lwip-users
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [lwip-users] Problems with sending UDP packets in FreeRTOS


From: goldsimon
Subject: Re: [lwip-users] Problems with sending UDP packets in FreeRTOS
Date: Thu, 28 Jul 2016 21:46:03 +0200
User-agent: AquaMail/1.6.2.9 (build: 27000209)

Actually, this is too much of a hardware issue for me to support right now. I do use the stm32 and stmcube, but not the Ethernet part. I can only imagine they have some bug in their driver under load. OTOH, are you absolutely sure your 4ms sleep is really 4ms long?

Simon


Gesendet mit AquaMail für Android
http://www.aqua-mail.com


Am 28. Juli 2016 8:48:51 nachm. schrieb Mike Fleetwood <address@hidden>:

Hi Simon,

That is exactly what I was trying to identify.

I put the "borrowed" code in to prove/check that fault wasn't in what I
had written - succeeded there, I think.

After that, it has to be at some lower level - that it, either in LWIP
or in ST's driver interface.  So, now we know the low level driver
returns ERR_USE (when maybe it would be more useful to return something
like ERR_DMA?).  I was kind of hoping that someone might be able to
suggest a reason why?

It has been suggested that the data rate is too high.  An overall packet
size of 67 bytes sent every 4mS is not exactly "SuperFast".  I have a
180MHz CPU and a 10MHz ethernet segment.  I know the whole packet send
gets processed, from creation to end of wire transmission in a maximum
of about 200uS - long before I start creating the next one - so what
possible reason is there for the DMA to run out of buffers?  There is
virtually no other traffic on this network segment that could cause
collision and no ARP to wait for (which would have happened at a higher
level anyway).

I have reduced the send rate to one every 100mS, and the packet data is
25 bytes - so far (since last compile) it has successfuly sent about 20k
packets.  This is with all the other threads enabled and regular access
to the HTTP status page to see how it's getting on (no warning
messages).  I'm using my original udp_test function - that checks for
errors and reports them (not the "borrowed" one!).

I'm going to try bigger or faster packets in a minute.

Many thanks,

Mike.


On 28/07/2016 16:27, goldsimon wrote:
Sylvain Rochet wrote:
The fact that it is your code and not your code doesn't change anything
about the fact that it is completely wrong.

Indeed. Not completely wrong maybe, but at least bad quality. However,
it changes everything about suspecting something about Mike driver's
code quality:

I assume your whole software quality is between
poor to critically wrong. Given that, the bug is
probably in your low level netif driver.

And I guess that could have been what Mike was complaining about... ;-)

Simon




Gesendet mit AquaMail für Android
http://www.aqua-mail.com



_______________________________________________
lwip-users mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/lwip-users

--
FACE Systems Ltd
The Old Boat House
Cadgwith
Cornwall TR12 7JX
T:01326 291031
M:07831 401464


_______________________________________________
lwip-users mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/lwip-users





reply via email to

[Prev in Thread] Current Thread [Next in Thread]