Atomwide Pocket Ethernet Adaptor

[MarkY...@yahoo.com]

I have just purchased a Atomwide Pocket Ethernet Adaptor (PEA) from CJE as I would like to get my A3010 onto my Ethernet (I have no room for a podule as that slot is full to the max), and I can see that the PEA is DCI-2 and therefore will not work with !Internet v5 that is in my Uniboot setup.

The question is how and where do I get the files in order to put an older version of !Internet on the machine that will accept the DCI-2 device?

Many thanks in advance.

Mark

[Theo Markettos]

MarkY...@yahoo.com wrote:
> The question is how and where do I get the files in order to put an older
> version of !Internet on the machine that will accept the DCI-2 device?

I think Internet 4 is around on one of the Acorn FTP site archives, but it's
in an obscure place I don't remember[1].

Another option is the Freenet stack, which was DCI2. Try using the Freenet
version of Acornet from here:
http://www.drobe.co.uk/archives//freenet.barnet.ac.uk/acornet/public_html/download.html
Older versions can be found here:
http://www.drobe.co.uk/archives/index.php?directory=/freenet.barnet.ac.uk/Acorn/acornet/

[1] Cryptic hints in this thread:
http://www.digipedia.pl/usenet/thread/1613/2041/

Theo

[MarkY...@yahoo.com]

Ah, I did find a copy of !Internet 4 on http://arcade.demon.co.uk/filepages/file63.htm

I didn't realise it supported DCI2. I thought it was only version 2 that did. I will have a play with Freenet and version 4 and see what I get.

[Theo Markettos]

MarkY...@yahoo.com wrote:
> Ah, I did find a copy of !Internet 4 on http://arcade.demon.co.uk/filepages/file63.htm

Thanks. I did look there, but missed it.

> I didn't realise it supported DCI2. I thought it was only version 2 that
> did. I will have a play with Freenet and version 4 and see what I get.

If you're really in for masochism, try !TCPIP from the same site. This is
KA9Q, the first free RISC OS internet stack which talks to DCI2 drivers.
However since it doesn't have a Sockets API, you can only use the supplied
programs to connect to the internet. I think it predates web browsers,
even.

Theo

[MarkY...@yahoo.com]

Well, I got nowhere with the acorn stack. Just kept getting that it couldn't find the interface. Tried tried and tried again and it did make the errors different, but there was no sign that I was getting any actual life from the interface.

So went on to Freestack.

Very odd, when I try and get the interface up via:-

ifconfig ep0 inet 10.0.0.2 up

It responds with the red light going out on the interface and the following message:-

Testing individual interrupts........................FAILED
ep0:interface hardware faulty.

Then if I do it again it crashes the computer.

Intrestingly I think the module must contain the MAC address too as when I use 'epinfo' it shows exactly the same information whether the interface is plugged in or not.

I am guessing these interfaces are a bit of a trick to get up and running!

[MarkYoungIW]

Something that does occur is that if I plug the adapter in and run just EtherP and nothing else, when I type 'EPInfo' I always get the same response (actually its the same whether the interface is plugged in or not!)

*EPInfo
etherp interface statistics

ep0: bussize 8 (1a), slot 1, disabled, hardware address **:**:**:**:**:**

Wondering why 'disabled'. This would seem to indicate that the driver is unable to see the device? I wouldn't have though that the stack has any bearing on this, if the driver can't see the device that would seem to be more likely where the problem lies?

On an A3010 are there any parallel port settings, just wondering if its possible to have the port in the wrong mode or something.

[MarkYoungIW]

At the moment Freestack seems to be the only one that causes anything to happen, which is that it reports the interface as faulty, once it has made that report, typing epinfo also shows that the drivers has recognised a couple of packet types and it too shows a faulty hardware error.

Hmm......Hoping this isn't the case!

Will try some older Acorn Stacks in a bit (Version 2 TCP/IP) and see what happens there, but at the moment things aren't looking too good!

[MarkYoungIW]

Just tried the module on an A5000 to and the MAC Address shown in EPInfo is different, so the module must create the MAC address from some ID in the computer. I was concerned perhaps I had the wrong module for the interface as the MAC that was showing wasn't the same as the one that had been pencilled on it.

So at the moment I think all I can do is keep playing with different stacks etc.

As far as I can see the EPInfo shows 'Disabled' until you try and bring the interface up with the stack, so at the moment it could just be a case that it doesn't work with Freestack.

As the manual gives instructions for use with Version 1+Version 2 of the Acorn Stack, I guess this is the way to go!

And there was me thinking that hopefully this would be plug and play!! Oh for a DCI-4 driver!!

[MarkYoungIW]

Thats not looking so good then. I've set up TCP/IP Version 2 as per the manual, and when I try and get the interface up I get a Faulty Hardware Error again.

And exactly the same as Freestack does, when I type EPInfo I get:-

etherp interface statistics

ep0: bussize 8 (1a), slot 1, faulty, hardware address **:**:**:**:**:**

tests failed

no transmit iterrupt
no receive interrupt
Frame types recognised: 0x0800, 0x0806

Log: ep0: interface hardware faultyep0: interface hardware faulty

This does kind of give the impression the adaptor is faulty!

Hmm....What to do.

[Theo Markettos]

MarkYoungIW <marky...@yahoo.com> wrote:
> Thats not looking so good then. I've set up TCP/IP Version 2 as per the
> manual, and when I try and get the interface up I get a Faulty Hardware
> Error again.

'TCP/IP Version 2' - do you mean KA9Q or Acorn's Internet 2 which I think
was a commercial product in circa 1991ish, ah:
http://acorn.chriswhy.co.uk/docs/Acorn/Brochures/Acorn_APP286_TCPIPProtocolSuite.pdf
I've never seen this released.

> And exactly the same as Freestack does, when I type EPInfo I get:-
>
> etherp interface statistics
>
> ep0: bussize 8 (1a), slot 1, faulty, hardware address **:**:**:**:**:**

The MAC address seems to be coming from the machine unique ID which is on
the motherboard, so that explains why it's available if there's no hardware
plugged in.

> tests failed
>
> no transmit iterrupt
> no receive interrupt
> Frame types recognised: 0x0800, 0x0806
>
> Log: ep0: interface hardware faultyep0: interface hardware faulty
>
> This does kind of give the impression the adaptor is faulty!

You should get some response even if there's no stack configured - just
loading EtherP should be sufficient to run *EPInfo.

Sounds like you need to fix the 'hardware faulty' first. Do you have
network connected to the dongle?

Theo

[MarkYoungIW]

> 'TCP/IP Version 2' - do you mean KA9Q or Acorn's Internet 2 which I think

Err.....!TCPIP v2.03 (21-Feb-1995)

Adam Goodfellow <ad...@computech.demon.co.uk> 04/03/95

So I guess not actually Acorn's v2 as I'd written on the disc. Oops!

>
> The MAC address seems to be coming from the machine unique ID which is on
>
> the motherboard, so that explains why it's available if there's no hardware
>
> plugged in.
>

Thats definately the case. Tried the module on an A5000 without the interface plugged in and got a different MAC address.

> You should get some response even if there's no stack configured - just
>
> loading EtherP should be sufficient to run *EPInfo.
>

I can run EPInfo as long as the module is loaded, but it comes back with:-

etherp interface statistics

ep0: bussize 8 (1a), slot 1, disabled, hardware address **:**:**:**:**:**

Suggesting that the interface isn't up.

>
> Sounds like you need to fix the 'hardware faulty' first. Do you have
>
> network connected to the dongle?
>

No dongle, just the adaptor plugged straight in to the parallel port.

Its got me at the moment that both stacks I've tried that seem to try and bring the interface up respond with a hardware error.

[Theo Markettos]

MarkYoungIW <marky...@yahoo.com> wrote:
>
> > 'TCP/IP Version 2' - do you mean KA9Q or Acorn's Internet 2 which I think
>
> Err.....!TCPIP v2.03 (21-Feb-1995)
>
> Adam Goodfellow <ad...@computech.demon.co.uk> 04/03/95
>
> So I guess not actually Acorn's v2 as I'd written on the disc. Oops!

Right, so that's KA9Q. Welcome to the funhouse ;-)

Did you read the instructions? On some of the stacks (can't remember which)
you have to put the driver module in a specific directory for the stack to
bring it up.

> > Sounds like you need to fix the 'hardware faulty' first. Do you have
> > network connected to the dongle?
> >
>
> No dongle, just the adaptor plugged straight in to the parallel port.

Dongle=adaptor, same thing. Do you have a 10base2 network - other host,
cables, terminators, the lot? It might be doing things like listening on
the network - if there's nothing plugged in, it'll never hear anything.

