[Xenomai] Context Switch with rtnet

[Xenomai] Context Switch with rtnet

[Franz Engel]

Hi,

I've a little problem. I'm using the rtnet.h-header and the following function:
ret = recvfrom ( ethernetConnection.sock, &recBuffer, sizeof ( recBuffer ),0, ( struct sockaddr * ) &ethernetConnection.dest_addr,&destlen );


I compiled the program with the xenomai compiling flags and library.

When I look to the xenomai state I get something like this:
CPU  PID    MSW        CSW        PF    STAT       %CPU  NAME
 22  0      0          834840     0     00000082    0.1  rtnet-stack
 22  0      0          1          0     00000082    0.0  rtnet-rtpc
  0  6262   2          2          0     00300380    0.0  qs_robotpositio
  2  6265   2          6          0     00300380    0.0  qs_robotpositio
  0  6266   1          1          0     00300380    0.0  robotrsiinterfa
  1  6268   1147       2365       0     00300182    0.8  robotrsiinterfa

You can see, I get a lote of context and module switches just as the rtnet-stack itself. Is that normal?

Thanks in advanced,
    Franz

Re: [Xenomai] Context Switch with rtnet

[Jan Kiszka]

On 2012-09-14 17:01, Franz Engel wrote:
> Hi,
> 
> I've a little problem. I'm using the rtnet.h-header and the following function:
> ret = recvfrom ( ethernetConnection.sock, &recBuffer, sizeof ( recBuffer ),0, ( struct sockaddr * ) &ethernetConnection.dest_addr,&destlen );
> 
> 
> I compiled the program with the xenomai compiling flags and library.
> 
> When I look to the xenomai state I get something like this:
> CPU  PID    MSW        CSW        PF    STAT       %CPU  NAME
>  22  0      0          834840     0     00000082    0.1  rtnet-stack
>  22  0      0          1          0     00000082    0.0  rtnet-rtpc
>   0  6262   2          2          0     00300380    0.0  qs_robotpositio
>   2  6265   2          6          0     00300380    0.0  qs_robotpositio
>   0  6266   1          1          0     00300380    0.0  robotrsiinterfa
>   1  6268   1147       2365       0     00300182    0.8  robotrsiinterfa
> 
> You can see, I get a lote of context and module switches just as the rtnet-stack itself. Is that normal?

Well, I have no clue what you configured, so I can only guess. If your
task 6268 is a pure RT task that is not supposed to interact with Linux
services, you first of all have a bug in your software stack. You should
look for the reason of those mode switches to Linux, e.g. by asking
Xenomai to raise SIGXCPU when this happens.

Regarding the rtnet-stack context switches: basically, every switch is
one packet arriving at your NIC. Either you are receiving a lot of small
but valid packets or you are on an open network, and that is basically
heavy unrelated broadcast traffic of other sources than the RT device(s).

Jan

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 259 bytes
Desc: OpenPGP digital signature
URL: <http://www.xenomai.org/pipermail/xenomai/attachments/20120916/21fcac64/attachment.pgp>

Re: [Xenomai] Context Switch with rtnet

[Franz Engel]

Hi Jan,

> > Hi,
> >
> > I've a little problem. I'm using the rtnet.h-header and the following
> function:
> > ret = recvfrom ( ethernetConnection.sock, &recBuffer, sizeof (
> > recBuffer ),0, ( struct sockaddr * )
> > &ethernetConnection.dest_addr,&destlen );
> >
> >
> > I compiled the program with the xenomai compiling flags and library.
> >
> > When I look to the xenomai state I get something like this:
> > CPU  PID    MSW        CSW        PF    STAT       %CPU  NAME
> >  22  0      0          834840     0     00000082    0.1  rtnet-stack
> >  22  0      0          1          0     00000082    0.0  rtnet-rtpc
> >   0  6262   2          2          0     00300380    0.0
> qs_robotpositio
> >   2  6265   2          6          0     00300380    0.0
> qs_robotpositio
> >   0  6266   1          1          0     00300380    0.0
> robotrsiinterfa
> >   1  6268   1147       2365       0     00300182    0.8
> robotrsiinterfa
> >
> > You can see, I get a lote of context and module switches just as the
> rtnet-stack itself. Is that normal?
> 
> Well, I have no clue what you configured, so I can only guess. If your
> task 6268 is a pure RT task that is not supposed to interact with Linux
> services, you first of all have a bug in your software stack. You
> should look for the reason of those mode switches to Linux, e.g. by
> asking Xenomai to raise SIGXCPU when this happens.

I use the SIGXCPU signal to look where I get a context switch. It is exactly
in the line of the recvfrom. I get the same result when I step through my
code. I debug from line to line and check the /proc/xenomai/state, so I can
see that the CSW and MSW increases when I step over the recvfrom. I tried my
code also with the native commands, like rt_recvfrom. But the result is the
same.

> 
> Regarding the rtnet-stack context switches: basically, every switch is
> one packet arriving at your NIC. Either you are receiving a lot of
> small but valid packets or you are on an open network, and that is
> basically heavy unrelated broadcast traffic of other sources than the
> RT device(s).
> 
> Jan

I'm connected to a robot that sends me his position every 4ms.

I send you a minimal example of my code. But I'm not sure if I understand
your question for the configuration correct. Depends the realtime capability
of rtnet only by my kdevelop project (cmake, *.c and *.h) or is it possible
that I miss some settings in the kernel? I start rtnet before I run my
program and I load the modules for my NIC and for UDP etc. Is there
something more what I have to do?

Thanks and regards,
    Franz

-------------- next part --------------
A non-text attachment was scrubbed...
Name: DemoCSW.tar.gz
Type: application/octet-stream
Size: 3601 bytes
Desc: not available
URL: <http://www.xenomai.org/pipermail/xenomai/attachments/20120920/c87f6809/attachment.obj>

Re: [Xenomai] Context Switch with rtnet

[Wolfgang Grandegger]

