAF0C5SJk-kLbWjfxC16hCD9zojJUGpbe22rLkgHPbgu7k50jLQ@mail.gmail.com>
	<50766930.7050006@vtt.fi> <50767076.5070502@vtt.fi>
Message-ID: <50767CFE.3090208@vtt.fi>

On 10/11/2012 10:08 AM, Markku Savela wrote:

> Apparently MIPS in big endian mode and calling htonq is not
> correct. I was using htonq in same sense as "htonl", which
> always works, whether host is big or little endian.

If anyone is insterested, following diff is my attempt to make
htonq/ntohq safe to call regardless of the host endianness?
(works for my MIPS -> x86 case, but of course, these things
are tricky, so some one needs to check this, or think a
better implementation).



-------------- next part --------------
A non-text attachment was scrubbed...
Name: htonq.diff
Type: text/x-patch
Size: 1068 bytes
Desc: not available
Url : http://amsterdam.lcs.mit.edu/pipermail/click/attachments/20121011/c84bed11/attachment.bin 

From ekohler at gmail.com  Thu Oct 11 11:43:29 2012
From: ekohler at gmail.com (Eddie Kohler)
Date: Thu, 11 Oct 2012 11:43:29 -0400
Subject: [Click] Functions htonq and ntohq in <click/integers.hh>
In-Reply-To: <50767CFE.3090208@vtt.fi>
References: <507564CB.6040905@vtt.fi>
	<CAF0C5SJk-kLbWjfxC16hCD9zojJUGpbe22rLkgHPbgu7k50jLQ@mail.gmail.com>
	<50766930.7050006@vtt.fi> <50767076.5070502@vtt.fi>
	<50767CFE.3090208@vtt.fi>
Message-ID: <CAF0C5SJQ768eYzQ+yvVn-bJMe7ZHut=FYSTBGX7fbHT6AhLSVQ@mail.gmail.com>

Ouch! Does the diff I just checked in work for you?

E


On Thu, Oct 11, 2012 at 4:02 AM, Markku Savela <Markku.Savela at vtt.fi> wrote:
> On 10/11/2012 10:08 AM, Markku Savela wrote:
>
>> Apparently MIPS in big endian mode and calling htonq is not
>> correct. I was using htonq in same sense as "htonl", which
>> always works, whether host is big or little endian.
>
>
> If anyone is insterested, following diff is my attempt to make
> htonq/ntohq safe to call regardless of the host endianness?
> (works for my MIPS -> x86 case, but of course, these things
> are tricky, so some one needs to check this, or think a
> better implementation).
>
>
>

From Markku.Savela at vtt.fi  Fri Oct 12 02:56:13 2012
From: Markku.Savela at vtt.fi (Markku Savela)
Date: Fri, 12 Oct 2012 09:56:13 +0300
Subject: [Click] Functions htonq and ntohq in <click/integers.hh>
In-Reply-To: <CAF0C5SJQ768eYzQ+yvVn-bJMe7ZHut=FYSTBGX7fbHT6AhLSVQ@mail.gmail.com>
References: <507564CB.6040905@vtt.fi>
	<CAF0C5SJk-kLbWjfxC16hCD9zojJUGpbe22rLkgHPbgu7k50jLQ@mail.gmail.com>
	<50766930.7050006@vtt.fi> <50767076.5070502@vtt.fi>
	<50767CFE.3090208@vtt.fi>
	<CAF0C5SJQ768eYzQ+yvVn-bJMe7ZHut=FYSTBGX7fbHT6AhLSVQ@mail.gmail.com>
Message-ID: <5077BF0D.9010807@vtt.fi>

On 10/11/2012 06:43 PM, Eddie Kohler wrote:
> Ouch! Does the diff I just checked in work for you?

Seems to work (just copied the integers.hh, as I didn't want
to start messing up with git merges at this point).

Just wondering: what about machines where endiannes can be
chosen dynamically (ARM?). Now you must at cross-compile time
know what endiannes your target is set for?

Would it be even possible for gcc to compile "endiannes
independent" code (e.g. not knowing the endiannes at
compile time)?



From ekohler at gmail.com  Fri Oct 12 07:54:47 2012
From: ekohler at gmail.com (Eddie Kohler)
Date: Fri, 12 Oct 2012 07:54:47 -0400
Subject: [Click] Functions htonq and ntohq in <click/integers.hh>
In-Reply-To: <5077BF0D.9010807@vtt.fi>
References: <507564CB.6040905@vtt.fi>
	<CAF0C5SJk-kLbWjfxC16hCD9zojJUGpbe22rLkgHPbgu7k50jLQ@mail.gmail.com>
	<50766930.7050006@vtt.fi> <50767076.5070502@vtt.fi>
	<50767CFE.3090208@vtt.fi>
	<CAF0C5SJQ768eYzQ+yvVn-bJMe7ZHut=FYSTBGX7fbHT6AhLSVQ@mail.gmail.com>
	<5077BF0D.9010807@vtt.fi>
Message-ID: <CAF0C5S+AjTi_zU4PFbzSTWGfTqnB0dsg87fQH+9bqoAs_TXZFA@mail.gmail.com>