Theo

[MarkYoungIW]

> Dongle=adaptor, same thing. Do you have a 10base2 network - other host,
>
> cables, terminators, the lot? It might be doing things like listening on
>
> the network - if there's nothing plugged in, it'll never hear anything.
>
>
>
> Theo

Yes, 10B2 via a 10B2-10BT Hub to my network. Will check I haven't messed up the terminator setting on the hub though in the morning as I know it has a switch, but dont remember checking what it was set to.

Sorry, when you said Dongle, I thought you meant like the pass through Computer Concepts type. :)

Will check the physical side of the network in the morning. And have another go.

Fingers crossed.

[Theo Markettos]

MarkYoungIW <marky...@yahoo.com> wrote:
> Yes, 10B2 via a 10B2-10BT Hub to my network. Will check I haven't messed
> up the terminator setting on the hub though in the morning as I know it
> has a switch, but dont remember checking what it was set to.

You'll need terminators on both ends:

!--T------T--!
| |
Acorn Hub

! = terminator
T = T-piece

Ideally you want the T-pieces right in the back of the machines, ie no drop
lead between the NIC and the T-piece. I never actually built a 10base2
network though, so I don't know how critical this is.

Theo

[MarkYoungIW]

I've checked the network and as far as I can see everything is correct. Tried it with an old network card in a PC and it can access my network through the 10B2 section, so hopefully this is correct.

The second I try and bring the interface up the power light on the adaptor goes out and it comes up faulty. This is with both Freestack and KA9Q.

In fact with KA9Q the second DCIDrv module runs the light on the adaptor goes out and brings up an error of:-

live network test with correct CRC.................FAILED

[MarkYoungIW]

Also when the PEA is attached to the hub the connection light on the hub never lights up no matter what I do, whereas when I tested it with an old network card in a PC it lit up the moment there was power.

[MarkYoungIW]

Might be bad termination. Not used Thinnet in a long time, but noticed that I have 2 open terminators rather than 1 open and 1 grounded. Possibly the PC network card is less fussy than the PEA. Ordered new terminators which will be with me Friday, so will try again then.

Fingers crossed!!

[Justin Fletcher]

On Thu, 11 Oct 2012, MarkY...@yahoo.com wrote:

> I have just purchased a Atomwide Pocket Ethernet Adaptor (PEA) from CJE
> as I would like to get my A3010 onto my Ethernet (I have no room for a
> podule as that slot is full to the max), and I can see that the PEA is
> DCI-2 and therefore will not work with !Internet v5 that is in my
> Uniboot setup.

If the idea of using the Internet on a A3010 hasn't already convinced you
that you're clearly nuts, then the fact that it uses DCI-2 should have
been an immediate sign that you're in for an excessive amount of pain.
Given that the pocket ethernet adapter costs £60 on CJE's site, I'd
suggest just saving up for a real machine and forget wasting your time
on massively obsolete hardware.

For a little over twice that price you can get one of the HP
Microservers[1] with 2GB of memory and a 250GB disc. Install linux, shove
RPCEmu on it and you're happy. This would be a world less pain than
wasting any time even considering getting the A3010 to any sort of useful
state. It's beyond an 'interesting exercise' to get a network stack and
useful applications running on it, and into well into realms of masochism.

All the best to you but... just... no.

[1] http://www.ebuyer.com/281915 - obviously other options exist, but the
microservers are great little systems.

--
Gerph <http://gerph.org/>
... Marking the space between the days;
Early hours pass away.

[spampling]

In article <alpine.DEB.1.10.1...@buttercup.gerph.org>,

Justin Fletcher <ge...@gerph.org> wrote:
> For a little over twice that price you can get one of the HP
> Microservers[1] with 2GB of memory and a 250GB disc. Install linux,
> shove RPCEmu on it and you're happy.

RPi plus peripherals would probably make them happier and secure in the
glow of ARM hardware. :-)

--

Steve Pampling

[MarkYoungIW]

Quite happy with just being nuts and going for the challenge. Loved the machine when I got it, and always wanted to see what it can do. After much help on here and getting Econet up and running on it, it seemed sensible that Ethernet should be the next challenge. Its not about speed or practicality. Its just whether it can be done!

I want the most expanded A3010 I can accomplish, just using expansions of that era!

I agree though, it is a shame the PEA was only DCI2. Things could have been so much simpler if it had been DCI4. Never seen any sign of anyone attempting a driver for it though!

[Jess]

In message <alpine.DEB.1.10.1...@buttercup.gerph.org>

Justin Fletcher <ge...@gerph.org> wrote:

> If the idea of using the Internet on a A3010 hasn't already convinced you
> that you're clearly nuts, then the fact that it uses DCI-2 should have
> been an immediate sign that you're in for an excessive amount of pain.
> Given that the pocket ethernet adapter costs £60 on CJE's site, I'd
> suggest just saving up for a real machine and forget wasting your time
> on massively obsolete hardware.

I suspect it's more a case of seeing what it will do rather than an
attempt to make a main internet machine.

I just plugged up my A3020 to see what happened. After resetting
everything I got it online after a fashion. FTP worked.

It was viable as an FTP client, or a chat client or to use email.

However the problem is that the protocols changed on IM and the new
programs are 3.5 and above.

Messenger was too big to run in the available ram with stronged
(though perhaps a different !boot might solve this).

Web browsing was pretty poor, because the programs that would run,
crashed when they got a nasty page.

I reckon a solid state drive, might lower the pain level, however i
would be very dubious about spending money on kit that can't be
deployed somewhere else if it still proves to be to painful to use.

--
Jess Iyonix

[Theo Markettos]

MarkYoungIW <marky...@yahoo.com> wrote:
> I agree though, it is a shame the PEA was only DCI2. Things could have
> been so much simpler if it had been DCI4. Never seen any sign of anyone
> attempting a driver for it though!

I don't think the PEA was common enough for there to be enough interest in a
new driver.

Alan Williams did some messing around with a DCI4 driver for a D-Link DE600
parallel ethernet adaptor. I don't think it was fully working, or know
where you'd get a DE600 these days. I've got the code somewhere if anyone
is interested.

Scratch that, there's even one on ebay:
http://cgi.ebay.co.uk/300754074439

Theo

[MarkYoungIW]

That photo is very similar to the the ANT PEA. Apart from lack of power light on the top, I would almost go that the ANT PEA is just the D-Link one rebranded!

[patric aristide]

On 2012-10-18, Jess <phant...@hotmail.com> wrote:
> In message <alpine.DEB.1.10.1...@buttercup.gerph.org>
> Justin Fletcher <ge...@gerph.org> wrote:
>
>> If the idea of using the Internet on a A3010 hasn't already convinced you
>> that you're clearly nuts, then the fact that it uses DCI-2 should have
>> been an immediate sign that you're in for an excessive amount of pain.

>> Given that the pocket ethernet adapter costs ?60 on CJE's site, I'd

>> suggest just saving up for a real machine and forget wasting your time
>> on massively obsolete hardware.

Well, it's just for sport, isn't it? I certainly got a lot of fun
out of my A4000 that way.
Not that I know anything about the PEA but my icubed EtherLan card
worked a treat. FTP turned out to be really useful, LIRC at least
provided some amusement. ArcWeb and TTFN for mail/news did work but
somehow left a lot to be desired. Especially the latter.
Telnetted into my shell account to use pine and slrn instead. To my
surprise Lynx and Elinks even made the web accesible to my 4MB machine.
Of course none of this is really useful in itself but it did teach me
one or two things.

Patric
--
GL

[Theo Markettos]

If you wanted to check, you could try doing things like reading the MAC
address according to the code in the DE600 Linux driver and seeing if it was
sane:
http://www.cs.fsu.edu/~baker/devices/lxr/http/source/linux/drivers/net/de600.c?v=2.6.25
(line 422 onwards)

Theo

[MarkYoungIW]

Now the fun starts. Sure enough it was just bad termination!! With a grounded terminator on one end and and open one on the other, I can ping my router sucessfully!!

10ms ping time on the A3010 1 ms on the laptop!! ;o)

Now what!

[MarkYoungIW]

I have decided to use !Freenet 1.1.2.1 as it is the only version of !Freenet I have been able to find that supports DCI2 (although there is meant to be a version 1.2 somewhere) and KA9Q has so far just messed with my head!!

My startup:-

RMRun FreeNet:FreeNet
RMRun FreeNet:Internet

RMRun FreeNet:Drivers:EtherP

RMRun FreeNet:DNSResolve

Set Inet$HostName AcornA3010

ifconfig lo0 inet 127.0.0.1 up
ifconfig eo0 inet 10.0.0.119 netmask 0xffffff00 arp up

route add default 10.0.0.1 1

So it brings up all necessary modules. Configures the interface and loopback and sets a default route.

I have set the DNSResolvers config to query my router, which it does and resolves names perfectly.

But...........................

There is something wrong with the default route.

If I put a route in for google.co.uk eg.

route add 173.194.75.94 10.0.0.1 then I can ping it, but just leaving it down to the default route I can't.

I think something is messing up the default route.

At the moment there are only 3 entries in the routing table:-

netstat -r

Destination (10) Gateway (10.0.0.119) Flags (UC) Refs (1) Use (0) Interface (ep0)

Destination (loopback) Gateway (localhost) Flags (UC) Refs (0) Use (0) Interface (lo0)

Destination (default) Gateway (10.0.0.1) Flags (UGCS) Refs (0) Use (0) Interface (ep0)

I can't help thinking that the route that gets inserted automatically when I bring up the interface (the top route with Destination (10) which I can't delete) is causing the problem.

Is there something glaringly obvious I am doing wrong here?

When I ping anything relying on the default route I get an error 65 (which is no route to host).

Interestingly though I can ping googles nameservers 8.8.8.8 and 8.8.4.4 which the A3010 should have no knowledge of. A problem at the router maybe? However the router doesn't have any knowledge of googles ns either as far as I am aware! It queries BT.

[spampling]

In article <e23df3a8-d4ba-4c8d...@googlegroups.com>,
MarkYoungIW <marky...@yahoo.com> wrote:
> ifconfig eo0 inet 10.0.0.119 netmask 0xff ff ff 00 arp up

> route add default 10.0.0.1 1

Some older implementations of TCP/IP stacks have a bug that treats address
ranges like 10.x.x.x as /8 mask (255.0.0.0) and similarly treats
192.168.x.x ranges as /24

Try setting your kit with a 192.168.x.x range on the router and matching
addresses on the various devices and use a 255.255.255.0 (/24) mask.

Unfortunately someone in the Acorn fraternity started encouraging people to
use 10.x.x.x as if it was a natural /24 and problems follow from there.

--

Steve Pampling

[MarkYoungIW]

Not looking forward to that. Got a few static devices on my network that will need changing too. But will give it a go.

Netstat -e does bring up:-

ep0: 10.0.0.119 netmask 0xffffff00 so it sort of looks like its coping.

Having a closer look and just pinging random websites, the machine pings www.facebook.com with no problems, and pings www.yahoo.com with no problems. However www.google.com returns a no route and www.bbc.co.uk returns no route. Seems very odd to me.

The stack does seem buggy as when I try and load yahoo.com in ArcWeb sometimes it does, other times the stack seems to freeze and will no longer ping anything.

Perhaps another stack? Although there seems to be little choice.

[MarkYoungIW]

Just got round to changing my home network around a bit (which was good for a cleanout of failed projects on the router).

Changed the router to 192.168.1.1 and the A3010 to 192.168.1.138 and sure enough now I can ping all websites I have tried on the internet, so the routing problem seems to be solved.

The stack does seem to hang alot though, but at least it seems the routing problem was because of the IP.

Thanks Steve for the help with that, I dont think I would have tried that if left to myself as it was a bit of info I had no come across while surfing the web.

Will get on and experiment more, but at least things are progressing......slowly!! :)

[MarkYoungIW]

Well, although a little closer, I am still having trouble with Freenet. I can load the stack up, then ping lots of sites. If I use arcweb to surf webpages it works for 2 or 3 pages, then starts timing out. As soon as that happens, I can no longer ping the router, or anything on the network other than localhost. Pinging the router is causing the router to rythmically flash showing that the machine is sending a ping, but I guess not hearing back.

Everything I query in the stack then seems to take ages to reply (eg Netstat takes much longer to bring back information than before the timeouts started).

So for example from a fresh start:-

I have just tried to access a site on homepages.demon and it got halfway through downloading the page then stopped.

If I don't quit the browser, *netstat returns

(Proto) tcp (Recv-Q) 0 (Send-Q) 0 (Local Address) 192.168.1.139:1025 (Foreign Address) 193.195.70.8www (State) Established

If I quit the browser then *netstat it returns

(Proto) tcp (Recv-Q) 0 (Send-Q) 0 (Local Address) 192.168.1.139:1025 (Foreign Address) 193.195.70.8www (State) FIN_WAIT_1

It retuns FIN_WAIT_1 for a long time (5 mins) after the browser has closed before netstat eventually returns nothing.

I have had a good look around the groups, but struggling to find an answer.

Regards

Mark

[Theo Markettos]

MarkYoungIW <marky...@yahoo.com> wrote:
> Well, although a little closer, I am still having trouble with Freenet. I
> can load the stack up, then ping lots of sites. If I use arcweb to surf
> webpages it works for 2 or 3 pages, then starts timing out. As soon as
> that happens, I can no longer ping the router, or anything on the network
> other than localhost. Pinging the router is causing the router to
> rythmically flash showing that the machine is sending a ping, but I guess
> not hearing back.

If it were Internet 5, I'd suggest looking at MBuf exhaustion. But it
isn't, and there aren't mbufs on other stacks. So I haven't any ideas, really :(

Actually, one just occurred to me. Have you tried reducing the MTU?

Theo

[MarkYoungIW]

Just having a look and I can't actually see how you set the MTU. It is currently set by default to 1500.

Just looking at the debug file and it goes a little something like this

select
in - r0 = 256
r1 = 000b181c -> 0
r2 = 000b17fc -> 0
r3 = 000b17dc -> 0
r4 = 000b17d4
out - r0 = 0
r1 = 000b181c -> 0
r2 = 000b17fc -> 0
r3 = 000b17dc -> 0
r4 = 000b17d4
r4 = 000b1650
select
in - r0 = 256
r1 = 000b1698 -> 0
r2 = 000b1678 -> 0
r3 = 000b17dc -> 0
r4 = 000b1650
out - r0 = 0
r1 = 000b1698 -> 0
r2 = 000b1678 -> 0
r3-[01]P[18]%~ ............(followed by pages and pages of gibberish!)

Not sure if that throws any light. If anyone could advise where I set the MCU I would be grateful.

[MarkYoungIW]

I've been playing for the last few days and to no avail. Freestack v1.00 and v1.01 work much better than 1.02 which was the final DCI-2 version. I cannot seem to change the MTU, no matter what I do. It accepts the command if issued in IFConfig eg. 'ifconfig ep0 inet 192.168.1.139 netmask 0xffffff00 arp mtu 1200 up' but this seems to do nothing as Netstat still reports the MTU to be 1500.

Once the stack seems to have frozen, it actually seems not to have, as it continues to fill the debug file with gibberish so much fairly quickly you do not have enough memory to open it, then no long after the hard drive is full!!

I can't find any documentation on what the log file is showing to help where the error is coming from.

I had another look at KA9Q, but like Theo said, its of little use without a sockets API. Really I could do with finding a way to get Freenet to go for it (I can't find any success stories relating to the PEA and Freenet (although I can't find much about the PEA full stop), or I could do with finding the Acorn Stack (which seems unlikely).

When it works it works great, just a shame it doesn't work for long!!

[MarkYoungIW]

Just looking quickly at the manual at:- http://acorn.chriswhy.co.uk/docs/Atomwide/Atomwide_Pocket_Ethernet.pdf

It gives instructions for installing the software with Acorn's Level4/AUN Suite, Acorn's TCP/IP suite version 2 or Acorn's TCP/IP suite version 1.

So to be fair the manual doesn't look like it had much else in mind.

So, if I can't get it working with Freenet, can I get any of the above?

[Alan Adams]

In message <224f9be7-a5c9-4f75...@googlegroups.com>

MarkYoungIW <marky...@yahoo.com> wrote:

> Not looking forward to that. Got a few static devices on my network
> that will need changing too. But will give it a go.

> Netstat -e does bring up:-

> ep0: 10.0.0.119 netmask 0xffffff00 so it sort of looks like its coping.

> Having a closer look and just pinging random websites, the machine
> pings www.facebook.com with no problems, and pings www.yahoo.com with
> no problems. However www.google.com returns a no route and
> www.bbc.co.uk returns no route. Seems very odd to me.

yahoo.com 87.248.112.181
facebook.com 31.13.72.20
bbc.co.uk 212.58.246.90
google.com 173.194.65.104

The working ones have the most significant bit of the address clear,
the failures have it set.
I suspect somewhere in the software is a signed compare which needs to
be an unsigned compare.
Alternatively a fault in the hardware related to that bit.

> The stack does seem buggy as when I try and load yahoo.com in ArcWeb
> sometimes it does, other times the stack seems to freeze and will no
> longer ping anything.