On 09/20/2012 05:24 PM, Franz Engel wrote:
> Hi Jan,
> 
>>> Hi,
>>>
>>> I've a little problem. I'm using the rtnet.h-header and the following
>> function:
>>> ret = recvfrom ( ethernetConnection.sock, &recBuffer, sizeof (
>>> recBuffer ),0, ( struct sockaddr * )
>>> &ethernetConnection.dest_addr,&destlen );
>>>
>>>
>>> I compiled the program with the xenomai compiling flags and library.
>>>
>>> When I look to the xenomai state I get something like this:
>>> CPU  PID    MSW        CSW        PF    STAT       %CPU  NAME
>>>  22  0      0          834840     0     00000082    0.1  rtnet-stack
>>>  22  0      0          1          0     00000082    0.0  rtnet-rtpc
>>>   0  6262   2          2          0     00300380    0.0
>> qs_robotpositio
>>>   2  6265   2          6          0     00300380    0.0
>> qs_robotpositio
>>>   0  6266   1          1          0     00300380    0.0
>> robotrsiinterfa
>>>   1  6268   1147       2365       0     00300182    0.8
>> robotrsiinterfa
>>>
>>> You can see, I get a lote of context and module switches just as the
>> rtnet-stack itself. Is that normal?
>>
>> Well, I have no clue what you configured, so I can only guess. If your
>> task 6268 is a pure RT task that is not supposed to interact with Linux
>> services, you first of all have a bug in your software stack. You
>> should look for the reason of those mode switches to Linux, e.g. by
>> asking Xenomai to raise SIGXCPU when this happens.
> 
> I use the SIGXCPU signal to look where I get a context switch. It is exactly
> in the line of the recvfrom. I get the same result when I step through my
> code. I debug from line to line and check the /proc/xenomai/state, so I can
> see that the CSW and MSW increases when I step over the recvfrom. I tried my
> code also with the native commands, like rt_recvfrom. But the result is the
> same.

What is the reason for the mode switch? How to get it see:

http://www.rts.uni-hannover.de/xenomai/lxr/source/examples/native/sigdebug.c

Wolfgang.

Re: [Xenomai] Context Switch with rtnet

[Gilles Chanteperdrix]

On 09/20/2012 05:24 PM, Franz Engel wrote:
> Hi Jan,
> 
>>> Hi,
>>>
>>> I've a little problem. I'm using the rtnet.h-header and the following
>> function:
>>> ret = recvfrom ( ethernetConnection.sock, &recBuffer, sizeof (
>>> recBuffer ),0, ( struct sockaddr * )
>>> &ethernetConnection.dest_addr,&destlen );
>>>
>>>
>>> I compiled the program with the xenomai compiling flags and library.
>>>
>>> When I look to the xenomai state I get something like this:
>>> CPU  PID    MSW        CSW        PF    STAT       %CPU  NAME
>>>  22  0      0          834840     0     00000082    0.1  rtnet-stack
>>>  22  0      0          1          0     00000082    0.0  rtnet-rtpc
>>>   0  6262   2          2          0     00300380    0.0
>> qs_robotpositio
>>>   2  6265   2          6          0     00300380    0.0
>> qs_robotpositio
>>>   0  6266   1          1          0     00300380    0.0
>> robotrsiinterfa
>>>   1  6268   1147       2365       0     00300182    0.8
>> robotrsiinterfa
>>>
>>> You can see, I get a lote of context and module switches just as the
>> rtnet-stack itself. Is that normal?
>>
>> Well, I have no clue what you configured, so I can only guess. If your
>> task 6268 is a pure RT task that is not supposed to interact with Linux
>> services, you first of all have a bug in your software stack. You
>> should look for the reason of those mode switches to Linux, e.g. by
>> asking Xenomai to raise SIGXCPU when this happens.
> 
> I use the SIGXCPU signal to look where I get a context switch. It is exactly
> in the line of the recvfrom. I get the same result when I step through my
> code. I debug from line to line and check the /proc/xenomai/state, so I can
> see that the CSW and MSW increases when I step over the recvfrom. I tried my
> code also with the native commands, like rt_recvfrom. But the result is the
> same.

Stepping in gdb causes switches to secondary mode (each step is a switch
to secondary mode).

Anyway, you did not tell us:
- what version of xenomai
- what version of the Linux kernel
- what version of the I-pipe patch
- what architecture
- what NIC

you are using.

-- 
					    Gilles.

Re: [Xenomai] Context Switch with rtnet

[Kai Bollue]

On 14.09.2012 17:01, Franz Engel wrote:
> Hi,
>
> I've a little problem. I'm using the rtnet.h-header and the following function:
> ret = recvfrom ( ethernetConnection.sock, &recBuffer, sizeof ( recBuffer ),0, ( struct sockaddr * ) &ethernetConnection.dest_addr,&destlen );

Hi,

stupid question: But you do use an rtnet socket, not a normal one, do 
you? In this case, shouldn't it be rt_dev_recvfrom(...)?


Kai

Re: [Xenomai] Context Switch with rtnet

[Gilles Chanteperdrix]

On 09/21/2012 11:58 AM, Kai Bollue wrote:
> On 14.09.2012 17:01, Franz Engel wrote:
>> Hi,
>>
>> I've a little problem. I'm using the rtnet.h-header and the following function:
>> ret = recvfrom ( ethernetConnection.sock, &recBuffer, sizeof ( recBuffer ),0, ( struct sockaddr * ) &ethernetConnection.dest_addr,&destlen );
> 
> Hi,
> 
> stupid question: But you do use an rtnet socket, not a normal one, do 
> you? In this case, shouldn't it be rt_dev_recvfrom(...)?

When using the posix skin, recvfrom is substituted at compilation (or
more exactly link-edit) time with __wrap_recvfrom, implemented in
libpthread_rt.so, allowing to work with real-time sockets.

(But that works only if you used socket too instead of rt_dev_socket).

-- 
					    Gilles.

Re: [Xenomai] Context Switch with rtnet

[Kai Bollue]

On 21.09.2012 12:02, Gilles Chanteperdrix wrote:
> On 09/21/2012 11:58 AM, Kai Bollue wrote:
>> On 14.09.2012 17:01, Franz Engel wrote:
>>> Hi,
>>>
>>> I've a little problem. I'm using the rtnet.h-header and the following function:
>>> ret = recvfrom ( ethernetConnection.sock, &recBuffer, sizeof ( recBuffer ),0, ( struct sockaddr * ) &ethernetConnection.dest_addr,&destlen );
>> Hi,
>>
>> stupid question: But you do use an rtnet socket, not a normal one, do
>> you? In this case, shouldn't it be rt_dev_recvfrom(...)?
> When using the posix skin, recvfrom is substituted at compilation (or
> more exactly link-edit) time with __wrap_recvfrom, implemented in
> libpthread_rt.so, allowing to work with real-time sockets.
>

Ah, sorry, I missed that.

Kai

[Xenomai] Context Switch with rtnet

[Franz Engel]

> On Fri, 21 Sep 2012 12:08:40 +0200, Kai Bollue <mlist1@bollue.de>
> > On 21.09.2012 12:02, Gilles Chanteperdrix wrote:
> >> On 09/21/2012 11:58 AM, Kai Bollue wrote:
> >>> On 14.09.2012 17:01, Franz Engel wrote:
> >>> Hi,
> >>>
> >>> I've a little problem. I'm using the rtnet.h-header and the
> following function:
> >>> ret = recvfrom ( ethernetConnection.sock, &recBuffer, sizeof (
> >>> recBuffer ),0, ( struct sockaddr * )
> >>> &ethernetConnection.dest_addr,&destlen );
> >> Hi,
> >>
> >> stupid question: But you do use an rtnet socket, not a normal one,
> do
> >> you? In this case, shouldn't it be rt_dev_recvfrom(...)?
> > When using the posix skin, recvfrom is substituted at compilation (or
> > more exactly link-edit) time with __wrap_recvfrom, implemented in
> > libpthread_rt.so, allowing to work with real-time sockets.
> >
> 
> Ah, sorry, I missed that.
> 

Sorry, I missed to answer on some questions:

Let me explain my problem again. When I run my program ("mytask") the CSW
and MSW increase every time the program receive a message:
CPU  PID    MSW        CSW        PF    STAT       %CPU  NAME
  0  0      0          396        0     00500080  100.0  ROOT/0
  1  0      0          0          0     00500080  100.0  ROOT/1
  0  1336   1          1          0     00300380    0.0  kogmo_man
  0  0      0          209        0     00000082    0.0  rtnet-stack
  0  0      0          1          0     00000082    0.0  rtnet-rtpc
  0  0      0          1          0     00000082    0.0  rtcfg-rx
  0  5182   1          15         0     00340182    0.0  mytask
  0  0      0          143        0     00000000    0.0  IRQ22: rteth0
  1  0      0          186        0     00000000    0.0  IRQ22: rteth0
  0  0      0          2054479    0     00000000    0.0  IRQ4364: [timer]
  1  0      0          547805     0     00000000    0.0  IRQ4364: [timer]

Wolfgang advised me to find the reason for this phenomenon like the program
on this link:
http://www.rts.uni-hannover.de/xenomai/lxr/source/examples/native/sigdebug.c
I tried it and modified my code (see the code further down). When I insert
an "cout" after "rt_task_set_mode(0, T_WARNSW, NULL);" I get a failure
message (so it should work). But when I run this program without the cout I
get no message, but the CSW and MSW counter are still increases.  


That's my versions:
rtnet 0.9.12
xenomai 2.5.5.2
linux kernel 2.6.32-20
ubuntu 10.04
i-pipe patch: adeos-ipipe-2.6.32.20-x86-2.7-03.patch
NIC driver: 8139too
CPU: Intel vPro


Any other ideas? Or is it not a problem but the normal state that the
counters increases?

Here is the code of my cmake-file:
cmake_minimum_required(VERSION 2.8)

##################################
# BEGIN OF EDITING THIS TEMPLATE #
##################################

############################################
# PLEASE ENTER HERE THE NAME OF YOUR PROJECT

set(ProjectName robotrsiVersuch)  #e.g. "robotdatalogger" (do not forget the
quotes)

PROJECT(${ProjectName})

###############################################
# ENTER HERE THE VERSION NUMBER OF YOUR PROJECT
set(${ProjectName}_VERSION_MAJOR 1)
set(${ProjectName}_VERSION_MINOR 0)

###############################################
# ENABLE OR DIABLE THE USE OF XENOMAI AND QT
option(USE_XENOMAI	"xenomai"		ON	)
option(USE_RTNET	"rtnet"			ON	)

#############################################
# NOW ADD ALL SOURCE FILES OF YOUR PROJECT
# IF YOU'RE USING QT, ALSO ADD ALL HEADERS OF 
# QT-OBJECTS, UI-FILES AND RESOURCES

#####################
# INCLUDE DIRECTORIES

set(MAIN_DIR ${PROJECT_SOURCE_DIR}/../..)
set(SOURCE_DIR src)


#############################################
# ENTER THE INSTALL DIRECTORY
# added by stefan.thumfart@profactor.at
set(INSTALLDIR ${MAIN_DIR}/bin)

set(INCLUDES
    ${SOURCE_DIR}
)


###########
# LIBRARIES

set(EXTRA_LIBS 

)

######
# CPPs

set(SOURCES 
	main.cpp
	src/rsiinterface.cpp
	src/rsiinterface.h
)

#############
# MOC-Headers
set(MOC_HEADERS

)

##########
# UI-Files
set(UIS

)

###########
# Resources
set(RESOURCES 

)

##################################
#  END OF EDITING THIS TEMPLATE  #
##################################


message (STATUS "Usage of Xenomai: ${USE_XENOMAI}")
message (STATUS "Usage of RTnet: ${USE_RTNET}")

message(STATUS "Generating buid files for project ${ProjectName}, version
${${ProjectName}_VERSION_MAJOR}.${${ProjectName}_VERSION_MINOR}")

include_directories(${INCLUDES})


IF(USE_XENOMAI)
    message(STATUS "Starting Xenomai configuration")

    message(STATUS "startign RTnet configuration")
    include_directories(/usr/local/rtnet/include)
    include_directories(/usr/xenomai/include/rtdm)
    message(STATUS "RTnet configuration finished")   

    find_library(XENO_rtdm rtdm /usr/xenomai/lib)
    set(EXTRA_LIBS ${EXTRA_LIBS} ${XENO_rtdm})
    message(STATUS ${EXTRA_LIBS} )

    set(XENOMAI_DIR "/usr/xenomai")
    set(XENOMAI_INCLUDE_DIR ${XENOMAI_DIR}/include)
    set(XENOMAI_INCLUDE_POSIX_DIR ${XENOMAI_DIR}/include/native)

    exec_program("${XENOMAI_DIR}/bin/xeno-config --skin=native --ldflags"
OUTPUT_VARIABLE XENOMAI_EXE_LINKER_FLAGS)
    set(CMAKE_EXE_LINKER_FLAGS ${CMAKE_EXE_LINKER_FLAGS}
${XENOMAI_EXE_LINKER_FLAGS})
    exec_program("${XENOMAI_DIR}/bin/xeno-config --skin=native --cflags"
OUTPUT_VARIABLE XENOMAI_DEFINITIONS)
    set(EXTRA_DEFINITIONS ${EXTRA_DEFINITIONS} ${XENOMAI_DEFINITIONS})
    ADD_DEFINITIONS(${EXTRA_DEFINITIONS})

    set(EXTRA_LIBS ${EXTRA_LIBS} rtdk)
    
    message(STATUS ${XENOMAI_EXE_LINKER_FLAGS})
    message(STATUS ${XENOMAI_DEFINITIONS})
    message(STATUS "Xenomai configuration finished")
ELSE(USE_XENOMAI)
    find_library(RT_TIME_LIB rt)
    set(EXTRA_LIBS ${EXTRA_LIBS} ${RT_TIME_LIB})
ENDIF(USE_XENOMAI)


IF (USE_RTNET)

ENDIF (USE_RTNET)


add_executable(${ProjectName} ${SOURCES})

