ttcp promblem in my 68360 board
Eric Norum
eric at cls.usask.ca
Mon Feb 28 20:53:16 UTC 2000
"Park, Chan Taek" wrote:
>
> Hello, rtems-users!
>
> Eric Norum wrote:
>
> > "Park, Chan Taek" wrote:
> > >
> > > Hello, rtems-list users!
> > >
> > > I'm now benchmarking the ttcp for my 68360 board's ethernet performance.
> > >
> > > There is no problem when I set the buflen to 8K bytes(this is default in
> > > the ttcp source code) or more than 512 bytes.
> > > In this case the ttcp transmits 16M bytes successfully from my target
> > > board to PC(winnt & cygwin) and vice versa.
> > > And when the ttcp finished it prints out statistics information about
> > > TCP throughput.
> > > But when I set the buflen to small such as 100 bytes with -l option, the
> > > ttcp never completed.
> > > It was blocked forever by the write function after some number of calls
> > > to the write function.
> > >
> > > My board configurations and host environments are that:
> > >
> > > Target
> > > MC68EN360 33MHz Rev.C
> > > 16MB DRAM
> > > SCC1 for ethernet(10Mbps)
> > > SMC1 for console(38.4kbps)
> > >
> > > Host
> > > Windows NT 4.0
> > > Cygwin32 B20.1
> > > egcs-1.1b.tar.bz2
> > > binutils-2.9.1.tar.gz
> > > newlib-1.8.0.tar.gz
> > > egcs-1.1b-rtems-diff-19981028.gz
> > > binutils-2.9.1-rtems-diff-19981027.gz
> > > newlib-1.8.0-rtems-diff-19981027.gz
> > > rtems-4.0.0.tgz
> > >
> > > Please help me for that problem.
> >
> > 1) There have been lots of changes made since 4.0.0. Perhaps that will
> > fix your problem.
>
> First, I'm sorry for my bad english!
> There is same problem in the current version of rtems, 20000118a.
> I'm working within Redhat 6.0.
>
This has been fixed for the 68360 in the 4.5.0 beta release.
The problem is present in all network drivers which use scatter/gather
DMA. As far as I know, the problem has been fixed only in the 68360
driver.
Joel has a list of drivers which need to be fixed.
--
Eric Norum eric at cls.usask.ca
Canadian Light Source Phone: (306) 966-6308
University of Saskatchewan FAX: (306) 966-6058
Saskatoon, Canada.
More information about the users
mailing list