> Perhaps another stack? Although there seems to be little choice.

--
Alan Adams, from Northamptonshire
al...@adamshome.org.uk
http://www.nckc.org.uk/

[MarkYoungIW]

OK. After a lot of searching around and a lot of help I have found Acorn TCP/IP Release 2 as recommended in the manual on the PEA.

The stack is working with the PEA (although I haven't managed to get DNS Resolver working yet).

Same problem though, after an amount of data, suddenly all stops.

Its not freezing the machine now though, so I can see using *EPInfo that it has skipped an unrecognised frame 0x0006 (If I'm correct thats a Network Device Unregistration!?) If so it sort of sounds like my router is booting the computer out?

Without a restart though I can't get reconnected which is awkward.

It does seem likely now though that the router is the problem. If I continue to try and ping it looks (flashing light on the hub) that its transmitting), but I am not getting a reply.

Right, will keep trying. Any suggestions would be good. Especially the resolver.

Thanks

[MarkYoungIW]

Right, I've tried all sorts.

I've tried my best to make sure it wasn't a problem somewhere along the network (as I was using Homeplugs as the Acorn was nowhere near the router and I am guessing these are some sort of repeater really), but plugging the Acorn directly into the router has made no difference.

I have tried pinging from both ends (acorn and router) and both times it transfers around 3mb of pings before going wrong.

Whatever I try I cannot manage to reconnect without restarting the Acorn. I have killed, restarted and reconfigured all associated modules but no joy.

So now I guess I will look at the packets using something such as wireshark to see if it sheds any light as to what is happening.

I sort of think the router is giving some response that the Acorn stack cannot deal with and so the connection dies.

I notice in these groups that there are some posts with identical symptoms to what I am experiencing, but no solutions, such as:-

https://groups.google.com/forum/?hl=en-GB&fromgroups#!searchin/comp.sys.acorn.networking/Arcweb$20stalls/comp.sys.acorn.networking/8bzIv5ZkwLo/7LV7Nl9imkwJ

I will post the findings of wireshark!!

[Alan Adams]

In message <4ec207a9-c75e-4ac8...@googlegroups.com>

A wild guess is that the router is usimg a more up to date version of
some protocol, which DCI2 pre-dates.

[MarkYoungIW]

I would agree with that. I will plug it all back in later, and see what the final frames are to see if I can work it out is getting sent that brings it all crashing down. The Router is running DD-WRT, so fingers crossed I may be able to persuade it not to send out whatever is upsetting the Acorn.

Whatever it is, the router only transfers it once in a while though!

Its so annoying where things start to work, and just as I get hopeful, they stop!!

[MarkYoungIW]

Definately has me stumped!! Wireshark shows nothing unusual! If I put either the acorn or the router to ping the other it records perfectly normal pings, until ones stops replying!

Once one stops replying the requests on the wire are still perfectly formed, but do not get a reply.

I can ping from the acorn (all showing as perfectly formed frames), but do not get a reply. I can then ping from the router (all showing as perfectly formed frames again), but no reply.

For some reason the 2 just stop talking to each other.

The only error I can see is immediately after communications halt, *epinfo which accesses info from the PEA, now shows (recieve errors = 1) & (recieve: something spurious = 1)

This error which the acorn is detecting is obviously what is halting the process, but whatever it is, Wireshark doesn't pick it up.

Watching the lights on the router, I see no collisions happening either.

Does anyone know something better than Wireshark? Or perhaps some way of finding out why the PEA stops on the very first recieve error it has. I am sure older networks must have been affected by more errors than modern ones!!

[MarkYoungIW]

Working!!!!!!!!!!!!!!!!! I think!!

Having tried everything I stumbled on this page on the internet:-

http://acorn.riscos.com/documents/appnotes/261-275/264.ps

Under Ethernet III card configuration it gives a number of options and even how to set them for the PEA.

Using *Set EtherP$Options "oldinet" which I know is meant to be of TCP/IP Release 1, suddenly data continues tranfering even once there has been a spurious packet recieved.

I am going to test this a bit and then try configuring to "newinet" too in case for some reason not setting this option is what the problem is.

Does anyone know what oldinet and newinet "actually do"!?

[MarkYoungIW]

Success was short lived.

It worked once, and not since.

I still haven't managed to work out how to set the MTU to see if that makes a difference.

I haven't managed to capture the spurious frame either that seems to be causing the connection to stop.

Its awkward to work out why the connection stops too as wireshark still shows both interfaces putting packets on the wire, and they look fine, its just neither interface respond to the other.

I've tried a crossover to my laptop too and its exactly the same as the router. It works for a bit, then packs up.

[Alan Adams]

In message <fd668164-bc9b-4287...@googlegroups.com>

MarkYoungIW <marky...@yahoo.com> wrote:

> Success was short lived.

> It worked once, and not since.

> I still haven't managed to work out how to set the MTU to see if that
> makes a difference.

> I haven't managed to capture the spurious frame either that seems to
> be causing the connection to stop.

How are you connecting wireshark to the A3020 and the router? If it
isn't a multi-port repeater (aka hub) then packets will be filtered.

(The other way is using a managed switch set to repeat packets to a
specified port. Not something which most people have access to.)

> Its awkward to work out why the connection stops too as wireshark
> still shows both interfaces putting packets on the wire, and they look
> fine, its just neither interface respond to the other.

> I've tried a crossover to my laptop too and its exactly the same as
> the router. It works for a bit, then packs up.

I'm wondering whether there is some form of flow control signal, that
the adapter/DCI stack doesn't understand.

[MarkYoungIW]

Connected it by putting a hub in between the A3010 and the router and connecting the PC running wireshark into that. Thought that was about right?

I am heading towards perhaps finding a router/bridge (something that regenerates the packets rather than just repeat them) with a 10B2 connection, but struggling to find one, as I am guessing whatever the signal is, is just being passed on by the hub, whereas a router with a 10B2 connection is probably going to be more likely to work.

I was going to try wireshark directly from the router using a method such as this:-

http://www.question-defense.com/2010/02/04/use-wireshark-and-dd-wrt-router-firmware-to-imitate-port-monitoring-on-a-router-switch-port

It is interesting that the now 5 stacks I have tried have differing levels of sucess.

Acorn 1.05 (The Best)
Acorn 2.01 (Second Best)
FreeNet 1.0 (Third Best)
FreeNet 1.01 (Forth Best)
Freenet 2.00 (Worse)

Its seems the older the stack, the better it runs, but eventually they all stall.

[Alan Adams]

In message <348b6d3f-bdaf-430f...@googlegroups.com>

MarkYoungIW <marky...@yahoo.com> wrote:

> Connected it by putting a hub in between the A3010 and the router and
> connecting the PC running wireshark into that. Thought that was about
> right?

That's exactly right. A hub is dumb, and copies everything everywhere.
It does it by replicating the bit pattern, making no effort to
understand anything.

> I am heading towards perhaps finding a router/bridge (something that
> regenerates the packets rather than just repeat them) with a 10B2
> connection, but struggling to find one, as I am guessing whatever the
> signal is, is just being passed on by the hub, whereas a router with a
> 10B2 connection is probably going to be more likely to work.

I have never seen anything with a 10B2 connection that wasn't 10MBit
only. I would suggest connecting your hub to a 10/100 switch, with the
other units connected to the switch. The switch will isolate the PEA
from unnecessary traffic, and you will still be able to monitor
traffic from one of the hub ports.

> I was going to try wireshark directly from the router using a method
> such as this:-

> http://www.question-defense.com/2010/02/04/use-wireshark-and-dd-wrt-ro
> uter-firmware-to-imitate-port-monitoring-on-a-router-switch-port

This technique limits you to TCP/IP packets. I wonder whether your
problem packet is an ICMP one, which you might be missing depending on
your capture settings. Are you catching everything, or limiting it by
IP address? You may need to include broadcast and multicast packets.

> It is interesting that the now 5 stacks I have tried have differing
> levels of sucess.

> Acorn 1.05 (The Best)
> Acorn 2.01 (Second Best)
> FreeNet 1.0 (Third Best)
> FreeNet 1.01 (Forth Best)
> Freenet 2.00 (Worse)

> Its seems the older the stack, the better it runs, but eventually they
> all stall.

[Theo Markettos]

MarkYoungIW <marky...@yahoo.com> wrote:
> Connected it by putting a hub in between the A3010 and the router and connecting the PC running wireshark into that. Thought that was about right?
>
> I am heading towards perhaps finding a router/bridge (something that
> regenerates the packets rather than just repeat them) with a 10B2
> connection, but struggling to find one, as I am guessing whatever the
> signal is, is just being passed on by the hub, whereas a router with a
> 10B2 connection is probably going to be more likely to work.