TARGET_LINK_LIBRARIES(${ProjectName} ${EXTRA_LIBS})

install(TARGETS ${ProjectName} DESTINATION ${INSTALLDIR})

And this is my cpp:
#include <unistd.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <signal.h>
#include <getopt.h>
#include <execinfo.h>
#include <sys/mman.h>

#include <stdexcept>
#include <arpa/inet.h>
#include <native/task.h>
#include "rtdk.h"
#include <rtnet.h>


#define RECBUFFERSIZE 600
#define SENDBUFFERSIZE 300

RT_TASK task;
const char *reason_str[7];

struct conn_t
{
    int sock;
    struct sockaddr_in dest_addr;
    struct sockaddr_in local_addr;
};

struct conn_t getSocket ()
{
    struct conn_t ethernetConnection;

    //ethernetConnection.dest_addr.sin_family         = AF_INET;
    ethernetConnection.dest_addr.sin_addr.s_addr    = inet_addr
("192.168.100.12");
    ethernetConnection.dest_addr.sin_port           = htons (6002);
    
    //ethernetConnection.local_addr.sin_family        = AF_INET;
    ethernetConnection.local_addr.sin_addr.s_addr   = inet_addr
("192.168.100.11");
    ethernetConnection.local_addr.sin_port          = htons (6008);

    if ( ( ethernetConnection.sock = socket ( AF_INET, SOCK_DGRAM,
IPPROTO_UDP ) ) < 0 )
    {
        throw std::runtime_error("Can not open socket");
    }
    else
    {
    }

    if ( bind ( ethernetConnection.sock, ( struct sockaddr * )
&ethernetConnection.local_addr, sizeof ( ethernetConnection.local_addr ) ) <
0 )
    {
        close ( ethernetConnection.sock );
        throw std::runtime_error("Can not bind socket");
    }
    else
    {
    }

    return ethernetConnection;
}

void task_body (void *cookie)
{  
    char recBuffer[RECBUFFERSIZE];

    socklen_t destlen;
    
    struct conn_t ethernetConnection;
    
    ethernetConnection = getSocket();

    destlen = sizeof ( ethernetConnection.dest_addr );
    
    int k;
    
    rt_task_set_mode(0, T_WARNSW, NULL);
    
    for(k=0;k<1000;k++)
    {
        recvfrom( ethernetConnection.sock, &recBuffer, sizeof ( recBuffer
),0, ( struct sockaddr * ) &ethernetConnection.dest_addr,&destlen );
	
        //ret = rt_dev_sendto ( ethernetConnection.sock, &sendBuffer, strlen
( sendBuffer ) +1, 0, ( struct sockaddr * )
&ethernetConnection.dest_addr,destlen);
    }
}


void warn_upon_switch(int sig, siginfo_t *si, void *context)
{
    unsigned int reason = si->si_value.sival_int;
    void *bt[32];
    int nentries;
    
    printf("\nSIGDEBUG received, reason %d: %s\n", reason,
           reason <= SIGDEBUG_WATCHDOG ? reason_str[reason] : "<unknown>");
    /* Dump a backtrace of the frame which caused the switch to
       secondary mode: */
    nentries = backtrace(bt,sizeof(bt) / sizeof(bt[0]));
    backtrace_symbols_fd(bt,nentries,fileno(stdout));
}

int main (int argc, char **argv)
{
    struct sigaction sa;
    int err;

    reason_str[SIGDEBUG_UNDEFINED] = "undefined";
    reason_str[SIGDEBUG_MIGRATE_SIGNAL] = "received signal";
    reason_str[SIGDEBUG_MIGRATE_SYSCALL] = "invoked syscall";
    reason_str[SIGDEBUG_MIGRATE_FAULT] = "triggered fault";
    reason_str[SIGDEBUG_MIGRATE_PRIOINV] = "affected by priority inversion";
    reason_str[SIGDEBUG_NOMLOCK] = "missing mlockall";
    reason_str[SIGDEBUG_WATCHDOG] = "runaway thread";
    
    mlockall(MCL_CURRENT | MCL_FUTURE);
    
    //rt_print_auto_init(0);

    sigemptyset(&sa.sa_mask);
    sa.sa_sigaction = warn_upon_switch;
    sa.sa_flags = SA_SIGINFO;
    sigaction(SIGDEBUG, &sa, NULL);

    err = rt_task_create(&task,"mytask",0,1,T_FPU);

    if (err) {
      fprintf(stderr,"failed to create task, code %d\n",err);
      return 0;
    }

    err = rt_task_start(&task,&task_body,NULL);

    if (err) {
      fprintf(stderr,"failed to start task, code %d\n",err);
      return 0;
    }

    pause();

    return 0;
}

Re: [Xenomai] Context Switch with rtnet

[Gilles Chanteperdrix]

On 09/22/2012 03:20 PM, Franz Engel wrote:

>> On Fri, 21 Sep 2012 12:08:40 +0200, Kai Bollue <mlist1@bollue.de>
>>> On 21.09.2012 12:02, Gilles Chanteperdrix wrote:
>>>> On 09/21/2012 11:58 AM, Kai Bollue wrote:
>>>>> On 14.09.2012 17:01, Franz Engel wrote:
>>>>> Hi,
>>>>>
>>>>> I've a little problem. I'm using the rtnet.h-header and the
>> following function:
>>>>> ret = recvfrom ( ethernetConnection.sock, &recBuffer, sizeof (
>>>>> recBuffer ),0, ( struct sockaddr * )
>>>>> &ethernetConnection.dest_addr,&destlen );
>>>> Hi,
>>>>
>>>> stupid question: But you do use an rtnet socket, not a normal one,
>> do
>>>> you? In this case, shouldn't it be rt_dev_recvfrom(...)?
>>> When using the posix skin, recvfrom is substituted at compilation (or
>>> more exactly link-edit) time with __wrap_recvfrom, implemented in
>>> libpthread_rt.so, allowing to work with real-time sockets.
>>>
>>
>> Ah, sorry, I missed that.
>>
> 
> Sorry, I missed to answer on some questions:
> 
> Let me explain my problem again. When I run my program ("mytask") the CSW
> and MSW increase every time the program receive a message:
> CPU  PID    MSW        CSW        PF    STAT       %CPU  NAME
>   0  0      0          396        0     00500080  100.0  ROOT/0
>   1  0      0          0          0     00500080  100.0  ROOT/1
>   0  1336   1          1          0     00300380    0.0  kogmo_man
>   0  0      0          209        0     00000082    0.0  rtnet-stack
>   0  0      0          1          0     00000082    0.0  rtnet-rtpc
>   0  0      0          1          0     00000082    0.0  rtcfg-rx
>   0  5182   1          15         0     00340182    0.0  mytask
>   0  0      0          143        0     00000000    0.0  IRQ22: rteth0
>   1  0      0          186        0     00000000    0.0  IRQ22: rteth0
>   0  0      0          2054479    0     00000000    0.0  IRQ4364: [timer]
>   1  0      0          547805     0     00000000    0.0  IRQ4364: [timer]
> 
> Wolfgang advised me to find the reason for this phenomenon like the program
> on this link:
> http://www.rts.uni-hannover.de/xenomai/lxr/source/examples/native/sigdebug.c
> I tried it and modified my code (see the code further down). When I insert
> an "cout" after "rt_task_set_mode(0, T_WARNSW, NULL);" I get a failure
> message (so it should work). But when I run this program without the cout I
> get no message, but the CSW and MSW counter are still increases.  
> 
> 
> That's my versions:
> rtnet 0.9.12
> xenomai 2.5.5.2


That is not the latest release in the 2.5 branch. You are advised to
upgrade to 2.5.6, it should work with the same linux kernel version.