GCC could potentially do it but Click is certainly not set up for such
things and neither I believe is the Linux kernel (__constant_htons)

E


On Fri, Oct 12, 2012 at 2:56 AM, Markku Savela <Markku.Savela at vtt.fi> wrote:
> On 10/11/2012 06:43 PM, Eddie Kohler wrote:
>>
>> Ouch! Does the diff I just checked in work for you?
>
>
> Seems to work (just copied the integers.hh, as I didn't want
> to start messing up with git merges at this point).
>
> Just wondering: what about machines where endiannes can be
> chosen dynamically (ARM?). Now you must at cross-compile time
> know what endiannes your target is set for?
>
> Would it be even possible for gcc to compile "endiannes
> independent" code (e.g. not knowing the endiannes at
> compile time)?
>
>

From j.herzen at gmail.com  Sun Oct 14 08:57:41 2012
From: j.herzen at gmail.com (Julien Herzen)
Date: Sun, 14 Oct 2012 14:57:41 +0200
Subject: [Click] Memory allocation for Click elements
Message-ID: <CAFyvpGAV4DjGLX0vYALSBJhdkSDVJ-d7HFAWxUxNUOuamkREGA@mail.gmail.com>

Dear all,

I'm writing my own elements (in userspace), and I compile everything
with "make". It works fine most of the time.

It can happen, however, that adding a single class variable to an
element (e.g., a Timer) causes click to systematically segfault (even
before calling the constructors).
When I do a "make clean" before "make", I don't have this problem.
Therefore, it seems that there is more to re-compile than only my own
element. There's probably something missing in the default Makefile,
but I didn't figure out what yet.

Is it a known issue? Or am I the only one to have this problem?

I should perhaps mention that I'm cross-compiling for openWrt.

Many thanks for your help,
Julien

From christian at icir.org  Mon Oct 15 18:31:06 2012
From: christian at icir.org (Christian Kreibich)
Date: Mon, 15 Oct 2012 15:31:06 -0700
Subject: [Click] Click as a router between real interface and loopback
Message-ID: <507C8EAA.6060702@icir.org>

I have a pretty standard Click setup in which a user-space Click running
on Linux realizes a basic router between two interfaces (eth0 and eth1),
grabbing and sending packets via FromDevice and ToDevice. When testing
this setup with a client process located on machine A, a server process
located on machine B, and the machine running Click sitting in between,
everything works as expected.

For various reasons it would be handy to run the server process and the
Click instance on the same host, with the server listening on loopback.
When I try this, packets appear on lo in tcpdump, but the server process
doesn't see them. I seem to recall that packets sent via ToDevice will
not be processed by the local kernel. Is that in fact the case, and if
so, what's a good way to make this work? Do I need to go the KernelTun
or ToHost route?

Thanks,
-C.

From j.herzen at gmail.com  Tue Oct 16 04:11:22 2012
From: j.herzen at gmail.com (Julien Herzen)
Date: Tue, 16 Oct 2012 10:11:22 +0200
Subject: [Click] Click as a router between real interface and loopback
In-Reply-To: <507C8EAA.6060702@icir.org>
References: <507C8EAA.6060702@icir.org>
Message-ID: <CAFyvpGAqXvnXNivG5iwuk5Z-OCAqioYpPgnmSitr=w6S2cag3A@mail.gmail.com>

You can use FromHost / ToHost.
These elements create a new TUN/TAP network interface, on which your
server can be listening to. You just need to be careful to set the ip
and netmasks correctly (i.e., so that Linux routes traffic to that new
interface instead of the real one). You will also need to reply to the
ARP requests coming from Linux, and take care of the actual ARP
requests/replies from Click (see FromHost documentation).

Best,
Julien

On Tue, Oct 16, 2012 at 12:31 AM, Christian Kreibich <christian at icir.org> wrote:
> I have a pretty standard Click setup in which a user-space Click running
> on Linux realizes a basic router between two interfaces (eth0 and eth1),
> grabbing and sending packets via FromDevice and ToDevice. When testing
> this setup with a client process located on machine A, a server process
> located on machine B, and the machine running Click sitting in between,
> everything works as expected.
>
> For various reasons it would be handy to run the server process and the
> Click instance on the same host, with the server listening on loopback.
> When I try this, packets appear on lo in tcpdump, but the server process
> doesn't see them. I seem to recall that packets sent via ToDevice will
> not be processed by the local kernel. Is that in fact the case, and if
> so, what's a good way to make this work? Do I need to go the KernelTun
> or ToHost route?
>
> Thanks,
> -C.
> _______________________________________________
> click mailing list
> click at amsterdam.lcs.mit.edu
> https://amsterdam.lcs.mit.edu/mailman/listinfo/click

From Markku.Savela at vtt.fi  Wed Oct 17 06:19:39 2012
From: Markku.Savela at vtt.fi (Markku Savela)
Date: Wed, 17 Oct 2012 13:19:39 +0300
Subject: [Click] Input flush action for userspace FromDevice?
Message-ID: <507E863B.9080109@vtt.fi>

I run into a problem while using userspace FromDevice for WLAN
monitor socket: when I reconfigure WIFI interface, I need to
ignore all before reconfiguration messages queued in the kernel
for the click FromDevice element.

