login
Header Space

 
 

RE: Question of Network 1Gbps Intel Packet Split

Score:
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
To: Raz <raziebe@...>
Cc: <netdev@...>
Date: Wednesday, April 23, 2008 - 11:46 am

Raz wrote:

If you'd included something about e1000/e1000e in the message our
filters would have marked this so we could reply to you :-)


All packets that are 
a) recognized, i.e. IPv4, TCP, UDP, IPv6, etc
b) not too short (sometimes packets < 70 bytes or so are not split, at
least on 82571)
c) sometimes the header after the TCP/UDP header can be split too, like
maybe for NFS, you're welcome to read the manual for PCIe parts posted
at e1000.sf.net


nr_frags will typically be 1, unless the frame is too short ( < 128 )
and then it is "unsplit" into the header buffer to avoid the page
free/allocation.

In the case of e1000/e1000e, splitting the packet is best used for jumbo
frames, and in that case multiple pages can be used to hold the packet,
with a single header buffer at skb->data, and nr_frags > 1

Hope this helps, and in the future you're welcome to post to
e1000-devel@lists.sf.net for these kinds of specific to Intel hardware
questions.
  
  Jesse
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

Messages in current thread:
RE: Question of Network 1Gbps Intel Packet Split, Brandeburg, Jesse, (Wed Apr 23, 11:46 am)
speck-geostationary