Can you try a 10base2 network card in a PC (which runs bridging or NAT or
something)? I think you could get 10base2 PCI cards if you have a PC with
that.

Theo

[Alan Adams]

In message <8hm*3D...@news.chiark.greenend.org.uk>

I've probably got one if you need one.

Alan

[MarkYoungIW]

OK. I have a switch on the way anyway as always useful to have one even if it doesn't work.

Going to try a few more things with Wireshark as although I had it monitoring the MAC address, to be fair I am fairly sure it only showed TCP packets, so I will have another look.

Something that is very very odd (which the switch will fix), is I plugged wireshark directly into my DD-WRT Router (which is set up as a wireless repeater), and the LAN ports are obviously just repeating too as I could see packets off the remote wireless network, let alone my own! This is obviously not going to be helping having that much traffic on the cable to the Acorn.

So I will try the switch, then if not as Theo has said a 10B2 interface of some sort (I have now spotted a couple of switches/managed hubs with both 10bT and 10b2 network connections, so there are still some options.

I have one ancient PC which I can put a 10b2 and 10bT card in and use it that way as a last resort.

Would be great to find out what throws it off though, so I'll have another go at Wireshark later and see if I can do better than just TCP packets.

[MarkYoungIW]

I think I see the problem. Another go with Wireshark and this is what I am seeing!

Everytime the router issues and ARP request the Acorn replies, but according to Wireshark the ARP reply has an invalid checksum.

The router does it again and again and again, then suddenly the acorn stops replying to the arp requests. The router queries 3 times, doesn't recieve a reply to the request 3 times, then the connection goes down!

The Frame Check Sequence that the Acorn is sending is 0x156bc765 instead of 0x92b77511.

[Alan Adams]

In message <476fa119-0f60-4cbd...@googlegroups.com>

It looks as though both the router and wireshark agree that the
checksum is invalid, so it probably is.

I wonder whether the ARP protocol exists in different versions?

Alternatively the checksum is done in hardware, and the hardware is
faulty.

[MarkYoungIW]

Tested, tested and tested again!! Reading a few bits on the net I'm not sure that the ARP's FCS is actually wrong. It just sounds like Wireshark isn't very good at that.

Every single time the connection stalls (at varying lengths of time into the connection), suddenly the Acorn doesn't reply to an ARP request.

If I am pinging at the time for example you get is:-

Ping Request
Ping Reply
Arp Request (who has 192.168.1.2 tell 192.168.1.1)
Arp Reply (192.168.1.2 is at Acorn Mac)
Ping Request
Ping Reply
.
.
.
.
.
Ping Request
Ping Reply
Ping Request
Ping Reply
Ping Request
Ping Reply
Arp Request (who has 192.168.1.2)
Ping Request
Ping Reply
Arp Request (who has 192.168.1.2)
Ping Request
Ping Reply
Arp Request (who has 192.168.1.2)
Ping Request
Ping Request
Ping Request
Ping Request
Ping Request
Ping Request

[MarkYoungIW]

http://noahdavids.org/traces/wiresharks_reports_bad_ARP_FCS.html

[MarkYoungIW]

OK. If I *IFConfig ep0 -arp to turn arp off I get the same result immediately. The router sends and arp request 3 times and then everything falls down, so the problem is definately that the Acorn suddenly stops responding to ARP requests.

I wonder if I can turn ARP off on the Acorn and perhaps create static link layer to IP mappings, or something along those lines.

Would be interesting to know why ARP fails on the Acorn though.

[MarkYoungIW]

OK, just looking a bit further into this, once the router has issued the 3 ARP request requesting who has 192.168.1.2 (it issues these 3 to the Acorn's MAC address). it then continues to issue ARP requests for quite some time afterwards asking who has 192.168.1.2 to the broadcast address of FF:FF:FF:FF:FF:FF

I am not sure whether the Acorn would recognise this broadcast address?

I can only set a IP Broadcast address of 192.168.1.255 but this obviously isn't going to help here. We can only make an assumption that the Acorn (PEA) is using a hardware broadcast address as well?

I am half wondering whether the 3 final ARP's issued before things go wrong are during a slow point as the Acorn does seem to struggle with regulating speed (one minute fast, then slow). Not sure if there is a way of adjusting how long an interface waits for an ARP reply.

This is a strange fault though, almost like a hardware fault, but odd how its affected by which stack I use if thats the case. I guess the older stacks are running faster which is perhaps why the connection holds together for longer.

[MarkYoungIW]

Actually not sure what I was thinking there. Of course the Acorn recognises FFFFFFFFFFFF, otherwise it wouldn't be able to use ARP in the first place.

It does just seem that ARP packs up on the Acorn after a while, causing the connection to cease.

Only thing I can think is to try and create some static mappings and see if the connection can hold together then.

[Alan Adams]

In message <2dd6a8f0-7585-4fec...@googlegroups.com>

It would be interesting to know whether the lack of response is
because the router has switched to broadcast, or whether the initial
failure to reply to 3 directed messages indicates a failure which also
causes the lack of response to the subsequent broadcasts.

If you restart the router after the failures, I would expect it to
start using directed messages again - in which case the PEA might
start to respond again.

[MarkYoungIW]

OK. I'm possibly getting ahead of myself again here, but it hasn't broken yet!! I've turned arp off on the Acorn and set a static ARP entry on the router (arp -i br0 -s 192.168.1.2 Ac:or:nM:AC:Ad:dr) and so far I can't break it!!

I haven't really managed to answered why the Acorn stops replying to ARP requests after a while from the router or directly to my laptop, but does seem to offer a workaround for me to use the Acorn on my network by turning off ARP and relying on a Static ARP entry.

[MarkYoungIW]

I have tried switching the router off and on, and it doesn't get the PEA Running again. The PEA Fails as soon as it gets the Spurious Packet error (which I cannot for the life of me work out what this spurious packet is), but without ARP running this spurious packet never shows up and the network does not crash.

I am guessing ARP is failing in the Acorn Stack, but not sure why.

[druck]

On 04/12/2012 13:58, Alan Adams wrote:
> It looks as though both the router and wireshark agree that the
> checksum is invalid, so it probably is.
>
> I wonder whether the ARP protocol exists in different versions?

Back in the DCI2 days machines were so underpowered many IP stack
implementations didn't check checksums in headers, and some didn't even
bother generating valid ones.

---druck

[MarkYoungIW]

After a very long time I have managed to get back to this, and even after disabling ARP and setting static ARP it does still drop out.

Wireshark doesn't really shed any light (other than the router or other station suddenly sending 3 ARP requests and then the connection dropping).

The only sign of life I can see once the stations have stopped exchanging packets is if the Acorn is still pinging, using *EPInfo the transmit interrupt's continue to increase (by 10 every ping) but nothing else does and if the other station/router is pinging the Acorn, the recieve interrupt's continue to increse, but again nothing else.

Turning the router/other station off and on does not cause the connection to resume, only however turning the Acorn off and back on does, or the following also re-establishes the connection:-

*ifconfig ep0 down
*rmkill internet
*rmkill etherp

Then reloading EtherP, then Internet and then issuing

*ifconfig ep0 inet 192.168.1.2 netmask 0xffffff00 up

I can only see once that the acorn during its connection ever issues and arp request for 192.168.1.1 (the router) and that is on *ifconfig ep0 up

Once the connection has ceased I have tried *ifconfig ep down then up to see if this would resume the connection, but all that happens is the machine freezes.

I cannot see using wireshark what is upsetting the acorn, all I can see is the connection ceases as soon as the spurious packet is recieved (if I have *epinfo running in a loop in a task window and the ping in another you can see this is the point where everything halts).

Is there a packet sniffer that I can run on the Acorn to perhaps try and see what this spurious packet that Wireshark appears not to see is?

Is there a way of getting the Acorn to manually issue an Arp request to the router once in a while to see if this prolongs the connection?

It would be so nice to find the route of the problem before the Acorn ends up back in its box again for a couple of months!! :)

Thanks again for all the help.

[Evan Clark]

In article <51bf9f79-8e5d-40b1...@googlegroups.com>,

MarkYoungIW <marky...@yahoo.com> wrote:
> Is there a packet sniffer that I can run on the Acorn to perhaps
> try and see what this spurious packet that Wireshark appears not to
> see is?

Try Wiresalmon:

<http://www.cp15.org/networking/>

Evan.

[SG nws]

Evan Clark wrote:

> MarkYoungIW wrote:
>> Is there a packet sniffer that I can run on the Acorn to perhaps try and
>> see what this spurious packet that Wireshark appears not to see is?

