Subject: Guide for tuning network buffers? (2.0.2_STABLE)
To: NetBSD User's Discussion List <netbsd-users@netbsd.org>
From: Peter Eisch <peter@boku.net>
List: netbsd-users
Date: 02/23/2006 14:44:52
> This message is in MIME format. Since your mail reader does not understand
this format, some or all of this message may not be legible.
--B_3223550692_46383370
Content-type: text/plain; charset="ISO-8859-1"
Content-transfer-encoding: quoted-printable
I've been running stock GENERIC values for my buffers on firewall-like
implementations. I'm now seeing packets get logged as blocked by ipfilter
even when their NAT entry is shown in 'ipnat -l' and the state for the
session is good and otherwise happy as shown by ipfstat. In many cases it
gets the 3-way connect and then the first push from the outside gets blocke=
d
(even logged as blocked) for no good reason.
Nothing in dmesg or .../messages is showing any bit of drama. The best I
can figure is that I'm running short of ideas. I don't have KMEMSTATS
enabled in the kernel, but my stats otherwise show no failures:
viper# vmstat -m
vmstat: Kmem statistics are not being gathered by the kernel.
Memory resource pool statistics
Name Size Requests Fail Releases Pgreq Pgrel Npage Hiwat Minpg Maxpg
Idle
phpool 40 406 0 0 5 0 5 5 0 inf
0
pcgpool 140 25 0 0 1 0 1 1 0 inf
0
pmappl 76 51638 0 51588 2 0 2 2 0 inf
1
pdppl 4096 68 0 14 54 0 54 54 0 inf
0
vmsppl 220 51638 0 51588 3 0 3 3 0 inf
0
vmmpepl 88 2369173 0 2367913 30 0 30 30 0 inf
2
vmmpekpl 88 120092 0 120087 1 0 1 1 0 inf
0
aobjpl 52 1 0 0 1 0 1 1 0 inf
0
amappl 40 581514 0 581010 6 0 6 6 0 inf
0
bufpl 120 429136 0 426155 94 0 94 94 0 inf
3
buf1k 1024 2128 0 0 34 0 34 34 1 1
0
buf2k 2048 53 0 1 2 0 2 2 1 1
0
buf4k 4096 27 0 0 2 0 2 2 1 1
0
buf8k 8192 774 0 0 97 0 97 97 1 1
0
mbpl 256 527 0 0 34 0 34 34 1 inf
1
mclpl 2048 332 0 0 170 0 170 170 4 1024
4
kqueuepl 260 4 0 1 1 0 1 1 0 inf
0
knotepl 72 8 0 1 1 0 1 1 0 inf
0
sockpl 200 286837 0 286676 9 0 9 9 0 inf
0
ttypl 264 71 0 0 5 0 5 5 0 inf
0
rndsample 528 10198 0 10192 1 0 1 1 0 inf
0
procpl 408 51654 0 51594 7 0 7 7 0 inf
1
lwppl 132 51682 0 51618 3 0 3 3 0 inf
0
pgrppl 16 8134 0 8092 1 0 1 1 0 inf
0
pcredpl 24 51654 0 51594 1 0 1 1 0 inf
0
plimitpl 156 221 0 208 1 0 1 1 0 inf
0
pstatspl 224 51654 0 51594 4 0 4 4 0 inf
0
rusgepl 72 51594 0 51594 1 0 1 1 0 inf
1
raspl 20 12 0 11 1 0 1 1 0 inf
0
sadatapl 40 8 0 7 1 0 1 1 0 inf
0
saupcpl 1608 9326719 0 9326719 2 0 2 2 0 inf
2
sastackpl 24 128 0 112 1 0 1 1 0 inf
0
savppl 52 8 0 7 1 0 1 1 0 inf
0
ptimerpl 124 6551 0 6526 1 0 1 1 0 inf
0
filepl 56 1202155 0 1201900 4 0 4 4 0 inf
0
cwdipl 12 51644 0 51594 1 0 1 1 0 inf
0
fdescpl 280 51644 0 51594 4 0 4 4 0 inf
0
pnbufpl 1024 9 0 0 3 0 3 3 0 inf
0
vnodepl 164 30304 0 914 1226 0 1226 1226 0 inf
0
ncachepl 84 30791 0 1401 613 0 613 613 0 inf
0
ffsinopl 176 30289 0 906 1279 0 1279 1279 0 inf
1
dino1pl 128 30289 0 906 949 0 949 949 0 inf
0
wdcspl 52 652382 0 652382 1 0 1 1 0 inf
1
extent 20 271 0 254 1 0 1 1 0 inf
0
rtentpl 132 2349 0 2208 6 0 6 6 0 inf
1
inpcbpl 108 276504 0 276457 2 0 2 2 0 inf
0
in6pcbpl 156 87 0 69 1 0 1 1 0 inf
0
inmltpl 32 11 0 1 1 0 1 1 0 inf
0
ipqepl 44 5 0 5 1 0 1 1 0 inf
1
tcpcbpl 316 16591 0 16564 3 0 3 3 0 inf
0
tcpipqepl 44 6312 0 6312 1 0 1 1 0 inf
1
synpl 184 6355 0 6355 1 0 1 1 0 inf
1
sigapl 2052 51644 0 51594 54 0 54 54 0 inf
4
siginfo 128 260727 0 260727 1 0 1 1 0 inf
1
ksiginfo 44 8 0 7 1 0 1 1 0 inf
0
pipepl 140 84874 0 84852 2 0 2 2 0 inf
1
In use 26178K, total allocated 27032K; utilization 96.8%
Any insight or suggestions would be appreciated. I=B9m truly sorry about the
html formatting, but the stats wrapping seemed to not serve my purposes.
