[tcpdump-workers] multiple libpcap ring buffers?

2006-02-16 Thread J.O. Leger

Hey,



If I have an application that opens two packet capturing sessions(ie eth0 and 
eth1) using the ring buffer version of libpcap, does libpcap create two 
separate ring buffers?



Thanks for your help



Cheers



 







___
Join Excite! - http://www.excite.com
The most personalized portal on the Web!


-
This is the tcpdump-workers list.
Visit https://lists.sandelman.ca/ to unsubscribe.


Re: [tcpdump-workers] PCAP Timestamp - HWClock or SWClock?

2005-05-11 Thread J.O. Leger

Actually, I really don't care how arcuate the timestamps are.



What I REALLY want to know is where does libpcap, not tcpdump, get its 
timestamp from when reading from a linux interface.



The application that is using libpcap sometimes displays unusual timestamps.  
I believe this is caused by the hardware clock and the software clock being set 
to different times.



So I go back to my original question.  Where does libpcap get its timestamp 
from in linux?  The hardware clock or the software clock?



Based on what you said below it seems to be the software clock.



Thanks for info.



J.O.





 --- On Fri 05/06, Guy Harris  [EMAIL PROTECTED]  wrote:

From: Guy Harris [mailto: [EMAIL PROTECTED]

To: tcpdump-workers@lists.tcpdump.org

Date: Fri, 06 May 2005 13:09:12 -0700

Subject: Re: [tcpdump-workers] PCAP Timestamp - HWClock or SWClock?



J.O. Leger wrote:br Is the timestamp in pcap_pkthdr from the hardware clock 
or the software clock?brbrThe timestamp is from whatever it's from. 
:-)brbrIf you're capturing on an interface on a UN*X or Windows (with 
WinPcap) brmachine, the time stamp is from the capture mechanism that libpcap 
uses. br  Those capture mechanisms almost always use a time stamp maintained 
by brthe OS, and those are usually the combination ofbrbr 1) a software 
clock (in the sense of a value updated periodically by an brinterrupt 
handler), giving the low-precision part of the time (probably brin 1/100ths 
of a second, maybe higher)brbrandbrbr   2) a hardware time stamp 
device of some sort, giving the high-precision brpart of the time.brbrIf 
you're capturing from a DAG card from Endace (recent versions of brlibpcap 
can be built with support for DAG cards), I suspect the time brstamp might 
come from a purely hardware clock.brbrI suspect, however, that you 
didn't ask a question the answer to which brtells you what you ultimately 
really want to know.brbrIf what you *really* want to know is how accurate 
are the time brstamps?, then the answer is if you care enough about that to 
ask, they brprobably aren't as accurate as you'd like, becausebrbr1) 
The time stamping isn't done at the instant that the first byte, or brthe 
last byte, of a packet being received is processed by the network bradapter 
hardware - it's done some time after that, at least for non-DAG brcaptures.  
The time stamping is done after the host sees the packet. brThat could be 
delayed by interrupt latency, polling (i.e., the driver brmight be running 
the adapter in a mode where it doesn't interrupt the brhost for every packet, 
but where the driver periodically polls the card brto see whether packets 
have arrived, so that it gets fewer interrupts brand gets many packets per 
interrupt, to reduce overhead), and the code brpath 
between the code that initially handles the device or clock brinterrupt and 
the code that gives the packet a time stamp.brbr  2) The hardware time 
stamp device of some sort might, or might not, be bra very high accuracy 
device - in some OSes on newer x86 boxes it might, brfor example, be the 
processor's time stamp counter, which doesn't brnecessarily tick at a rate 
that corresponds well to high-precision brfractions of a second (I think that 
can be an issue with WinPcap when it bruses an internal Windows kernel 
routine to read that counter, for example).brbr(There might be other issues 
I've missed.)brbrSo if you need, for example, time stamps with microsecond 
accuracy, brlibpcap probably won't give it to you if you're not capturing on 
a DAG brcard.  Even *relative* time stamps - i.e., differences *between* time 
brstamps - probably won't have microsecond accuracy, because the various 
britems I listed above don't put a constant offset into the 
time stamps, brthe offset from reality can vary from time stamp to time 
stamp.brbrCapturing with a DAG card might help.  I can't speak for the 
Endace brpeople on that; you should contact them if you need high-accuracy 
time brstamps:brbrhttp://www.endace.com/brbrThere might be other 
hardware that you could use, although you'd have to brmodify libpcap to 
support it.br-brThis is the tcpdump-workers list.brVisit 
https://lists.sandelman.ca/ to unsubscribe.br

___
Join Excite! - http://www.excite.com
The most personalized portal on the Web!


-
This is the tcpdump-workers list.
Visit https://lists.sandelman.ca/ to unsubscribe.


[tcpdump-workers] PCAP Timestamp - HWClock or SWClock?

2005-05-06 Thread J.O. Leger

Is the timestamp in pcap_pkthdr from the hardware clock or the software clock?



Thanks



J.O.







___
Join Excite! - http://www.excite.com
The most personalized portal on the Web!


-
This is the tcpdump-workers list.
Visit https://lists.sandelman.ca/ to unsubscribe.