> Try Wiresalmon: http://www.cp15.org/networking/

Failing that you could try:
Wireflounder
Wireplaice
Wirecod
Wirestickleback
Wirehaddock
Wirebaracuda
...

Forgive me. :-)

--
Stewart Goldwater
http://janusg.co.nr

[MarkYoungIW]

Wiresalmon is unfortuantely RO3.5 and my poor old A3010 is only RO3.1.

I have tried to get the PEA running on a RPC600 but no such luck. It sits there quietly even with the right stack doing nothing.

Something is still going on after the connection fails though as like I say the transmit/recieve interrupt to continue to increase if there is anything on the wire.

I know there was one called !Sniffer which was DCI4, but unfortunately the PEA is DCI2 only.

I wonder if perhaps there is something better for wireshark for this, or a better way of using wireshark? I have been watching all packets with it, and there are no obvious problem packets, so I wonder if this 'spurious' packet the Acorn is reporting is perhaps not considered by wireshark (or the network card in my laptop), as a packet.

[Alan Adams]

In message <51bf9f79-8e5d-40b1...@googlegroups.com>
MarkYoungIW <marky...@yahoo.com> wrote:
<snip>

> Is there a packet sniffer that I can run on the Acorn to perhaps try
> and see what this spurious packet that Wireshark appears not to see
> is?

I would be surprised if Wireshark couldn't see it. however, if you are
telling it to capture packets to and from the IP address, I could
understand it. Not all packets go to IP addresses. It could be an
administrative packet sent to the hardware address for instance.

> Is there a way of getting the Acorn to manually issue an Arp request
> to the router once in a while to see if this prolongs the connection?

arp -d will delete a hostname.
arp -a will show the table.

If you delete the router entry, then ping the router, you should see
an arp dialog taking place.

> It would be so nice to find the route of the problem before the Acorn
> ends up back in its box again for a couple of months!! :)

> Thanks again for all the help.

[Alan Adams]

In message <a2f8eceb-83d0-43ed...@googlegroups.com>

How are you connecting wireshark? Remember that switched filter
packets - they only send packets on the port where they know the
destination to be. They only broadcast packets when they haven't seen
the destination machine before.

The old 10M hubs are useful in this sort of situation, as they are
more correctly called multi-port repeaters. They send every packet out
on every port.

The other way requires a high-end manageable switch, with the facility
to mirror traffic from one port to another. In this case you would
mirror traffic from the port with the RISC OS system to the port with
the PC running Wireshark.

[druck]

On 5 Mar 2013 MarkYoungIW <marky...@yahoo.com> wrote:
> Is there a packet sniffer that I can run on the Acorn to perhaps try
> and see what this spurious packet that Wireshark appears not to see
> is?

Is this the DCI2 stack thing again? If so I suspect that any sniffers
wont work, !Wiresalmon is DCI4, as is !Sniffer by the look of it.

---druck

--
The ARM Club Free Software - http://www.armclub.org.uk/free/
32 bit Conversions Page - http://www.armclub.org.uk/32bit/

[MarkYoungIW]

The router port is working as a 10M switch (100M and Auto are disabled) which in turn is connected to a 10B2/10BT Netgear hub which both the Acorn (PEA 10B2) and the Laptop (10BT) running Wireshark are connected to.

Wireshark is showing all traffic, no filters set, so in theory it is showing all traffic.

Nothing looks out of place as far as I can see, and no obvious reason why the connection suddenly drops.

Unfortuantely the only driver for the PEA is DCI2 so it does somewhat limit options.

[MarkYoungIW]

Probably the best way to show what is going on is to post the capture file.

It seems like all goes well until eventually the router issues and arp request that the acorn doesn't reply to. The router tries a couple more times, with no reply, at which point I am guessing the router shuts the connection, and does not reopen it until things are reinitiated.

I have watched a few devices and they all seem to manage arp in this way, so I am not sure why the acorn is giving up responding.

I did try static arp routes, so I will give these another go as I think something must have still been a little wrong as the router was still issuing requests regardless and the acorn was still answering them regardless for a while.

http://www50.zippyshare.com/v/75922761/file.html

[MarkYoungIW]

Hmmmm.............Playing a copy of the arp packets from the Acorn using Colasoft Packet Player over the network has stopped the router constantly asking where the Acorn is, but the connection still ceases. This time it does just halt. Watching wireshark the pinging station continues pinging, but the other stops replying. Might continue throwing a few packets onto the network to see if I can get things back up, but I still cannot see this spurious packet the Acorn reports. The illusive packet that halts everything!

Going to fiddle a bit more later, but its looking like ARP only stops as the connection already has!

[Theo Markettos]

MarkYoungIW <marky...@yahoo.com> wrote:
> Hmmmm.............Playing a copy of the arp packets from the Acorn using
> Colasoft Packet Player over the network has stopped the router constantly
> asking where the Acorn is, but the connection still ceases. This time it
> does just halt. Watching wireshark the pinging station continues pinging,
> but the other stops replying. Might continue throwing a few packets onto
> the network to see if I can get things back up, but I still cannot see
> this spurious packet the Acorn reports. The illusive packet that halts
> everything!

Since you lack a packet sniffer for DCI2, have you tried looking at the
stats emitted by either *epinfo or *ifconfig? That should at least count
packets - will tell you how many the Acorn thinks are going through.

Other than that I think you're going to have to poke about with the packet
handler routines - on DCI4 these are reported by *eXinfo and you could write
some intercept code. I've never seen nor read documentation for DCI2 so I
don't know how it behaves.

Theo

[MarkYoungIW]

I've spent most of the day playing packets onto the network and here is the result.

For all of these tests I have used constant pinging as data.

If I turn arp off on both the router and the Acorn and set static arps the network runs for a bit then stops. However the Acorn continues to send out pings which the router recieves and replies to, but the acorn doesn't hear the reply (although the recieve interupt reported by *epinfo does continue to rise).

If I inject correct arp packets onto the network informing the router of the Acorns hardware address exactly the same as above. Acorn keeps pinging, router keeps replying, acorn doesn't hear reply but receive interupts increase.

All other bits and pieces I have tried result in the same. Eventually the Acorn stops receiving packets, so it doesn't appear to be an arp problem, it just seems the Acorn suddenly stops recieving packets, but does continue to transmit.

The only way of getting the link back up is to either Ctrl-Brk the Acorn or to *ifconfig ep0 down, *rmkill internet *rmkill etherp and then load them back up and reconfigure them. Nothing else I have tried re-establishes the Acorns ability to continue recieving.

The spurious packet does not show through wireshark, so all I can assume is this happens when the Acorn stops listening (perhaps gets half way through recieving a packet then stops).

At least I am fairly sure I can rule out the router side being a problem, and the Acorns transmit side continues working. It is definately the acorn suddenly ceases recieving packets, and only registers recieve interupts.

Something I have found is the bad checksum on the arp reply only happens with Acorns stack, Freenet doesn't suffer from it so I guess slightly better programming in that respect!

[MarkYoungIW]

It would be good to know where the recieving packets are falling short as when the connection fails *epinfo is still showing the receive interupts but they are obviously not being read as recieved packets.

[MarkYoungIW]

I've been trawling through the internet again and other old documents again and I have come across a few other people who have described these symptoms, and tracked it down to the stack enabling ICQ's whilst the receive frame handler is executing. Apparently the frame handler is not re-entrant, so this means the end of receiving anything.

I find it unlikely this is the problem as the I am using the PEA with the stack recommended by Atomwide, but in the same breath, maybe the more active modern network is enough to cause issues. More back to back packets?

[Theo Markettos]

MarkYoungIW <marky...@yahoo.com> wrote:
> I find it unlikely this is the problem as the I am using the PEA with the
> stack recommended by Atomwide, but in the same breath, maybe the more
> active modern network is enough to cause issues. More back to back
> packets?

Maybe there is some tool that will slow down packets from the other end?
I'm not sure how you would do that without dropping packets, which is
specifically what you want to avoid. Maybe some evil bridging of ethernet
to PPP over serial or something? Or generate tons of collisions. 10Mbit/s
is quite fast for a parallel port and it might be being overloaded.

Pity you can't do 2.94Mbit/s ethernet any more...

Theo

[MarkYoungIW]

My router is running DD-WRT and under QOS it does allow you to select the rate of the Ethernet ports. I have just tried 5M, 2M then 256k, but unfortuantely the same problem arises. Assuming that the function actually works in DD-WRT

I am starting to once again wonder if its a hardware fault with the PEA, but if that is, its a pretty strange one. Freenet does allow you to save a large debugging file, but I have no idea what its contents means.

I wonder if there would be any merit in monitoring whats going on over the parallel port.