>

>     exec_program("${XENOMAI_DIR}/bin/xeno-config --skin=native --ldflags"
> OUTPUT_VARIABLE XENOMAI_EXE_LINKER_FLAGS)
>     set(CMAKE_EXE_LINKER_FLAGS ${CMAKE_EXE_LINKER_FLAGS}
> ${XENOMAI_EXE_LINKER_FLAGS})
>     exec_program("${XENOMAI_DIR}/bin/xeno-config --skin=native --cflags"
> OUTPUT_VARIABLE XENOMAI_DEFINITIONS)


Here you are retrieving the flags for the native skin, not for the posix
skin, unless there is some other part of your makefile which gets them.

>     if ( ( ethernetConnection.sock = socket ( AF_INET, SOCK_DGRAM,
> IPPROTO_UDP ) ) < 0 )

> (...)
>     if ( bind ( ethernetConnection.sock, ( struct sockaddr * )
> &ethernetConnection.local_addr, sizeof ( ethernetConnection.local_addr
) ) <
> 0 )
> (...)
>         recvfrom( ethernetConnection.sock, &recBuffer, sizeof ( recBuffer
> ),0, ( struct sockaddr * ) &ethernetConnection.dest_addr,&destlen );

Here you are using the posix calls socket, bind and recvfrom. It will
allow you to work with an RTDM socket only if you compile for the posix
skin. See:

http://www.xenomai.org/index.php/Porting_POSIX_applications_to_Xenomai#Compilation_for_the_Xenomai_POSIX_skin

Now, I have no idea why you do not get any signal when using these calls.

-- 
                                                                Gilles.

Re: [Xenomai] Context Switch with rtnet

[Gilles Chanteperdrix]

On 09/22/2012 03:20 PM, Franz Engel wrote:

