Daniel
PF_RING attempts to reduce the number of polls, by using the watermark
that is a minimum number of packets that need to be received before
PF_RING wakes up the application. The drawback is the one you have seen
in case no packets are flowing. If you want to avoid that add "-w 1" in
pfcount
Luca
On 01/30/2012 04:20 AM, Daniel, Christopher wrote:
Hi Luca,
I am using the following config pfRing (on my linux version 2.6.18-194.17.1.el5)
[root@localhost ~]# cat /proc/net/pf_ring/info
PF_RING Version : 4.7.0 ($Revision: exported$)
Ring slots : 4096
Slot version : 13
Capture TX : Yes [RX+TX]
IP Defragment : No
Socket Mode : Standard
Transparent mode : Yes (mode 1)
Total rings : 0
Total plugins : 0
My pfRing is capturing on eth1.
I want to capture the occasional packet I insert (Tx) on the eth1 interface. It
looks like if there is very little traffic on
the Rx (20 pkts/sec) then the packet I insert (Tx) is captured after a 3-4
seconds. However if there is a continuous stream of packets coming in (1000
pkts/sec or more) on the Rx then the inserted packet (tx ) is captured
immediately.
Looks like a buffering issue. Is this a known issue. Is there a configuration I
need to do?
Any ideas?
Thanks,
Chris
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Baeder, Jason
Sent: Saturday, January 28, 2012 9:07 AM
To: [email protected]
Cc: [email protected]
Subject: Re: [Ntop-misc] CentOS 6.2 with PF_Ring?
Alfredo,
Actually we are aware that the bnx2 PF_RING-aware driver is an old release
I figured as much. Thanks for the confirmation. I have a less capable system
that does have Intel NICs. If I can show that working as a proof of concept,
I'll have justification to purchase an Intel NIC card for the Dell.
Jason
Sent from my iPad
On Jan 28, 2012, at 4:32 AM, "Alfredo Cardigliano"<[email protected]> wrote:
Sorry guys
we don't have such hardware in our lab to reproduce the issue. Actually we are
aware that the bnx2 PF_RING-aware driver is an old release and needs to be
updated. If someone want to port the new driver
(http://www.broadcom.com/support/license.php?file=NXII/linux-6.2.23.zip) is
welcome, it would be much appreciate.
Regards
Alfredo
On Jan 27, 2012, at 10:22 PM, Baeder, Jason wrote:
Richard,
You are not alone. I've been working on a Dell 2950 with two Broadcom 5708's
on board. I've seen the same behavior: pfcount (or tcpdump) only captures
packets with transparent_mode set to 0. Maybe it's a problem in the bnx2
driver.
Regards,
Jason
From: [email protected]
[mailto:[email protected]] On Behalf Of Richard Ullrich
Sent: Friday, January 27, 2012 3:35 PM
To: [email protected]
Subject: Re: [Ntop-misc] CentOS 6.2 with PF_Ring?
Thank You! That fix worked well.
I was able to compile both the PF_Ring and the bnx2 driver.
Now I seem to have another problem though...
I removed the current bnx2 'standard' driver (rmmod bnx2),
Then insmod bnx2.ko to install the PF_Ring aware driver, no errors reported.
ifconfig eth1 up, interface comes up as expected.
I then "insmod pf_ring.ko transparent_mode=0 enable_tx_capture=0 min_num_slots=32768
enable_ip_defrag=1"
(transparent_mode=0)
No errors reported,
if I do ./pcount -i eth1 - I see packets incoming
if I do ./pfcount -i eth1 - I see packets incoming. (shouldn't this ONLY read
from the Ring ??)
But, when I rmmod pf_ring, and then
insmod pf_ring.ko transparent_mode=2 enable_tx_capture=0 min_num_slots=32768
enable_ip_defrag=1"
(no errors reported) (transparent_mode=2)
if I do ./pcount -i eth1, I see no packets (which I expect, this should only
read from NAPI??)
if I do .pfcount -i eth1, I see no packets (which I don't expect!. - This
should be reading from the Ring?)
The other odd thing is, if I use transport_mode=1, the same as transport_mode=2
occurs.
Am I doing something wrong ??
Very much appreciate the assistance on this - I've been struggling with it for
weeks!<G>
Richard
Alfredo Cardigliano<[email protected]> 1/26/2012 6:15 PM>>>
Richard
please update to svn rev.5134 and let us know
Regards
Alfredo
On Jan 27, 2012, at 1:48 AM, Richard Ullrich wrote:
Has anyone successfully compiled and used the current svn build (5133) with
CentOS 6.2 ?
When I attempt the /PF_RING/kernel/make, I get :
make -C /lib/modules/2.6.18-274.7.1.el5/build SUBDIRS=/root/PF_RING/kernel
EXTRA_CFLAGS='-I/root/PF_RING/kernel -DSVN_REV="\"5133\""' modules
make[2]: Entering directory `/usr/src/kernels/2.6.18-274.7.1.el5-x86_64'
CC [M] /root/PF_RING/kernel/pf_ring.o
/root/PF_RING/kernel/pf_ring.c: In function âallocate_extra_dma_memoryâ:
/root/PF_RING/kernel/pf_ring.c:4282: warning: passing argument 1 of
âdma_mapping_errorâ makes integer from pointer without a cast
/root/PF_RING/kernel/pf_ring.c:4282: error: too many arguments to function
âdma_mapping_errorâ
make[3]: *** [/root/PF_RING/kernel/pf_ring.o] Error 1
make[2]: *** [_module_/root/PF_RING/kernel] Error 2
make[2]: Leaving directory `/usr/src/kernels/2.6.18-274.7.1.el5-x86_64'
make[1]: *** [all] Error 2
make[1]: Leaving directory `/root/PF_RING/kernel'
make: *** [all] Error 2
Am I just not 'holding my teeth' right??<G>
Thanks!
Richard
_______________________________________________
Ntop-misc mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop-misc
_______________________________________________
Ntop-misc mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop-misc
_______________________________________________
Ntop-misc mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop-misc
_______________________________________________
Ntop-misc mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop-misc
_______________________________________________
Ntop-misc mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop-misc
_______________________________________________
Ntop-misc mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop-misc