[MarkYoungIW]

Just checked and my router has a chipset that doesn't allow the rate to be changed on the ethernet port, so I will have a look at other ways of doing this in the morning. It would certainly seem worth persuing!!

[MarkYoungIW]

Thinking aloud again I could generate and inject a pause frame loop to slow the transmission from the router to the Acorn's address down.

Fairly sure this can be done asymmetrically as I don't think the Acorn would have a clue what the frame was if the router sent one back!!

[MarkYoungIW]

OK, doing the above doesn't actually exactly slow the rate down, but its traffic does help add congestion slowing the rate to the Acorn. As a result the interface is running for a much greater number of pings before eventually suffering the spurious packet fate.

Although I have had a number of occasions that I have thought I was on the right lines here, once again reducing the speed can only help. Now I just have to work out how I am going to do this properly!!

[MarkYoungIW]

Well, very interesting couple of days, but still really no exact progress.

Using Linux's Classful Queueing Disciplines I have managed to control the rate that the router transmits, and even with the connection to the Acorn down to a crawl it still gives up the same as if the network was at the full 10Mbits.

I have tried different speeds and there is no real conclusive proof to suggest any speeds have any benefit.

Things I have found is sometimes using Freenet when the Acorn stops listening if I rmreinit etherP I can reconfigure and get things back up, so it seems like the driver is perhaps more at fault than the stack.

Interestingly I can receive a huge amount of packets before it halts using Freenet's -f (Flood) while pinging. Many more packets than normal pinging.

When it quits is hit and miss. Sometimes it will run for half an hour, sometimes a couple of minutes.

I wonder if perhaps the Acorn is waiting for a frame that modern equipment doesnt send?

[MarkYoungIW]

Well I am at a total loss again.

Here's how it is:-

The Acorn just attached to the hub will ping itself on its IP address all night, and with wireshark attached to that hub I can see the packets are actually going through the hub, so just talking to itself the acorn is fine.

Pinging the router or and of my PCs the acorn will eventually stop listening.

Leaving the acorn attached to the router all night not doing anything the acorn is still fine in the morning, but will stop listening again when pinging.

Examining the last packet recieved by the Acorn and the ones before that and immediately after nothing is wrong with them. Replaying those exact packets onto the network does not cause the acorn to stop listening, so this indicates there is nothing wrong with the packets.

Question is if the packets are all correct as they appear to be, what on earth is causing the problem. I have slowed the rate of the network to a crawl which makes no difference. I have changed the networks MTU down just in case, but no change. Can there be something that wireshark does not see? Is there some other signal on the cable that isn't a packet that the router introduces when transmitting?

I have left the machine flooding pings to itself again to makes sure I can't get it to fail when its talking to itself, but at the moment I am coming up blank. I have even changed the hub in case that made a difference and tried another router, but nothing.

[Theo Markettos]

MarkYoungIW <marky...@yahoo.com> wrote:
> Well I am at a total loss again.
>
> Here's how it is:-
>
> The Acorn just attached to the hub will ping itself on its IP address all
> night, and with wireshark attached to that hub I can see the packets are
> actually going through the hub, so just talking to itself the acorn is
> fine.

Self-pings don't always go out on the network, because not all interfaces
can receive and transmit at the same time. That they are doesn't mean the
received pings are coming from the network itself, rather than being
short-circuited at a higher level.

> Pinging the router or and of my PCs the acorn will eventually stop listening.

What happens if you ping the router/etc at the same time as pinging
yourself?

> Question is if the packets are all correct as they appear to be, what on
> earth is causing the problem. I have slowed the rate of the network to a
> crawl which makes no difference. I have changed the networks MTU down
> just in case, but no change. Can there be something that wireshark does
> not see? Is there some other signal on the cable that isn't a packet that
> the router introduces when transmitting?

There is the influence of collisions. Does your hub have a light for those?

Wireshark might also not see frames whose ethernet (LLC) CRC is wrong. You
may need to poke the (Linux?) driver to prevent it discarding such frames
(not sure if promiscuous mode enables that or you need to do explicitly, it
will depend on the driver).
http://stackoverflow.com/questions/1867992/ethernet-checksum-checking-in-wireless-stack

Theo

[MarkYoungIW]

> Self-pings don't always go out on the network, because not all interfaces
>
> can receive and transmit at the same time. That they are doesn't mean the
>
> received pings are coming from the network itself, rather than being
>
> short-circuited at a higher level.

I kind of assumed as I could read the packets on the hub with Wireshark that they were going out and coming back in.

> What happens if you ping the router/etc at the same time as pinging
>
> yourself?

It gives up. Any significant amount of packets going into the Acorn, even if from the Acorn seems to give up shortly after plugging the hub into the router or another computer.

> Wireshark might also not see frames whose ethernet (LLC) CRC is wrong. You
>
> may need to poke the (Linux?) driver to prevent it discarding such frames
>
> (not sure if promiscuous mode enables that or you need to do explicitly, it
>
> will depend on the driver).
>
> http://stackoverflow.com/questions/1867992/ethernet-checksum-checking-in-wireless-stack
>
>
>
> Theo

Thanks for this. I will have a good dig around and see what I can do there. I haven't seen a frame in Wireshark with a bad CRC, so maybe it does ignore them. I will do some research. Nothing worse than not being able to solve something!!

In the last hour I have obtained an amount of equipment from my Dad's loft, and I have found a D-Link DI-524 which as long as the Acorn is the only thing plugged into it, it is happy to exchange frames at rate with no errors (been on flood ping with it for an hour now and not 1 reply missed). The moment anything is attached to the routers switch, within a few minutes the Acorn once again stops listening!! Shame the DI-524 has very little you can do with it though!! I guess a little sucess though, it is proving its something in modern devices that the Acorn and PEA have no clue what to do with!!

[druck]

On 10 Mar 2013 Theo Markettos <theom...@chiark.greenend.org.uk>
wrote:

> There is the influence of collisions. Does your hub have a light for those?

My money would be on a collision causing the stack to get terminally
confused. These tend to be very difficult to spot without an ethernet
analyser.

[MarkYoungIW]

One of the hubs I have tried using does have a collision light, but the only time I have seen it come on is when the hub is turned on, and once when the network went really made while I was playing around trying to change the rate.

epinfo is meant to report collsions, and so far that is one thing it hasn't reported (I have even had dribble errors, packet too short and buffer errors reported while trying to solve this).

Not getting much further with the DLink router yet. Doesn't matter what I do, the moment I plug anything else in it goes strange.

If it were collisions, does that seem likely with what I am seeing with the DLink (eg it works fine until something else is plugged in). As the router is regeneratring the signals to fit in with what its already sending, where would things be going wrong?

I was trying to think what else might work like perhaps an Aironet AP or something as they have 10B2 and some of those are of a similar age, but should be able to bridge wirelessly.

[Alan Adams]

In message <fb4d442a...@druck.org.uk>

druck <ne...@druck.org.uk> wrote:

> On 10 Mar 2013 Theo Markettos <theom...@chiark.greenend.org.uk>
> wrote:
>> There is the influence of collisions. Does your hub have a light for those?

> My money would be on a collision causing the stack to get terminally
> confused. These tend to be very difficult to spot without an ethernet
> analyser.

> ---druck

Have you tried with the Acorn plugged into a switch? Since these
effectively filter packets, only transmitting what is needed by the
attached device, it might stop the problem, since it reduces the
likelihood of collisions, and reduces the load on the stack.

If it doesn't then at least you know that the packets causing the
problem are either addressed to the Acorn or are broadcast/multicast
ones.

I don't know what the switch would do with bad CRCs - my guess is that
it would regenerate a good CRC.

[MarkYoungIW]

Yep, tried a switch. Exactly the same. At the moment the only packets Wireshark is showing on the hub (which is seperated from everything with a switch) are frames destined for the Acorn or multicast. None of which seem to cause the problem (at least from what I can see, normally if I am pinging so I can see the sequence of packets and when it quits), the frames around that area don't seem to have any problems.

This one is beyond me. I only have one dedicated switch and the switches inside routers which I have also used as switches, and I had hoped these would solve the problem, but not so far.

I think maybe another brand of switch is in order (as both the dedicated switch and the 2 routers with switches I have are all Linksys).

[MarkYoungIW]

Well I've spent a little more time again on this today......hope to work out the problem within 1 year!! :)

I have tried every device under the sun, begged stolen and borrowed. The only device I had that the PEA/A3010 will keep a continued connection with is a D-Link DI-524.

Monitoring the connections every device I have tried does put an ARP request onto the wire once in a while apart from the DI-524 which only does one at the initial handshake and thats it.

The DI-524 doesn't put any broadcasts onto the wire either, whereas all the other devices have the odd one or two broadcasts once in a while.