Apparently there is no such "native flush" action for a socket,
and flushing input is just reading and discarding until socket
indicates "no data".

I cannot find such function in FromDevice or is there? First
that comes to mind is some kind of flush handler that
would read and discard input, until none... (there is, of
course, a small concern of reading forever and never seeing
"no data" if new data for socket is generated faster than
click can read).

Any thoughts? Should handler been written or is there another
way to trigger the "input flush" for FromDevice from another
click element? Would handler be "read" or "write" handler?


From sandeep048 at gmail.com  Wed Oct 17 10:55:39 2012
From: sandeep048 at gmail.com (sandeep)
Date: Wed, 17 Oct 2012 15:55:39 +0100
Subject: [Click] click netmap(20120813) patch
Message-ID: <CAKE=LV1WMYr0QnPPAVeeYm3RHN0=fw5N6oKRZVC5G0Xk8Ks79g@mail.gmail.com>

Hi,

netmap_if structure is changed in netmap v3 (20120813). I am attaching a
patch to fix netmap's integration. This assumes that updated netmap.h and
netmap_user.h files are copied to /usr/include/net directory.

----------------------------------

--- /home/pathivas/click/elements/userlevel/netmapinfo.cc       2012-06-25
13:09:29.375014693 +0100
+++ ../elements/userlevel/netmapinfo.cc 2012-10-17 17:18:11.216540166 +0100
@@ -16,8 +16,6 @@
  * legally binding.
  */

-#include <iostream>
-
 #include <click/config.h>
 #include <click/glue.hh>
 #include "netmapinfo.hh"
@@ -94,7 +91,7 @@ NetmapInfo::ring::initialize_rings_rx(in
 {
     ring_begin = 0;
     // 0 means "same count as the converse direction"
-    ring_end = nifp->ni_rx_queues ? nifp->ni_rx_queues :
nifp->ni_tx_queues;
+    ring_end = nifp->ni_rx_rings ? nifp->ni_rx_rings : nifp->ni_tx_rings;
     if (timestamp >= 0) {
        int flags = (timestamp > 0 ? NR_TIMESTAMP : 0);
        for (unsigned i = ring_begin; i != ring_end; ++i)
@@ -106,7 +103,7 @@ void
 NetmapInfo::ring::initialize_rings_tx()
 {
     ring_begin = 0;
-    ring_end = nifp->ni_tx_queues ? nifp->ni_tx_queues :
nifp->ni_rx_queues;
+    ring_end = nifp->ni_tx_rings ? nifp->ni_tx_rings : nifp->ni_rx_rings;
 }

----------------------------------

Kind Regards,
Sandeep,
PhD Student,
CTVR <http://ctvr.ie/> - Trinity College Dublin

From Markku.Savela at vtt.fi  Fri Oct 19 01:47:37 2012
From: Markku.Savela at vtt.fi (Markku Savela)
Date: Fri, 19 Oct 2012 08:47:37 +0300
Subject: [Click] Input flush action for userspace FromDevice?
In-Reply-To: <507E863B.9080109@vtt.fi>
References: <507E863B.9080109@vtt.fi>
Message-ID: <5080E979.40505@vtt.fi>

On 10/17/2012 01:19 PM, Markku Savela wrote:

> I cannot find such function in FromDevice or is there? First
> that comes to mind is some kind of flush handler that
> would read and discard input, until none... (there is, of
> course, a small concern of reading forever and never seeing
> "no data" if new data for socket is generated faster than
> click can read).

Ok, here is my test for such handler. This is not anything
that could be incorporated as is: it is hard coded to assume
the linux method for reading -- I didn't have time to study
how to achieve same with PCAP or NETMAP...

I suppose I should have wrapped this in FROMDEVICE_ALLOW_LINUX
and check the selected _method too.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: flush.diff
Type: text/x-patch
Size: 1124 bytes
Desc: not available
Url : http://amsterdam.lcs.mit.edu/pipermail/click/attachments/20121019/f668aada/attachment.bin 

From zaykalov at gmail.com  Thu Oct 25 08:22:12 2012
From: zaykalov at gmail.com (Konstantin Zaykalov)
Date: Thu, 25 Oct 2012 13:22:12 +0100
Subject: [Click] kernel patching elimination
Message-ID: <CALdZA=WQ9o-U2YHJow-WJ7NHOtBsCS8zYwsDigeko5R=es36NQ@mail.gmail.com>

Hi,

I've got a few questions regarding the design of Click router:
1) Why does Click do kernel patching (and fixing include files via
fixincludes.pl script) for g++ compiler instead of providing wrapper
functions, surrounded by "extern "C" "? Or even some abstraction layer
written in C?
2) Why doesn't Click use netfilter hooks for intercepting packets?
3) Why doesn't Click use NAPI for device polling?

Basically I am looking into possibilities to eliminate kernel patching
altogether and make it more independent/standalone.

Thank you,

Konstantin