>> On Fri, 21 Sep 2012 12:08:40 +0200, Kai Bollue <mlist1@bollue.de>
>>> On 21.09.2012 12:02, Gilles Chanteperdrix wrote:
>>>> On 09/21/2012 11:58 AM, Kai Bollue wrote:
>>>>> On 14.09.2012 17:01, Franz Engel wrote:
>>>>> Hi,
>>>>>
>>>>> I've a little problem. I'm using the rtnet.h-header and the
>> following function:
>>>>> ret = recvfrom ( ethernetConnection.sock, &recBuffer, sizeof (
>>>>> recBuffer ),0, ( struct sockaddr * )
>>>>> &ethernetConnection.dest_addr,&destlen );
>>>> Hi,
>>>>
>>>> stupid question: But you do use an rtnet socket, not a normal one,
>> do
>>>> you? In this case, shouldn't it be rt_dev_recvfrom(...)?
>>> When using the posix skin, recvfrom is substituted at compilation (or
>>> more exactly link-edit) time with __wrap_recvfrom, implemented in
>>> libpthread_rt.so, allowing to work with real-time sockets.
>>>
>>
>> Ah, sorry, I missed that.
>>
> 
> Sorry, I missed to answer on some questions:
> 
> Let me explain my problem again. When I run my program ("mytask") the CSW
> and MSW increase every time the program receive a message:
> CPU  PID    MSW        CSW        PF    STAT       %CPU  NAME
>   0  0      0          396        0     00500080  100.0  ROOT/0
>   1  0      0          0          0     00500080  100.0  ROOT/1
>   0  1336   1          1          0     00300380    0.0  kogmo_man
>   0  0      0          209        0     00000082    0.0  rtnet-stack
>   0  0      0          1          0     00000082    0.0  rtnet-rtpc
>   0  0      0          1          0     00000082    0.0  rtcfg-rx
>   0  5182   1          15         0     00340182    0.0  mytask
>   0  0      0          143        0     00000000    0.0  IRQ22: rteth0
>   1  0      0          186        0     00000000    0.0  IRQ22: rteth0
>   0  0      0          2054479    0     00000000    0.0  IRQ4364: [timer]
>   1  0      0          547805     0     00000000    0.0  IRQ4364: [timer]
> 
> Wolfgang advised me to find the reason for this phenomenon like the program
> on this link:
> http://www.rts.uni-hannover.de/xenomai/lxr/source/examples/native/sigdebug.c
> I tried it and modified my code (see the code further down). When I insert
> an "cout" after "rt_task_set_mode(0, T_WARNSW, NULL);" I get a failure
> message (so it should work). But when I run this program without the cout I
> get no message, but the CSW and MSW counter are still increases.  
> 
> 
> That's my versions:
> rtnet 0.9.12
> xenomai 2.5.5.2
> linux kernel 2.6.32-20
> ubuntu 10.04
> i-pipe patch: adeos-ipipe-2.6.32.20-x86-2.7-03.patch
> NIC driver: 8139too
> CPU: Intel vPro
> 
> 
> Any other ideas? Or is it not a problem but the normal state that the
> counters increases?
> 
> Here is the code of my cmake-file:
> cmake_minimum_required(VERSION 2.8)
> 
> ##################################
> # BEGIN OF EDITING THIS TEMPLATE #
> ##################################
> 
> ############################################
> # PLEASE ENTER HERE THE NAME OF YOUR PROJECT
> 
> set(ProjectName robotrsiVersuch)  #e.g. "robotdatalogger" (do not forget the
> quotes)
> 
> PROJECT(${ProjectName})
> 
> ###############################################
> # ENTER HERE THE VERSION NUMBER OF YOUR PROJECT
> set(${ProjectName}_VERSION_MAJOR 1)
> set(${ProjectName}_VERSION_MINOR 0)
> 
> ###############################################
> # ENABLE OR DIABLE THE USE OF XENOMAI AND QT
> option(USE_XENOMAI	"xenomai"		ON	)
> option(USE_RTNET	"rtnet"			ON	)
> 
> #############################################
> # NOW ADD ALL SOURCE FILES OF YOUR PROJECT
> # IF YOU'RE USING QT, ALSO ADD ALL HEADERS OF 
> # QT-OBJECTS, UI-FILES AND RESOURCES
> 
> #####################
> # INCLUDE DIRECTORIES
> 
> set(MAIN_DIR ${PROJECT_SOURCE_DIR}/../..)
> set(SOURCE_DIR src)
> 
> 
> #############################################
> # ENTER THE INSTALL DIRECTORY
> # added by stefan.thumfart@profactor.at
> set(INSTALLDIR ${MAIN_DIR}/bin)
> 
> set(INCLUDES
>     ${SOURCE_DIR}
> )
> 
> 
> ###########
> # LIBRARIES
> 
> set(EXTRA_LIBS 
> 
> )
> 
> ######
> # CPPs
> 
> set(SOURCES 
> 	main.cpp
> 	src/rsiinterface.cpp
> 	src/rsiinterface.h
> )
> 
> #############
> # MOC-Headers
> set(MOC_HEADERS
> 
> )
> 
> ##########
> # UI-Files
> set(UIS
> 
> )
> 
> ###########
> # Resources
> set(RESOURCES 
> 
> )
> 
> ##################################
> #  END OF EDITING THIS TEMPLATE  #
> ##################################
> 
> 
> message (STATUS "Usage of Xenomai: ${USE_XENOMAI}")
> message (STATUS "Usage of RTnet: ${USE_RTNET}")
> 
> message(STATUS "Generating buid files for project ${ProjectName}, version
> ${${ProjectName}_VERSION_MAJOR}.${${ProjectName}_VERSION_MINOR}")
> 
> include_directories(${INCLUDES})
> 
> 
> IF(USE_XENOMAI)
>     message(STATUS "Starting Xenomai configuration")
> 
>     message(STATUS "startign RTnet configuration")
>     include_directories(/usr/local/rtnet/include)
>     include_directories(/usr/xenomai/include/rtdm)
>     message(STATUS "RTnet configuration finished")   
> 
>     find_library(XENO_rtdm rtdm /usr/xenomai/lib)
>     set(EXTRA_LIBS ${EXTRA_LIBS} ${XENO_rtdm})
>     message(STATUS ${EXTRA_LIBS} )
> 
>     set(XENOMAI_DIR "/usr/xenomai")
>     set(XENOMAI_INCLUDE_DIR ${XENOMAI_DIR}/include)
>     set(XENOMAI_INCLUDE_POSIX_DIR ${XENOMAI_DIR}/include/native)
> 
>     exec_program("${XENOMAI_DIR}/bin/xeno-config --skin=native --ldflags"
> OUTPUT_VARIABLE XENOMAI_EXE_LINKER_FLAGS)
>     set(CMAKE_EXE_LINKER_FLAGS ${CMAKE_EXE_LINKER_FLAGS}
> ${XENOMAI_EXE_LINKER_FLAGS})
>     exec_program("${XENOMAI_DIR}/bin/xeno-config --skin=native --cflags"
> OUTPUT_VARIABLE XENOMAI_DEFINITIONS)
>     set(EXTRA_DEFINITIONS ${EXTRA_DEFINITIONS} ${XENOMAI_DEFINITIONS})
>     ADD_DEFINITIONS(${EXTRA_DEFINITIONS})
> 
>     set(EXTRA_LIBS ${EXTRA_LIBS} rtdk)
>     
>     message(STATUS ${XENOMAI_EXE_LINKER_FLAGS})
>     message(STATUS ${XENOMAI_DEFINITIONS})
>     message(STATUS "Xenomai configuration finished")
> ELSE(USE_XENOMAI)
>     find_library(RT_TIME_LIB rt)
>     set(EXTRA_LIBS ${EXTRA_LIBS} ${RT_TIME_LIB})
> ENDIF(USE_XENOMAI)
> 
> 
> IF (USE_RTNET)
> 
> ENDIF (USE_RTNET)
> 
> 
> add_executable(${ProjectName} ${SOURCES})
> 
> TARGET_LINK_LIBRARIES(${ProjectName} ${EXTRA_LIBS})
> 
> install(TARGETS ${ProjectName} DESTINATION ${INSTALLDIR})
> 
> And this is my cpp:
> #include <unistd.h>
> #include <stdlib.h>
> #include <stdio.h>
> #include <string.h>
> #include <signal.h>
> #include <getopt.h>
> #include <execinfo.h>
> #include <sys/mman.h>
> 
> #include <stdexcept>
> #include <arpa/inet.h>
> #include <native/task.h>
> #include "rtdk.h"
> #include <rtnet.h>
> 
> 
> #define RECBUFFERSIZE 600
> #define SENDBUFFERSIZE 300
> 
> RT_TASK task;
> const char *reason_str[7];
> 
> struct conn_t
> {
>     int sock;
>     struct sockaddr_in dest_addr;
>     struct sockaddr_in local_addr;
> };
> 
> struct conn_t getSocket ()
> {
>     struct conn_t ethernetConnection;
> 
>     //ethernetConnection.dest_addr.sin_family         = AF_INET;
>     ethernetConnection.dest_addr.sin_addr.s_addr    = inet_addr
> ("192.168.100.12");
>     ethernetConnection.dest_addr.sin_port           = htons (6002);
>     
>     //ethernetConnection.local_addr.sin_family        = AF_INET;
>     ethernetConnection.local_addr.sin_addr.s_addr   = inet_addr
> ("192.168.100.11");
>     ethernetConnection.local_addr.sin_port          = htons (6008);
> 
>     if ( ( ethernetConnection.sock = socket ( AF_INET, SOCK_DGRAM,
> IPPROTO_UDP ) ) < 0 )
>     {
>         throw std::runtime_error("Can not open socket");
>     }
>     else
>     {
>     }
> 
>     if ( bind ( ethernetConnection.sock, ( struct sockaddr * )
> &ethernetConnection.local_addr, sizeof ( ethernetConnection.local_addr ) ) <
> 0 )
>     {
>         close ( ethernetConnection.sock );
>         throw std::runtime_error("Can not bind socket");
>     }
>     else
>     {
>     }
> 
>     return ethernetConnection;
> }
> 
> void task_body (void *cookie)
> {  
>     char recBuffer[RECBUFFERSIZE];
> 
>     socklen_t destlen;
>     
>     struct conn_t ethernetConnection;
>     
>     ethernetConnection = getSocket();
> 
>     destlen = sizeof ( ethernetConnection.dest_addr );
>     
>     int k;
>     
>     rt_task_set_mode(0, T_WARNSW, NULL);
>     
>     for(k=0;k<1000;k++)
>     {
>         recvfrom( ethernetConnection.sock, &recBuffer, sizeof ( recBuffer
> ),0, ( struct sockaddr * ) &ethernetConnection.dest_addr,&destlen );
> 	
>         //ret = rt_dev_sendto ( ethernetConnection.sock, &sendBuffer, strlen
> ( sendBuffer ) +1, 0, ( struct sockaddr * )
> &ethernetConnection.dest_addr,destlen);
>     }
> }
> 
> 
> void warn_upon_switch(int sig, siginfo_t *si, void *context)
> {
>     unsigned int reason = si->si_value.sival_int;
>     void *bt[32];
>     int nentries;
>     
>     printf("\nSIGDEBUG received, reason %d: %s\n", reason,
>            reason <= SIGDEBUG_WATCHDOG ? reason_str[reason] : "<unknown>");
>     /* Dump a backtrace of the frame which caused the switch to
>        secondary mode: */
>     nentries = backtrace(bt,sizeof(bt) / sizeof(bt[0]));
>     backtrace_symbols_fd(bt,nentries,fileno(stdout));
> }