I have tried a wireless bridge onto the DI-524 to my internet router, but this then caused the connection to fail so hopefully this rules out something on the wire that isn't a packet.

As soon as anything else is plugged into the DI-524 the odd ARP and Broadcast packets go through, the connection fails not too long after.

It looks like ARP's or Broadcast packets seem to be the only thing that can be bringing things down, as I really cannot see any other difference between the packets on the DI-524 and any other devices packets.

They don't directly bring the connection down though, the connection is often lost a considerable few packets after an ARP or Broadcast.

I am going to have another look through Linux (DD-WRT) and see if there is any way of controlling ARP's and Broadcasts to an interface.

Seems interesting why broadcasts/ARP's would cause it to fail? I can only assume that there we very few of these/none on a network an 1996?

Still very odd. I just hope I can get this sorted as it is so annoying it is happy talking to the DI-524 until anything is plugged in to it!

[MarkYoungIW]

15 Hours later and on my existing router with nothing but the WAN interface up for the internet and a static arp entry on both router and Acorn the connection has stayed up. There has not been a single broadcast or ARP request according to Wireshark in this time.

The Acorn has been flood pinging the router too.

So I think to sum up modern equipment keep requesting ARP updates does seem to cause a problem and so do broadcasts.

To get round this I hope to assign a seperate VLAN to one of the router ports for the Acorn so that I have no outside broadcasts. And I will keep the static ARP entry on the router so that the router does not request ARP updates.

Hopefully this is the solution. Then I will try something better than pinging to soak test it, maybe streaming media through something like DigitalCD as that apparently supports streaming audio in RO3.1?

This is the most stable it appears to have been so far!!

[Chris Evans]

In article <c81e85df-7389-4db4...@googlegroups.com>,

MarkYoungIW <URL:mailto:marky...@yahoo.com> wrote:


> Seems interesting why broadcasts/ARP's would cause it to fail? I can only assume that
> there we very few of these/none on a network an 1996?

Not sure when Access/ShareFS appeared but it uses brodcasts a lot.
I don't recall hearing at the time that they were incompatible with the
Atomwide unit.

Wikipedia says the A3020 was launched just a few months after the A4 in 1992.

The A3020 had Access/ShareFS either at launch or soon after!

Chris Evans

--
CJE Micro's / 4D 'RISC OS Specialists'
Telephone: 01903 523222 Fax: 01903 523679
ch...@cjemicros.co.uk http://www.cjemicros.co.uk/
78 Brighton Road, Worthing, West Sussex, BN11 2EN
The most beautiful thing anyone can wear, is a smile!

[MarkYoungIW]

Very odd. It definately seems to be that though, no ARP's and no Broadcasts and the interface is still running happily at the moment flood pinging.

Struggling to set my VLAN up quite how I want it, but it does definately seem to have solved the issue. Like I say as soon as I have it networked as I want it, I will try DigitalCD streaming something so its getting tested with real data instead of pings, but I am guessing it should all work the same now!

Not used ShareFS or Access before, but once I have it running I will perhaps pop another machine on the same VLAN/Subnet and see if the broadcast packets from that cause any problems.

[Theo Markettos]

MarkYoungIW <marky...@yahoo.com> wrote:
> Very odd. It definately seems to be that though, no ARP's and no
> Broadcasts and the interface is still running happily at the moment flood
> pinging.

Just a thought. The PEA was originally intended for use with an A4, while
you're running it on an A3020 I believe. The A4 has a 24MHz ARM3, while you
have an ARM250. It may be the driver was written on an ARM3 machine and the
slowness of an ARM250 might be a reason for it falling over?

Theo

[MarkYoungIW]

It does seem as long as no broadcasts and not a single ARP goes onto the wire then it doesn't fall over, but interestingly I have left it running constantly pinging www.bbc.co.uk and 250000 pings later its still going, I play one arp onto the wire (copy of one from the acorn), the router replies, 723 pings later the network ceases. Tried this twice now second time 211 pings later.

I can't understand why 1 single ARP could cause a difference (especially where its not immediate). It seems the same with Broadcasts, although I haven't tried a lot of these.

Yes, I'm running it on a A3010 slightly overclocked but still slightly slower than ARM3.

I am going to find something that makes it transfer 'proper' data all day to day and see if thats any different to pings.

[MarkYoungIW]

OK, I've tried to see if the speed of the A3010 is causing the issue, but so far I have nothing to show it is.

It seems to fall over whether I have the machine running at its standard 7.2mips its slightly overclocked 9.47mips or its more overclocked 10.47mips.

I know the A5000 was supposed to run at 13.5mips but not sure about the A4. Was it faster because of a shorter bus, or was it held back by the memory?

I am kind of wondering how fast my A3010 can go, but I've ran out of Crystals!! :)

[druck]

On 01/04/2013 16:07, MarkYoungIW wrote:
> OK, I've tried to see if the speed of the A3010 is causing the issue,
> but so far I have nothing to show it is.
>
> It seems to fall over whether I have the machine running at its
> standard 7.2mips its slightly overclocked 9.47mips or its more
> overclocked 10.47mips.

Even with over clocking you aren't going to get anywhere near the speed
of the 24MHz ARM3 with an ARM250.

> I know the A5000 was supposed to run at 13.5mips but not sure about
> the A4. Was it faster because of a shorter bus, or was it held back
> by the memory?
>
> I am kind of wondering how fast my A3010 can go, but I've ran out of
> Crystals!! :)

The ARM250 did not have any cache, and had to share the memory bus
between the processor and video. In real world, the ARM3 is closer to 4x
faster rather than the 2x the clock speed would suggest. That is in low
resolution modes, and even quicker in high resolution when there is more
contention between processor and video.

It is more than likely an ARM250 cannot keep up with normal traffic on a
10mbit ethernet. Remember pinging is a very ordered one out one back,
and not very indicative of the pseudo random nature of normal traffic.

---druck

[MarkYoungIW]

> It is more than likely an ARM250 cannot keep up with normal traffic on a
>
> 10mbit ethernet. Remember pinging is a very ordered one out one back,
>
> and not very indicative of the pseudo random nature of normal traffic.
>
>
>
> ---druck

Agreed, I can keep the network up with just pings by removing arps and broadcasts, but the minute I try to download a small file for instance it just dies again.

I think perhaps a sensible test is if I get hold of a A5k or A4 (if I can find one), and try it with that to eliminate the speed problem. Trouble is ideally I what I am trying to achieve was Ethernet through any means other than the expansion port on the A3010.

Interestingly enough the connection lasts longer with the speed at the full 10mbits than if I bring it down to 2.9mbits or even 256kbit. Its seems the slower I make the connection through the router, the less it holds out.

As said though, I think an A5k might be the next thing to try. Shame I can't find anyone else who still has a PEA and see what their experience is on modern networks.

[MarkYoungIW]

I've just dug out a 60mhz crystal and the machine is running with it (at 20mhz) Doesn't seem too flakey either (not yet anyway), but still exactly the same ethernet runs for a very short period of time then spurious packet error (dribble error as well now).

I could always fit ARM3 to the Adelaide I guess, not ideal though as they are thin on the ground as it is!!

Looking out for an A5k.

[Dave Higton]

In message <8c9912b4-e413-4786...@googlegroups.com>

MarkYoungIW <marky...@yahoo.com> wrote:

>Looking out for an A5k.

Don't waste your time and money.

Buy some up-to-date RISC OS hardware: a BeagleBoard xM if your budget
will run to it, otherwise a Raspberry Pi.

Far faster, far more capable, far better support.

Dave

[MarkYoungIW]

I would, but the aim is to get the A3010 running Ethernet (and not through the expansion socket as that is running everything else).

Its not really about the money, more about doing it.

I can't help but thinking that in the manual for the PEA it says 'The Atomwide Pocket Ethernet Adaptor (PEA) works on any Acorn 32bit computer with a bi-directional parallel port'. This would sound like they intended it for the A3010/A3020/A4000/A5000/A4 etc!

http://acorn.chriswhy.co.uk/docs/Atomwide/Atomwide_Pocket_Ethernet.pdf

I keep wondering if perhaps the PEA is faulty, but I can't imagine what could be faulty about it as it runs for a while.

Back to the drawing board I think. I might try and get a 10B2 card to go in one of my other A3010's and see if it will talk to that properly to rule out a fault on the PEA.

[Chris Evans]

In article <9c174f395...@my.inbox.com>, Dave Higton

Pandaboards are not a lot more than Beagleboards but have a number of
significant advantages e.g. faster and 1920 x 1080 @ 60Hz!

You can even get them as a 'complete' miniITX cased system:
www.pandaro.co.uk

I suspect though that Mark isn't looking for that type of solution!