Thanks,
Peter
--B_3223550692_46383370
Content-type: text/html; charset="ISO-8859-1"
Content-transfer-encoding: quoted-printable
<HTML>
<HEAD>
<TITLE>Guide for tuning network buffers? (2.0.2_STABLE)</TITLE>
</HEAD>
<BODY>
<FONT FACE=3D"Verdana"><SPAN STYLE=3D'font-size:12.0px'><BR>
I've been running stock GENERIC values for my buffers on firewall-like impl=
ementations. I'm now seeing packets get logged as blocked by ipfilter =
even when their NAT entry is shown in 'ipnat -l' and the state for the sessi=
on is good and otherwise happy as shown by ipfstat. In many cases it g=
ets the 3-way connect and then the first push from the outside gets blocked =
(even logged as blocked) for no good reason.<BR>
<BR>
Nothing in dmesg or .../messages is showing any bit of drama. The bes=
t I can figure is that I'm running short of ideas. I don't have KMEMST=
ATS enabled in the kernel, but my stats otherwise show no failures:<BR>
<BR>
</SPAN></FONT><SPAN STYLE=3D'font-size:12.0px'><FONT FACE=3D"Courier">viper# vm=
stat -m<BR>
vmstat: Kmem statistics are not being gathered by the kernel.<BR>
Memory resource pool statistics<BR>
Name Size Requests Fail Releases =
Pgreq Pgrel Npage Hiwat Minpg Maxpg Idle<BR>
phpool 40  =
; 406 0 0 &=
nbsp; 5 0 5=
5 0 inf &=
nbsp; 0<BR>
pcgpool 140 &nb=
sp;25 0 0 =
1 0 1 &nbs=
p; 1 0 inf =
0<BR>
pmappl 76 51638=
0 51588 2  =
; 0 2 2 &nb=
sp; 0 inf 1<BR>
pdppl 4096 &nbs=
p; 68 0 14 =
54 0 54 &n=
bsp;54 0 inf 0<BR>
vmsppl 220 51638 &nbs=
p; 0 51588 3 &nbs=
p; 0 3 3 &n=
bsp; 0 inf 0<BR>
vmmpepl 88 2369173 &n=
bsp;0 2367913 30 0 &nb=
sp; 30 30 0 inf &=
nbsp; 2<BR>
vmmpekpl 88 120092 &n=
bsp;0 120087 1 0=
1 1  =
;0 inf 0<BR>
aobjpl 52  =
; 1 0  =
; 0 1 0 &nb=
sp; 1 1 0 i=
nf 0<BR>
amappl 40 581514 &nbs=
p; 0 581010 6 &nb=
sp; 0 6 6 &=
nbsp; 0 inf 0<BR>
bufpl 120 429136 &nbs=
p; 0 426155 94 &n=
bsp;0 94 94 0 &n=
bsp; inf 3<BR>
buf1k 1024 2128=
0 0  =
; 34 0 34 3=
4 1 1 0<BR=
>
buf2k 2048 &nbs=
p; 53 0 1 &=
nbsp; 2 0 2=
2 1  =
;1 0<BR>
buf4k 4096 &nbs=
p; 27 0 0 &=
nbsp; 2 0 2=
2 1  =
;1 0<BR>
buf8k 8192 &nbs=
p;774 0 0 =
97 0 97 &n=
bsp;97 1 1  =
;0<BR>
mbpl 256 =
527 0 &nbs=
p;0 34 0 34 &nbs=
p; 34 1 inf  =
;1<BR>
mclpl 2048 &nbs=
p;332 0 0 =
170 0 170 170 &n=
bsp; 4 1024 4<BR>
kqueuepl 260 &n=
bsp;4 0 1 =
1 0 1 &nbs=
p; 1 0 inf =
0<BR>
knotepl 72 &nbs=
p; 8 0 &nbs=
p;1 1 0 &n=
bsp;1 1 0 inf &n=
bsp; 0<BR>
sockpl 200 286837 &nb=
sp; 0 286676 9 &n=
bsp;0 9 9 =
0 inf 0<BR>
ttypl 264  =
; 71 0 &nbs=
p;0 5 0 &n=
bsp;5 5 0 inf &n=
bsp; 0<BR>
rndsample 528 10198 0=
10192 1 0=
1 1  =
;0 inf 0<BR>
procpl 408 51654 &nbs=
p; 0 51594 7 &nbs=
p; 0 7 7 &n=
bsp; 0 inf 1<BR>
lwppl 132 51682=
0 51618 3  =
; 0 3 3 &nb=
sp; 0 inf 0<BR>
pgrppl 16  =
;8134 0 8092 &nb=
sp;1 0 1 &=
nbsp;1 0 inf 0<BR>
pcredpl 24 51654 &nbs=
p; 0 51594 1 &nbs=
p; 0 1 1 &n=
bsp; 0 inf 0<BR>
plimitpl 156 221 &nbs=
p; 0 208 1 =
0 1 =
1 0 inf 0<BR>
pstatspl 224 51654 &n=
bsp;0 51594 4 &n=
bsp;0 4 4 =
0 inf 0<BR>
rusgepl 72 51594 &nbs=
p; 0 51594 1 &nbs=
p; 0 1 1 &n=
bsp; 0 inf 1<BR>
raspl 20 =
12 0  =
;11 1 0 &n=
bsp;1 1 0 inf &n=
bsp; 0<BR>
sadatapl 40 &nb=
sp; 8 0 7 &=
nbsp; 1 0 1=
1 0 inf &=
nbsp; 0<BR>
saupcpl 1608 9326719 0 &nbs=
p;9326719 2 0 &n=
bsp; 2 2 0 =
inf 2<BR>
sastackpl 24 128 &nbs=
p; 0 112 1 =
0 1 =
1 0 inf 0<BR>
savppl 52  =
; 8 0  =
; 7 1 0 &nb=
sp; 1 1 0 i=
nf 0<BR>
ptimerpl 124 6551 &nb=
sp; 0 6526 1 &nbs=
p; 0 1 1 &n=
bsp; 0 inf 0<BR>
filepl 56 1202155 &nb=
sp; 0 1201900 4 0=
4 4  =
;0 inf 0<BR>
cwdipl 12 51644=
0 51594 1  =
; 0 1 1 &nb=
sp; 0 inf 0<BR>
fdescpl 280 51644 &nb=
sp; 0 51594 4 &nb=
sp; 0 4 4 &=
nbsp; 0 inf 0<BR>
pnbufpl 1024 &n=
bsp;9 0 0 =
3 0 3 &nbs=
p; 3 0 inf =
0<BR>
vnodepl 164 30304 &nb=
sp; 0 914 1226 &n=
bsp;0 1226 1226 0 inf =
0<BR>
ncachepl 84 30791 &nb=
sp; 0 1401 613 &n=
bsp;0 613 613 0 =
inf 0<BR>
ffsinopl 176 30289 &n=
bsp;0 906 1279 0=
1279 1279 0 inf  =
; 1<BR>
dino1pl 128 30289 &nb=
sp; 0 906 949 &nb=
sp; 0 949 949 0 &=
nbsp;inf 0<BR>
wdcspl 52 652382 &nbs=
p; 0 652382 1 &nb=
sp; 0 1 1 &=
nbsp; 0 inf 1<BR>
extent 20  =
; 271 0 254  =
; 1 0 1 &nb=
sp; 1 0 inf  =
;0<BR>
rtentpl 132 2349 &nbs=
p; 0 2208 6  =
; 0 6 6 &nb=
sp; 0 inf 1<BR>
inpcbpl 108 276504 &n=
bsp;0 276457 2 0=
2 2  =
;0 inf 0<BR>
in6pcbpl 156 87=
0 69 &nbs=
p; 1 0 1 &n=
bsp; 1 0 inf 0<BR=
>
inmltpl 32 &nbs=
p; 11 0 1 &=
nbsp; 1 0 1=
1 0 inf &=
nbsp; 0<BR>
ipqepl 44  =
; 5 0  =
; 5 1 0 &nb=
sp; 1 1 0 i=
nf 1<BR>
tcpcbpl 316 16591 &nb=
sp; 0 16564 3 &nb=
sp; 0 3 3 &=
nbsp; 0 inf 0<BR>
tcpipqepl 44 6312 &nb=
sp; 0 6312 1 &nbs=
p; 0 1 1 &n=
bsp; 0 inf 1<BR>
synpl 184  =
;6355 0 6355 &nb=
sp;1 0 1 &=
nbsp;1 0 inf 1<BR>
sigapl 2052 51644 &nb=
sp; 0 51594 54 &n=
bsp;0 54 54 0 &n=
bsp; inf 4<BR>
siginfo 128 260727 &n=
bsp;0 260727 1 0=
1 1  =
;0 inf 1<BR>
ksiginfo 44 &nb=
sp; 8 0 7 &=
nbsp; 1 0 1=
1 0 inf &=
nbsp; 0<BR>
pipepl 140 84874 &nbs=
p; 0 84852 2 &nbs=
p; 0 2 2 &n=
bsp; 0 inf 1<BR>
<BR>
In use 26178K, total allocated 27032K; utilization 96.8%<BR>
</FONT><FONT FACE=3D"Verdana"><BR>
<BR>
Any insight or suggestions would be appreciated. I’m truly sorr=
y about the html formatting, but the stats wrapping seemed to not serve my p=
urposes.<BR>
<BR>
Thanks,<BR>
<BR>
Peter<BR>
<BR>
<BR>
</FONT></SPAN>
</BODY>
</HTML>
--B_3223550692_46383370--