|
Sean Boudreau(deleted)
|
Re: io-net - problem every 20mins
|
|
Sean Boudreau(deleted)
09/03/2008 10:16 PM
post12795
|
Re: io-net - problem every 20mins
Yes, probably an arp timeout which happens at 20 minutes.
As Xiaodan mentioned setting it permanent would be a good
test. Should also be fixed in io-pkt which will try to
refresh the entry before a full deletion.
-seanb
On Wed, Sep 03, 2008 at 08:13:35PM -0400, Xiaodan Tang wrote:
> I have same situation in the pass, where every 20 minutes or so, some
> UDP packets get dropped.
>
> Turns out the ARP cache is timed out. And the stack is trying to ARP the
> destination again. Using one of the "arp -s" option to set a permanent
> ARP entry solve the problem.
>
> No idea of the stack blown up though, whatelse is running inside the
> io-net?
>
> -xtang
>
> -----Original Message-----
> From: Dave Bott [mailto:community-noreply@qnx.com]
> Sent: September 3, 2008 8:09 PM
> To: general-networking
> Subject: io-net - problem every 20mins
>
> A customer is using io-net on 6.3.2 w. a speedo chip on an 850MHz X86.
>
> all traffic is UDP - trying to wring every drop out - currently around
> 40Mps one way, 20Mbps the other. The CPU is pretty much maxed out.
>
> They drop a couple of packets every 40 minutes, pretty much to the
> second.
> They want to know if anything in the stack has a timeout of 2400s ?
>
> I see ntp at 1200s, but nothing specifically at 2400 - any ideas ?
> They wonder if it could be arp table updates and, if so, how they can
> disable them...
>
> I think that if they're this maxed out, then it could be almost
> anything...
>
> They've also seen this message occasionally:
> tcpip: blown stack handling 0x102. See "stacksize" option
>
> Any ideas ?
>
> Thanks !
>
> Dave
>
> _______________________________________________
> General
> http://community.qnx.com/sf/go/post12792
>
>
> _______________________________________________
> General
> http://community.qnx.com/sf/go/post12793
>
|
|
|
|
|
