Re: [Xen-users] Use of "ethtool - K eth0 tx off"

["Leonardo Pinto" <listas@xxxxxxxxxxxxxxxx>]

Much thankful by the answers, had been very valuable.
Not, really this bug is not corrected still. At least in the DomU
it continues being necessary disable checksum. I am using the version:
xen-3.0.2-0.FC5.3 Before (in <= 3.0.1) it needed in both domains, DomU
and Dom0. I also perceived that checksum in the internal domains was
not thus much-needed. Beyond this control being made by the proper
net-board (in Dom0), "could save some CPU cycles". Evidently,
improving the performance in the transmitted packages.

My grammatical corrections are not so necessary, seem that you, for
the little that understood, it answered me to all the questions.
It forgives me the bad English.

- The question that remained is: When fixed this bug, WE MUST then
disable checksum in internal domains, in favor of the performance?

Much thankful,

--
Leonardo Pinto
listas#openlogic dot com br


On Sun, 28 May 2006 23:04:56 -0300 (BRT), xen-users-request wrote
> On Sun, May 28 '06 at 09:39, Leonardo Pinto wrote:
> > I am with problems using method NAT and having that to make use of
> > "ethtool - K eth0 tx off"
> 
> we all have (well, at last with 3.0.1, I'm not 100% sure if this has
> been fixed with 3.0.2)
> 
> > - Why is being necessary the indiscriminate use of "tx off"?
> 
> EPARSE
> 
> > - That is one bug or resource? It will be corrected?
> 
> It's a bug. And shortly before 3.0.2 was relesed there were some
> attempts to fix it.
> 
> It is intended as an optimisztion, but it failed for some (or IMHO most
> real world) setups.
> 
> > Another fact is, if the use of "tx off", we would not be compromising
> > the security?
> 
> EPARSE
> 
> But disableing TX checksum offloading (that's what ethtool -K tx off
> does) does not affect security at all. The CPU only has to compute
> the checksum off all outgoing packages, even if the package is intended
> for another domU or the Dom0 where not computing the checksum could save
> some CPU cycles (as coruption of the package is VERY unlikely).
> --  
> /"\ Goetz Bock at blacknet dot de  --  secure mobile Linux


_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users