Beware the stack, you probably can not use printf with the default
thread stack size (this is the reason why backtrace_symbols_fd is used).

-- 
                                                                Gilles.

[Xenomai] Context Switch with rtnet

[Franz Engel]

> On 09/22/2012 03:40 PM, Gilles Chanteperdrix wrote:
> > On 09/22/2012 03:20 PM, Franz Engel wrote:
> >> On Fri, 21 Sep 2012 12:08:40 +0200, Kai Bollue <mlist1@bollue.de>
> >>> On 21.09.2012 12:02, Gilles Chanteperdrix wrote:
> >>>> On 09/21/2012 11:58 AM, Kai Bollue wrote:
> >>>>> On 14.09.2012 17:01, Franz Engel wrote:
> >>>>> Hi,
> >>>>>
> >>>>> I've a little problem. I'm using the rtnet.h-header and the
> >> following function:
> >>>>> ret = recvfrom ( ethernetConnection.sock, &recBuffer, sizeof (
> >>>>> recBuffer ),0, ( struct sockaddr * )
> >>>>> &ethernetConnection.dest_addr,&destlen );
> >>>> Hi,
> >>>>
> >>>> stupid question: But you do use an rtnet socket, not a normal one,
> >> do
> >>>> you? In this case, shouldn't it be rt_dev_recvfrom(...)?
> >>> When using the posix skin, recvfrom is substituted at compilation
> >>> (or more exactly link-edit) time with __wrap_recvfrom, implemented
> >>> in libpthread_rt.so, allowing to work with real-time sockets.
> >>>
> >>
> >> Ah, sorry, I missed that.
> >>
> >
> > Sorry, I missed to answer on some questions:
> >
> > Let me explain my problem again. When I run my program ("mytask") the
> > CSW and MSW increase every time the program receive a message:
> > CPU  PID    MSW        CSW        PF    STAT       %CPU  NAME
> >   0  0      0          396        0     00500080  100.0  ROOT/0
> >   1  0      0          0          0     00500080  100.0  ROOT/1
> >   0  1336   1          1          0     00300380    0.0  kogmo_man
> >   0  0      0          209        0     00000082    0.0  rtnet-stack
> >   0  0      0          1          0     00000082    0.0  rtnet-rtpc
> >   0  0      0          1          0     00000082    0.0  rtcfg-rx
> >   0  5182   1          15         0     00340182    0.0  mytask
> >   0  0      0          143        0     00000000    0.0  IRQ22:
> rteth0
> >   1  0      0          186        0     00000000    0.0  IRQ22:
> rteth0
> >   0  0      0          2054479    0     00000000    0.0  IRQ4364:
> [timer]
> >   1  0      0          547805     0     00000000    0.0  IRQ4364:
> [timer]
> >
> > Wolfgang advised me to find the reason for this phenomenon like the
> > program on this link:
> > http://www.rts.uni-
> hannover.de/xenomai/lxr/source/examples/native/sigd
> > ebug.c I tried it and modified my code (see the code further down).
> > When I insert an "cout" after "rt_task_set_mode(0, T_WARNSW, NULL);"
> I
> > get a failure message (so it should work). But when I run this
> program
> > without the cout I get no message, but the CSW and MSW counter are
> > still increases.
> >
> >
> > That's my versions:
> > rtnet 0.9.12
> > xenomai 2.5.5.2
> 
> 
> That is not the latest release in the 2.5 branch. You are advised to
> upgrade to 2.5.6, it should work with the same linux kernel version.
> 
> >
> 
> >     exec_program("${XENOMAI_DIR}/bin/xeno-config --skin=native --
> ldflags"
> > OUTPUT_VARIABLE XENOMAI_EXE_LINKER_FLAGS)
> >     set(CMAKE_EXE_LINKER_FLAGS ${CMAKE_EXE_LINKER_FLAGS}
> > ${XENOMAI_EXE_LINKER_FLAGS})
> >     exec_program("${XENOMAI_DIR}/bin/xeno-config --skin=native --
> cflags"
> > OUTPUT_VARIABLE XENOMAI_DEFINITIONS)
> 
> 
> Here you are retrieving the flags for the native skin, not for the
> posix skin, unless there is some other part of your makefile which gets
> them.

Oh, sorry. I mixed two of my demo projects. In the correct cmake file I use the posix skin. In my second project I use the native skin and the rt_dev_* commands. But it has the same result.

> 
> >     if ( ( ethernetConnection.sock = socket ( AF_INET, SOCK_DGRAM,
> > IPPROTO_UDP ) ) < 0 )
> 
> > (...)
> >     if ( bind ( ethernetConnection.sock, ( struct sockaddr * )
> > &ethernetConnection.local_addr, sizeof (
> ethernetConnection.local_addr
> ) ) <
> > 0 )
> > (...)
> >         recvfrom( ethernetConnection.sock, &recBuffer, sizeof (
> > recBuffer ),0, ( struct sockaddr * )
> > &ethernetConnection.dest_addr,&destlen );
> 
> Here you are using the posix calls socket, bind and recvfrom. It will
> allow you to work with an RTDM socket only if you compile for the posix
> skin. See:
> 
> http://www.xenomai.org/index.php/Porting_POSIX_applications_to_Xenomai#
> Compilation_for_the_Xenomai_POSIX_skin
> 
> Now, I have no idea why you do not get any signal when using these
> calls.

Here is my "good" cmake file:
cmake_minimum_required(VERSION 2.8)

##################################
# BEGIN OF EDITING THIS TEMPLATE #
##################################

############################################
# PLEASE ENTER HERE THE NAME OF YOUR PROJECT

set(ProjectName robotrsiVersuch)  #e.g. "robotdatalogger" (do not forget the quotes)

PROJECT(${ProjectName})

###############################################
# ENTER HERE THE VERSION NUMBER OF YOUR PROJECT
set(${ProjectName}_VERSION_MAJOR 1)
set(${ProjectName}_VERSION_MINOR 0)

###############################################
# ENABLE OR DIABLE THE USE OF XENOMAI AND QT
option(USE_XENOMAI	"xenomai"		ON	)
option(USE_RTNET	"rtnet"			ON	)

#############################################
# NOW ADD ALL SOURCE FILES OF YOUR PROJECT
# IF YOU'RE USING QT, ALSO ADD ALL HEADERS OF 
# QT-OBJECTS, UI-FILES AND RESOURCES

#####################
# INCLUDE DIRECTORIES

set(MAIN_DIR ${PROJECT_SOURCE_DIR}/../..)
set(SOURCE_DIR src)


#############################################
# ENTER THE INSTALL DIRECTORY
set(INSTALLDIR ${MAIN_DIR}/bin)

set(INCLUDES
    ${SOURCE_DIR}
)


###########
# LIBRARIES

set(EXTRA_LIBS 

)

######
# CPPs

set(SOURCES 
	main.cpp
	src/rsiinterface.cpp
	src/rsiinterface.h
)

#############
# MOC-Headers
set(MOC_HEADERS

)

##########
# UI-Files
set(UIS

)

###########
# Resources
set(RESOURCES 

)

##################################
#  END OF EDITING THIS TEMPLATE  #
##################################


message (STATUS "Usage of Xenomai: ${USE_XENOMAI}")
message (STATUS "Usage of RTnet: ${USE_RTNET}")

message(STATUS "Generating buid files for project ${ProjectName}, version ${${ProjectName}_VERSION_MAJOR}.${${ProjectName}_VERSION_MINOR}")

include_directories(${INCLUDES})


IF(USE_XENOMAI)
    message(STATUS "Starting Xenomai configuration")

    message(STATUS "startign RTnet configuration")
    include_directories(/usr/local/rtnet/include)
    include_directories(/usr/xenomai/include/rtdm)
    message(STATUS "RTnet configuration finished")   

    find_library(XENO_rtdm rtdm /usr/xenomai/lib)
    set(EXTRA_LIBS ${EXTRA_LIBS} ${XENO_rtdm})
    message(STATUS ${EXTRA_LIBS} )

    set(XENOMAI_DIR "/usr/xenomai")
    set(XENOMAI_INCLUDE_DIR ${XENOMAI_DIR}/include)
    set(XENOMAI_INCLUDE_POSIX_DIR ${XENOMAI_DIR}/include/posix)

    exec_program("${XENOMAI_DIR}/bin/xeno-config --skin=posix --ldflags" OUTPUT_VARIABLE XENOMAI_EXE_LINKER_FLAGS)
    set(CMAKE_EXE_LINKER_FLAGS ${CMAKE_EXE_LINKER_FLAGS} ${XENOMAI_EXE_LINKER_FLAGS})
    exec_program("${XENOMAI_DIR}/bin/xeno-config --skin=posix --cflags" OUTPUT_VARIABLE XENOMAI_DEFINITIONS)
    set(EXTRA_DEFINITIONS ${EXTRA_DEFINITIONS} ${XENOMAI_DEFINITIONS})
    ADD_DEFINITIONS(${EXTRA_DEFINITIONS})

    set(EXTRA_LIBS ${EXTRA_LIBS} rtdk)
    
    message(STATUS ${XENOMAI_EXE_LINKER_FLAGS})
    message(STATUS ${XENOMAI_DEFINITIONS})
    message(STATUS "Xenomai configuration finished")
ELSE(USE_XENOMAI)
    find_library(RT_TIME_LIB rt)
    set(EXTRA_LIBS ${EXTRA_LIBS} ${RT_TIME_LIB})
ENDIF(USE_XENOMAI)


IF (USE_RTNET)

ENDIF (USE_RTNET)


add_executable(${ProjectName} ${SOURCES})

TARGET_LINK_LIBRARIES(${ProjectName} ${EXTRA_LIBS})

install(TARGETS ${ProjectName} DESTINATION ${INSTALLDIR})

Re: [Xenomai] Context Switch with rtnet

[Gilles Chanteperdrix]

On 09/22/2012 03:55 PM, Franz Engel wrote:

>> On 09/22/2012 03:40 PM, Gilles Chanteperdrix wrote:
>>> On 09/22/2012 03:20 PM, Franz Engel wrote:
>>>> On Fri, 21 Sep 2012 12:08:40 +0200, Kai Bollue <mlist1@bollue.de>
>>>>> On 21.09.2012 12:02, Gilles Chanteperdrix wrote:
>>>>>> On 09/21/2012 11:58 AM, Kai Bollue wrote:
>>>>>>> On 14.09.2012 17:01, Franz Engel wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> I've a little problem. I'm using the rtnet.h-header and the
>>>> following function:
>>>>>>> ret = recvfrom ( ethernetConnection.sock, &recBuffer, sizeof (
>>>>>>> recBuffer ),0, ( struct sockaddr * )
>>>>>>> &ethernetConnection.dest_addr,&destlen );
>>>>>> Hi,
>>>>>>
>>>>>> stupid question: But you do use an rtnet socket, not a normal one,
>>>> do
>>>>>> you? In this case, shouldn't it be rt_dev_recvfrom(...)?
>>>>> When using the posix skin, recvfrom is substituted at compilation
>>>>> (or more exactly link-edit) time with __wrap_recvfrom, implemented
>>>>> in libpthread_rt.so, allowing to work with real-time sockets.
>>>>>
>>>>
>>>> Ah, sorry, I missed that.
>>>>
>>>
>>> Sorry, I missed to answer on some questions:
>>>
>>> Let me explain my problem again. When I run my program ("mytask") the
>>> CSW and MSW increase every time the program receive a message:
>>> CPU  PID    MSW        CSW        PF    STAT       %CPU  NAME
>>>   0  0      0          396        0     00500080  100.0  ROOT/0
>>>   1  0      0          0          0     00500080  100.0  ROOT/1
>>>   0  1336   1          1          0     00300380    0.0  kogmo_man
>>>   0  0      0          209        0     00000082    0.0  rtnet-stack
>>>   0  0      0          1          0     00000082    0.0  rtnet-rtpc
>>>   0  0      0          1          0     00000082    0.0  rtcfg-rx
>>>   0  5182   1          15         0     00340182    0.0  mytask


The thing is, we only see 1 MSW for mytask, I think it is normal to have
an MSW during task startup. CSW are completely normal for a recvfrom loop.

-- 
                                                                Gilles.