Subject: misc/17006: no buffer space available (kernel misconfiguration?)
To: None <gnats-bugs@gnats.netbsd.org>
From: None <ks@ks.ping-viini.org>
List: netbsd-bugs
Date: 05/24/2002 22:18:58
>Number:         17006
>Category:       misc
>Synopsis:       network dies with rtk NIC
>Confidential:   no
>Severity:       critical
>Priority:       high
>Responsible:    misc-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Fri May 24 12:20:00 PDT 2002
>Closed-Date:
>Last-Modified:
>Originator:     Karo Salminen
>Release:        NetBSD 1.5.3_RC2
>Organization:
	
>Environment:
	
	
System: NetBSD experiment.kortex.jyu.fi 1.5.3_RC2 NetBSD 1.5.3_RC2 (experiment) #0: Thu May 23 15:35:50 EEST 2002 ks@experiment.kortex.jyu.fi:/sources/src/sys/arch/i386/compile/experiment i386
Architecture: i386
Machine: i386
>Description:

Sorry, but the description of this problem is very vague, but I have been struggling with this many weeks asking from #NetBSD IRC channel and from NetBSD mailinglists and still haven't get any precise description...

I'm using rtk NIC. My network works fine but sometimes (not that
frequently) it goes down and I can't connect it to it. SSH says "no route to host" when I try to connect it from the
outside. Then I'll try to look what's wrong locally. If I ping some IP, all I get is this:
ping: sendto: no buffer space available 

Netstat -m says: 
455 mbufs in use:
        359 mbufs allocated to data
        96 mbufs allocated to packet headers
76/86 mapped pages in use
292 Kbytes allocated to network (91% in use)
0 requests for memory denied
0 requests for memory delayed
0 calls to protocol drain routines 

netstat -id doesn't report anything odd... (I didn't save it because it seemed to be normal, sorry... my loss?).

If I do 'ifconfig rtk0 down; ifconfig rtk0 up" then it's back online. This usually does not happen frequently _except_
it always happens when the backup script is sending backup packages to remote host. The script
can be found at http://ks.ping-viini.org:81/~ks/backup2.sh. This does not happen during _heavy_ traffic, which
is very odd. Last time I mailed about this one proposed solution was to increase NMBCLUSTERS. I increased it to 8192,
but it didn't help. Some other posts was about throwing the NIC away, but I'm pretty sure it's not because
of the NIC, many people have no problems with it. I doubt the configs... because the problem did not appear on
Linux.
My kernel config can be found at http://ks.ping-viini.org:81/~ks/experiment. 

The previous mail about the issue can be found at http://mail-index.netbsd.org/netbsd-help/2002/05/15/0006.html.
>How-To-Repeat:
	
>Fix:
	
>Release-Note:
>Audit-Trail:
>Unformatted: