This is the mail archive of the
ecos-discuss@sources.redhat.com
mailing list for the eCos project.
Re: "eth_recv out of MBUFs" of FreeBSD
On Thursday 26 September 2002 10:08 pm, Jimmy Lin \(林俊明\) wrote:
> Hi, All,
>
> I tried to evaluate the stability of network function of our
> MIPS R3000 series system so I pinged our system with heavy
> traffic in Linux by:
> ping -i 0.1 -s 35000 172.20.15.6
> I had networking buffer of 512Kbytes(524288).
>
> I found if I use the FreeBSD, I will get "eth_recv out of MBUFs"
> within 1 min! I looked into the source code and I found it actually
> run out of clusters. And after it run out of clusters, the clusters have
> gone forever and never recoverd. So I lost the networking forever.
>
> But if I use the OpenBSD, everything is well. Although there will
> be some packets lost due to the power of our CPU, I still can get
> ping reply without any problem.
>
> Did anyone encounter this kind of problem before? Or can anyone
> point out where I made mistakes? Thanks!
>
> Regards,
> Jimmy
I have found with my build that if I start up 50 pings such as:
ping -s 6000 192.268.0.100
That I loose connectivity with GDB and for all practical purposes, it appears
the eCos box has crashed. I'm using eCos 2.x something in a i386. I've not
updated to the latest sources yet though either. I know similar problems
have been discussed before and patches applied to address it.
-Rich
--
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss