[mythtv-users] Cheap frontend for HD

[Gary Buhrmaster]

On Tue, Jun 30, 2015 at 4:25 PM, Jim Stichnoth <stichnot at gmail.com> wrote:
.....
> Has anyone gotten this ECS Liva to PXE-boot at a reasonable speed?  Or
> solved similar slow-PXE-boot problems for other Linux systems?

"Typically" the problem is tftp udp transaction issues (tftp is optimized
for low end systems and the protocol is very chatty, and sends single
512 byte packets at a time).  One can sometimes mitigate against
some of the issue by increasing the packet size to a larger number.
Depending on your network infrastructure, you may want to keep the
number under the MTU of your network, or 1468, but 65464 can work
in some situations.

The support must be enabled on both the server and client, although
some clients will attempt to use the blocksize enhancements without
any additional configuration and some, likely not (I *thought* the normal
pxeboot will do this, but it has been some time since I looked at this).
On the server, look at the --blocksize/-B parm to be sure that the
server is enabled for larger blocksizes.

And the " tftp: client does not accept options" suggests that the client
may not be accepting one or more options, and may fall back
interestingly.  You may need to use the --refuse/-r option to disable some
option negotiation to make the client happy.

See RFC 2348 for details (and sample graph on performance).
https://tools.ietf.org/html/rfc2348


If all else fails, a tcpdump should show whether you are using
a block size of 512 or some other interesting transaction is
occurring (and/or verbose logging).

That is the limit of my random recollections at this point.