[Click] Re: [Roofnet] click.roofnet statically compiled

[John Bicket]

Hi Andrea,

The retries should probably be a element configuration option,
but I just hadn't gotten that far.  
I put them in wifi.h because we were testing different rate control
algorithms and I wanted to make sure they all use the same max_retry
values (hence WIFI_MAX_RETRIES).

The atheros cards can specifiy a schedule of bitrates to send at on
subsequent retries (ie send 11 mbps the first try, then do 5.5 for 3).
In the wifi_extra_header, this what the max_retries and alt_max_retries do.
If alt_rate and max_retries are nonzero, then after the packet exceeds
max_retries, the card will change to alt_rate for alt_max_retries.

Let me know if that doesn't help.

--john


Andrea Marasco [marasco3000 at yahoo.com] wrote:
>Hi John,
>
>it worked fine with your options. Thanks.
>CXXFLAGS=-static is the way to go.
>
>I'd like to test roofnet varying some MAC-layer
>parameters. I found several of them in wifi.h but I
>have a question regarding the number of link-level
>retransmissions. I found two variables max_retries and
>max_alt_retries in several files (autoratefallback.cc,
>madwifirate.cc, probetxrate.cc and setrate.cc). Which
>is the difference between the two variables? From my
>understanding I shouldn't vary the retries value in
>probetxrate.cc but I should change the values in the
>other files. Is that correct?
>
>I'd really appreciate your guidance on that.
>Thanks for your help.
>
>Best,
>
>Andrea
>
>--- John Bicket <jbicket at amsterdam.lcs.mit.edu> wrote:
>
>> Hi Andrea,
>> 
>> Here is what I run configure with to run on our
>> nodes without glibc:
>> /home/am5/jbicket/co/click/configure
>> --enable-userlevel --enable-wifi
>> --disable-linuxmodule --enable-tools=no
>> CXXFLAGS=-static
>> 
>> Let me know how that works.
>> 
>> I'm not sure if --enable-static and --disable-shared
>> work.. Maybe someone knows on the click list.
>> 
>> --john
>> 
>> Andrea Marasco [marasco3000 at yahoo.com] wrote:
>> >Thanks John,
>> > 
>> >I compiled click but I am having some portability
>> problems due to shared library. I noticed that your
>> executable click.roofnet is compiled statically.
>> > 
>> >I tried to compiled click with no shared libraries
>> running
>> > 
>> >./configure --disable-sahred --enalbe-static
>> --target=i386-pc-linux-gnu 
>> --prefix=/roof-distr-directory --enable-userlevel
>> --enable-wifi --disable-linuxmodule
>> >
>> >and than running
>> > 
>> >gmake all
>> > 
>> >However, the executable I get still needs some
>> shared libraries from my host distribution. Could
>> you please tell me what options you use to compile
>> statically the executable for the roofnet
>> distribution?
>> > 
>> >Thanks,
>> > 
>> >Andrea
>> > 
>> >
>> >John Bicket <jbicket at amsterdam.lcs.mit.edu> wrote:
>> >We get the same huge binary - if you run "strip" on
>> the binary, it'll 
>> >be less than 2 megs.
>> >
>> >Thanks for the fix - I checked it in.
>> >
>> >--john
>> >
>> >
>> >Andrea Marasco [marasco3000 at yahoo.com] wrote:
>> >>Thanks John,
>> >>
>> >>now it works great!
>> >>
>> >>Compiling click from source I got an executable
>> file 
>> >>CLICKDIR/userlevel/click of 14 MB that it is much
>> >>bigger than bin/click.roofnet in the tarball (just
>> >>1.7MB). Could you please explain me how I can
>> compile
>> >>click from source and get the smaller version of
>> the
>> >>executable?
>> >>
>> >>Thanks for you help.
>> >>
>> >>Andrea
>> >>
>> >>
>> >>P.S.
>> >>
>> >>Check scripts/Roofnet.pm line 200, you are missing
>> a
>> >>$gw_ad in the return statement.
>> >>It should be like this:
>> >>return ($gw_dev, $gw_ip, $gw_nm, $gw_gw, $gw_ad,
>> >>$download_cap, $upload_cap);
>> >>
>> >>--- John Bicket wrote:
>> >>
>> >>> 
>> >>> Hi Andrea,
>> >>> Andrea Marasco [marasco3000 at yahoo.com] wrote:
>> >>> >Hi,
>> >>> >
>> >>> >I upgraded to your latest tarball. The nodes
>> work.
>> >>> I
>> >>> >am able to ping other nodes and ssh into them.
>> >>> >However, the /click directory is completely
>> empty.
>> >>> >Therefore, I am not able to look at the routing
>> >>> table,
>> >>> >etc...
>> >>> >
>> >>> >Do you have any idea why that might happen?
>> >>> 
>> >>> This is because we're running at userlevel now -
>> we
>> >>> use the scripts
>> >>> /home/roofnet/scripts/read_handler.pl and 
>> >>> '' write_handler.pl
>> >>> to access the handlers that were in click.
>> >>> (for instance, we run 
>> >>> read_handler.pl srcr/es.bcast_stats 
>> >>> to look at neighbor info and
>> >>> read_handler.pl srcr/lt.routes
>> >>> for the routes).
>> >>> 
>> >>> 
>> >>> You can also just telnet to port 7777 for an
>> >>> interactive session with the
>> >>> ControlSocket.
>> >>> 
>> >>> >
>> >>> >Moreover, I think you should take a look at
>> >>> >find_srcr.pl. In my case the 5.x.x.x device is
>> >>> tun0
>> >>> >that is not in the search space of
>> find_srcr.pl.
>> >>> 
>> >>> I just fixed this. Thanks for letting us know.
>> >>> 
>> >>> >
>> >>> >Thanks,
>> >>> >
>> >>> >Andrea
>> >>> >
>> >>> >
>> >>>
>> >__________________________________________________
>> >>> >Do You Yahoo!?
>> >>> >Tired of spam? Yahoo! Mail has the best spam
>> >>> protection around 
>> >>> >http://mail.yahoo.com 
>> >>> 
>> >>
>> >>
>> >> 
>> >>__________________________________ 
>> >>Do you Yahoo!? 
>> >>Check out the new Yahoo! Front Page. 
>> >>www.yahoo.com 
>> >> 
>> >
>> >__________________________________________________
>> >Do You Yahoo!?
>> >Tired of spam?  Yahoo! Mail has the best spam
>> protection around 
>> >http://mail.yahoo.com 
>> 
>
>
>		
>__________________________________ 
>Do you Yahoo!? 
>Check out the new Yahoo! Front Page. 
>www.yahoo.com 
>