kern/132984: [netgraph] swi1: net 100% cpu usage

[Vladimir Kobal]

>Number:         132984
>Category:       kern
>Synopsis:       [netgraph] swi1: net 100% cpu usage
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    freebsd-bugs
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Mon Mar 23 20:50:03 UTC 2009
>Closed-Date:
>Last-Modified:
>Originator:     Vladimir Kobal
>Release:        7.1-STABLE
>Organization:
ProKK SE
>Environment:
FreeBSD test.prokk.net 7.1-STABLE FreeBSD 7.1-STABLE #0: Sat Feb 28 13:19:11 EET 2009     root at test.prokk.net:/usr/src/sys/amd64/compile/TEST  amd64
>Description:
The system is a NAS: mpd 5.2 (PPTP), SMP, em (pci-e) interfaces. Dummynet, divert and pf are disabled in the kernel. NAT and shaping aren't used.

Periodically (from 2 minutes up to 2 weeks) the system hangs with swi1: net 100% cpu usage.

While swi1: net (netgraph) is using 100% of CPU time m_tag_locate() and ng_iface_output() are constantly called from the infinite loop according to the hwpmc stats.

The problem's appearance rate doesn't depend on cpu/users/traffic load.

Related to http://lists.freebsd.org/pipermail/freebsd-net/2009-February/021120.html
>How-To-Repeat:

>Fix:


>Release-Note:
>Audit-Trail:
>Unformatted: