Re: CheckUDP problem
- From: Maarten <Maarten@xxxxxxxxxxxxxxxxxxxxxxxxx>
- Date: Mon, 26 Nov 2007 08:50:02 -0800
Dear Paul,
I did search for ChechUPD already and will do it again, but I can not
understand why the same platformbuilder runs OK with another system. The only
difference is the PHY I'm using (LAN8700 of SMSC). Suprising is that the
download most of the times hangs at exactly the same amout of data
transfered. Once checkUDP occurs (EbootEtherReadData), I see with a TCPIP
sniffer that PB retries correctly, but my platform keeps giving CheckUDP
errors. I am affraid that I getting into loopback tests and so......
--
Maarten
"Paul G. Tobey [eMVP]" wrote:
A search for "CheckUDP" in the archives of.
microsoft.public.windowsce.platbuilder gives fifty-some results. I think
I'd read all of those before moving forward. I seem to recall some download
problems when certain settings on the PC, in the regsitry, were
non-standard, for example.
http://groups.google.com/advanced_search?q=&hl=en&;
Paul T.
"Maarten" <Maarten@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:0C2E6C39-63C8-4FDB-9552-BE6BD5EB06FF@xxxxxxxxxxxxxxxx
Dear GraemeW,
I tried this, but no success. I see that the (switched off) verify
checksum
function may not causing the problem, During communication I already get
some
!CheckUDP: UDP header checksum failure messages before the system 'hangs'.
The problem also is that EBOOT runs OK on a similar system (same processor
and memory and internal mac) with a different PHY, so I guess I have to
focus
on hardware
any hints on how to investigate this?
--
Maarten
"GraemeW" wrote:
I used to see problems with downloads a long time ago with some
platforms,
the downloads would mostly continue but there were UDP timeout errrors.
The way an nk.bin is downloaded, its basically a set of SRecords or
blocks
of data each with a checksum, the bootloader checksums each record AFTER
its
downloaded, this is normally quick as the blocks are small, but some
blocks
can be big, i.e. 1Mb in size, this takes a little while to calculate the
checksum and the next UDP record from the TFTP server (i.e. PB) is lost
as
the bootloader wasn't servicing the network card whilst it was doing the
checksum. So I calculated this checksum as the record is being
downloaded,
this increased download speed and got rid of the UDP errors, but for a
test
you can just try disabling it.
Maybe worth a try, its easy to do, copy the BLCOMMON out of platform, and
just comment out the code in
static BOOL VerifyChecksum (DWORD cbRecord, LPBYTE pbRecord, DWORD
dwChksum)
Always return TRUE, see if it helps... you can speed up the download on
slowish CPUs by doing the calculation of the checksum as each packet it
received, i got a 10-15% increase in download speed
--
GraemeW
Blog - http://ce4all.blogspot.com
"Maarten" <Maarten@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:9B0DBE24-91A2-49A4-8079-FDC576A0FFBB@xxxxxxxxxxxxxxxx
I'm struggling with a CheckUDP problem in my 5.00 platform during
kernel
download. Suprisingly the error always occurs at the same moment (about
2.3Mb
of kernel transfer), so I am more and more asking myself if it is
hardware
or
software
To be sure my mac/phy operate correct, I would like to test these
separate
from the rest of the project. Could anyone give some hints how to do
this?
Could I e.g. use OALKitlethSendFrame/OALKitlethGetFrame to test the
buffer
contents with a loopback?
--
Maarten
- Follow-Ups:
- Re: CheckUDP problem
- From: Paul G. Tobey [eMVP]
- Re: CheckUDP problem
- References:
- Re: CheckUDP problem
- From: GraemeW
- Re: CheckUDP problem
- From: Maarten
- Re: CheckUDP problem
- From: Paul G. Tobey [eMVP]
- Re: CheckUDP problem
- Prev by Date: Re: CheckUDP problem
- Next by Date: Re: Strange Registry behaviour
- Previous by thread: Re: CheckUDP problem
- Next by thread: Re: CheckUDP problem
- Index(es):
Relevant Pages
|
