Hi everybody,

I have ported the server application to a "pure" Linux application (enclosed), 
i.e. nothing with rtnet.
I just wanted to see if the eepro100 driver of "standard" Linux shows the same 
effect.
And actually it does!
My setup is the following:
I have two 3.2 GHz Pentium 4 PCs, both having a eepro100 as PCI board.
The eepro100-boards are connected via a crossover cable.
On client side, I ran the rtnet kernel module, on server side, I checked the 
standard
linux application. I tried with a RTAI-3.3 patched kernel 2.4.33 (UP) (RTAI not 
loaded),
and with an adeos patched 2.6.17.7 (SMP), xenomai not active.
It takes longer to get the error (typically more than 500000 cycles), however 
the effect is the same.
This shows me, that there seems to be actually an issued with the eepro100 or 
the linux driver approach.
I am going to do the same test with the Intel e100 driver. 

Regards

Mathias

> M. Koehrer wrote:
> > Hi everbody,
> > 
> > I have made a test that allows to reproduce the issue (enclosed to this
> mail).
> > You need two PCs that run the rtnet software.
> > The test is written for RTAI-3.3cv, however it should be easy to move it
> > to xenomai.
> > 
> > One of the PCs acts as server. This one should have the eepro100 board, as
> this is the candidate that
> > shows the buggy behaviour.
> > The other PC acts as client. This one can have any rtnet-support NIC (I
> use also a eepro100 for it).
> > Both PCs are connected via an cross-over cable.
> > The test sequence is the following:
> >   Client         Server
> >     ---- 1472 ---->
> >     ----   32 ---->
> >     <---- 800 -----
> > The client sends two UDP messages, one with 1472 the second with 32 byte
> data.
> > The server confirms the reception of both messages by a 800 byte
> response.
> > This test will be repeated permanently. Whenever the reception of the
> > second message fails, it will be printed out (dmesg to see the results).
> > On my system, it takes about 280 microseconds for one cycle.
> > 
> > The test is realized in two kernel modules, to ease startup I have also
> written
> > two scripts.
> > Please read the readme.txt in the archive.
> > 
> > The shorter the second frame is, the more often the critical situation
> appears.
> > It takes only a few seconds on my system to get the error (especially if
> some
> > other process is doing lot of harddisk/ethernet action on the server PC).
> > 
> > Perhaps there is anybody out there that could run this test.
> > Interesting could be, if it behaves the same under xenomai.
> > 
> 
> I ported it over (Native user-space) and gave it a first try. Archive is
> attached. So far this was without "success", i.e. everything works as it
> should (almost everything, but that's an unrelated story).
> 
> Test scenario here:
> 
> Pentium 266 <-> Pentium 700 (tried both roles for each node)
> Kernel 2.6.17.13, i-pipe-1.5-00, Xenomai trunk, RTnet 0.9.6
> 
> Load on the server: "latency" test + "find /" (CF-card as disk, rather
> low IRQ load).
> 
> I have no 2.4 system at hand (more precisely: no working libc), so I
> cannot test in this direction. Another question is if the system speed
> (specifically of the 266 MHz box) has any influence. When time permits I
> will try on some other boxes as well, but that may take a few days.


-- 
Mathias Koehrer
[EMAIL PROTECTED]


Viel oder wenig? Schnell oder langsam? Unbegrenzt surfen + telefonieren
ohne Zeit- und Volumenbegrenzung? DAS TOP ANGEBOT JETZT bei Arcor: günstig
und schnell mit DSL - das All-Inclusive-Paket für clevere Doppel-Sparer,
nur  44,85 €  inkl. DSL- und ISDN-Grundgebühr!
http://www.arcor.de/rd/emf-dsl-2

Attachment: netrecvtest.c
Description: Binary data

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
RTnet-users mailing list
RTnet-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/rtnet-users

Reply via email to