Synopsis: process swi1:net gives 100% CPU usage.
Sergey Pronin
sepron at gmail.com
Wed Feb 18 04:45:43 PST 2009
Good day.
As I'm using mpd4 I have mpd.conf and mpd.links. There are lots of static
ngs created. So I'll post just a pattern of my mpd.conf:
pppoe_def:
set bundle disable multilink
set bundle disable round-robin
set bundle enable noretry
set ipcp dns X.X.X.X Y.Y.Y.Y
set iface disable on-demand
set iface disable proxy-arp
set iface enable tcpmssfix
set link keep-alive 3 12
# set link mtu 1300
# set link mru 1460
set link max-redial -1
set pppoe service "*"
set pppoe enable incoming
set link enable chap
set link enable chap-md5
set link enable chap-msv1
set link enable chap-msv2
set link disable passive callback no-orig-auth check-magic
log -chat -bund -fsm -pptp
set iface up-script "/root/scripts/up-script.pl"
set iface down-script "/root/scripts/down-script.pl"
set ipcp disable vjcomp
set auth disable internal
set auth enable radius-auth
set radius server Z.Z.Z.Z password 1812 1813
set radius timeout 10
set radius retries 3
#ngctl types
There are 27 total types:
Type name Number of living nodes
--------- ----------------------
pppoe 144
socket 2222
iface 2077
hole 0
gif_demux 0
gif 0
frame_relay 0
ether 560
echo 0
cisco 0
vjc 3
tty 0
tee 0
tcpmss 233
bridge 0
rfc1490 0
pptpgre 0
bpf 0
ppp 2076
one2many 0
async 0
mppc 0
lmi 0
UI 0
l2tp 0
ksocket 0
ip_input 0
I've tried to compile the kernel with and without netgraph options. I
haven't seen any benefit. swi1:net still uses 100% CPU.
I have the same problem with netgraph on 7.1-RELEASE, mpd 5.2, AMD Phenom.
> Dummynet, divert and pf
> are disabled in the kernel.
>
> While swi1:net (aka netgraph) is using 100% of CPU time m_tag_locate() and
> ng_iface_output()
> are constantly called from the infinite loop.
>
> Sergey, could you provide the content of mpd.conf and the output of "ngctl
> types".
>
More information about the freebsd-net
mailing list