Date: Wed, 30 Jul 2003 16:30:44 -0400 From: "Lapinski, Michael (Research)" <lapinski@crd.ge.com> To: "'Haesu'" <haesu@towardex.com>, Dave Byrne <dave-lists@glynjones.com>, freebsd-isp@freebsd.org, freebsd-questions@freebsd.org Subject: RE: sendto: No buffer space available Message-ID: <E4AAC34FE3CF564D8AE89EB8AC333FD7091651FA@xmb03crdge.crd.ge.com>
next in thread | raw e-mail | index | archive | help
I think heading down the path of switchign out network
gear is a bad idea, this is definitely something in
the software.
I have had this error a few times when messing with
the TCP window sizes, net.inet.tcp.sendspace and
net.inet.tcp.recvspace. When I set them to something
over 128000 I would get the error. The solution was
set the number of mbufs to 128000. Doing so allowed
me to make the window sizes 256000 and eliminated
the error.
-good luck
-mtl
--------------------------------------------------
Michael Lapinski
Computer Scientist
GE Research
"I think there is a world market for maybe five computers."
- IBM Chairman Thomas Watson, 1943
->-----Original Message-----
->From: Haesu [mailto:haesu@towardex.com]
->Sent: Wednesday, July 30, 2003 4:26 PM
->To: Dave Byrne; freebsd-isp@freebsd.org; freebsd-questions@freebsd.org
->Subject: Re: sendto: No buffer space available
->
->
->Hmmm... i had truss running but the moment it died it was
->running gettimeoftheday() so i am not sure :-/
->
->I tried different ports on the switch.. It's a cisco switch
->btw, and other freebsd boxes on that switch
->are not exhibiting similar problem....
->
->I'll try putting this behind a hub or something other than
->cisco just for kicks but if anyone has any further
->ideas/suggestions, i'd really appreciate it.
->
->Thank you!
->
->-hc
->
->--
->Sincerely,
-> Haesu C.
-> TowardEX Technologies, Inc.
-> WWW: http://www.towardex.com
-> E-mail: haesu@towardex.com
-> Cell: (978) 394-2867
->
->On Wed, Jul 30, 2003 at 11:36:35AM -0700, Dave Byrne wrote:
->> I had the same exact problem. I traced it to be a bug in
->some software
->> that opened a domain socket(2) but could not connect(2) and
->never closed
->> the descriptor returned.
->>
->> something like:
->>
->> sd = socket(AF_UNIX, SOCK_STREAM, 0);
->> ...
->> if(connect(sd, (struct sockaddr *)&saddr, sizeof(saddr)) < 0) {
->> return -1;
->> }
->>
->> where close(2) was skipped on sd if the connect failed.
->>
->> over a period of time (12-18hrs), these unconnected sockets
->would fill
->> up the available buffer space, and exhibit the same symptoms you are
->> having. even running ifconfig would fail with No buffer
->space available.
->>
->> from intro(2):
->> 55 ENOBUFS No buffer space available. An operation on a
->socket or pipe
->> was not performed because the system lacked sufficient
->buffer space or
->> because a queue was full.
->>
->> Fixing that bug fixed the problem. I doubt you have a
->hardware problem,
->> I would try narrowing down what software is causing the
->lockup. truss(1)
->> might help you out here.
->>
->>
->>
->> Dave
->>
->>
->>
->>
->> On Wed, 2003-07-30 at 09:32, Haesu wrote:
->> > Hello,
->> >
->> > We have a FreeBSD box here that we use to route some GRE
->tunnels and ipv6 gif tunnels. We use zebra for dynamic
->routing running zebra, bgpd, ospfd, and ospf6d.
->> >
->> > We have about 12 FreeBSD boxes with exact same
->configuration, the only
->> > difference is just the IP address of each interface.
->> >
->> > None of them fail but this one box...
->> >
->> > Everyday, this box stops all networking. I can still
->console in and stuff.. When
->> > I typed 'ping 127.0.0.1' at the console after networking
->locked up, it says:
->> > ping: sendto: No buffer space available
->> >
->> > The only solution seems to be rebooting it everyday... It
->happens every 12 hours
->> > or so...
->> >
->> > This is not related with mbuf, etc either, as netstat -m
->doesn't show any
->> > issues.
->> >
->> > The box has one IP address and IPv6 address in addition
->to 127.0.0.1 on lo0
->> > interface. It also has a ds0 interface with 10.5.5.5/30
->assigned to ds0.
->> > This is exact same configuration on all other boxes, and
->none of them fail but
->> > this one.
->> >
->> > I've swapped out NICs with different vendors 3 times
->(tried, xl, dc, and now rl)
->> >
->> > I've also swapped out the whole box, and also swapped out
->the whole hard drive
->> > and did full reinstall. And problem still persists and
->it's definately not
->> > hardware as I swapped everything out... (unless the 3 NIC
->vendors above are all
->> > exhibiting same issue)
->> >
->> > I tried to look on Google but nothing useful that
->corelates to this particular
->> > issue..
->> >
->> > Any help would be very appreciated :)
->> >
->> > Thanks,
->> > -hc
->> >
->> > The box is running FreeBSD 4.8-STABLE
->> > FreeBSD necsis 4.8-STABLE FreeBSD 4.8-STABLE #0: Tue Jul
->29 13:10:11 GMT 2003 root@necsis:/usr/src/sys/compile/router i386
->> >
->> > Following is output of netstat -s AFTER the networking
->locks up with no buffer
->> > space available error:
->>
->>
->> _______________________________________________
->> freebsd-isp@freebsd.org mailing list
->> http://lists.freebsd.org/mailman/listinfo/freebsd-isp
->> To unsubscribe, send any mail to
->"freebsd-isp-unsubscribe@freebsd.org"
->
->_______________________________________________
->freebsd-isp@freebsd.org mailing list
->http://lists.freebsd.org/mailman/listinfo/freebsd-isp
->To unsubscribe, send any mail to "freebsd-isp-unsubscribe@freebsd.org"
->
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?E4AAC34FE3CF564D8AE89EB8AC333FD7091651FA>