RE: PacketSendPackets and latency

"David Atkins" <[email protected]>
Newsgroups gmane.comp.security.nmap.devel
Message-ID <[email protected]>
Hi,

 

So the delay increases with the size of buffer in a linear fashion, the packets send with the correct delays between them what we get is a big “hole” in sending between chunks of packets.

 

As far as I can see direction mode is used which should mean no copies to a  buffer for transfers to the driver should be done which is what seems to be the most likely cause of the delay.

 

I have not evaluated if we see the same issue on reads from the driver but will probably aim to do so.

 

David

 

 

From: Varunram Ganesh [mailto:[email protected]] 
Sent: Thursday, March 23, 2017 9:10 AM
To: Nmap-dev
Cc: Daniel Miller; David Atkins
Subject: Re: PacketSendPackets and latency

 

Hi David,

 

Disregard what I said about -d2. I thought you were trying to send data using --send-eth and similar using nmap.

 

Coming back to the problem at hand, from what I know with regard to npf, NPF_BufferedWrite() scans the buffer and then calls NdisSend to send the same. So, if the buffer has enough space, it should queue the data, send and immediately return. I have seen such errors only when we receive and send packets both at the same time, but I haven't sent 30Mb or seen issues which report the same. A quick question - Does Sending a smaller buffer cause the same 13ms delays? My guess is that its taking time to scan the buffer (assuming you are resetting the buffer contents), but I could be wrong.

 

 

Cheers,

 

Varunram

_______________________________________________
Sent through the dev mailing list
https://nmap.org/mailman/listinfo/dev
Archived at http://seclists.org/nmap-dev/
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.