Acorn Level 4 Fileserver Issue 3

[philpem]

In my quest for things Econet, I stumbled across a copy of Level 4 Fileserver Issue 2: https://acorn.huininga.nl/pub/disks/ori ... Issue%202/

I also came across mentions of Level 4 Fileserver Issue 3.

Does anyone know if L4FS 3 has been found and archived?

I have InterTalk which I'm currently in the process of archiving - this will include disc images and manual scans.

Cheers
Phil

[BeebMaster]

The version linked is 1.24, and he has 1.00 as Issue 1. I'm running version 1.30, which I downloaded from somewhere or other, so maybe that counts as issue 3.

There's also the Advanced Level 4 file server, I have version 2.10 of that.

[philpem]

The version linked is 1.24, and he has 1.00 as Issue 1. I'm running version 1.30, which I downloaded from somewhere or other, so maybe that counts as issue 3.

There's also the Advanced Level 4 file server, I have version 2.10 of that.

Looks like you might be right. I'm running 1.24 on the A3000.

I don't suppose you might remember where you got 1.30 or Advanced Level 4? I haven't found downloadable copies of those.

Cheers
Phil

[dandoore]

The version linked is 1.24, and he has 1.00 as Issue 1. I'm running version 1.30, which I downloaded from somewhere or other, so maybe that counts as issue 3.

There's also the Advanced Level 4 file server, I have version 2.10 of that.

I'd be interested in both of those as I have L4FS 0.25, 1.00 and 1.24 but not 1.30 (although I found a patch for 'Issue 3' in the archive of Acorn FTP) or the Advanced L4FS one.

Do you have a link or a copy of 1.30/AL4FS you could share?

[BeebMaster]

Right...I've tracked down version 1.30, which I am running on my RISC PC, but I clicked that many !Server applications till I got a different version that I didn't remember where it originated.

It's part of aun.zip but where that actual file came from I don't know, possibly somebody sent it me, but anyway here it is.

aun.zip

(176.77 KiB) Downloaded 102 times

I've temporarily forgotten why I stopped using 1.24 and went to 1.30.

[philpem]

Right...I've tracked down version 1.30, which I am running on my RISC PC, but I clicked that many !Server applications till I got a different version that I didn't remember where it originated.

It's part of aun.zip but where that actual file came from I don't know, possibly somebody sent it me, but anyway here it is.

aun.zip

I've temporarily forgotten why I stopped using 1.24 and went to 1.30.

Nice one -- thanks for that. I'll give it a try on my SARPC later.

I've been having issues where the Filer crashes when L4FS is running - data aborts in RISC OS ROM space. Maybe 1.30 will be a little better behaved!


Edit - Given it a try. The failure mode is different but it still crashes the machine sometimes. What's worse is, the clients can connect to it but can't transfer files. "No reply from Station 254" is fairly common... I get the impression L4FS doesn't get on well with RISC PCs.

[BeebMaster]

Could that be a clock speed or termination problem?

I don't really rate the L4FS at all, ever since I first tried to use it to copy a Domesday Disc in 2006. It's very unreliable with 8 bit stations, but it's extremely handy as "bridge" between my Linux shared drive and Beebs.

[philpem]

Could that be a clock speed or termination problem?

I don't really rate the L4FS at all, ever since I first tried to use it to copy a Domesday Disc in 2006. It's very unreliable with 8 bit stations, but it's extremely handy as "bridge" between my Linux shared drive and Beebs.

I've been testing with awServer and it fails in the same way with that too. I can log into the fileserver, but accessing files gets me a "station 123 is not responding".

I can't see it being a clock or termination issue, the setup is the same as the one I used for the BBC/A3000 test which was absolutely textbook:

Code: Select all

                   [ clockbox ]
                       |   |
T ---[ socketbox ]-----+   +-----[ socketbox ]---- T
         |   |                       |   |
         |   x                       x   |
         |                               |
      A4000                             RISC PC

The clockbox is on the default settings.

The RISC PC introduces RISC OS 4, the PC Card and the Rev J 200MHz StrongARM as variables, so it's possible the Econet driver might be twitchy due to their influence.

I found a StrongARM patch for the Econet module here -- viewtopic.php?f=16&t=8155&start=30 -- though it sounds like it might only be applicable to the DIY Econet Podule. I may give it a try for a lark. I'm currently running Econet module 5.70 (27 Jul 1994).

*Help Station says the RISC PC is Station 123, with a 200kHz clock.

edit - I've just had a Filer crash without having run Level4 (I was using awServer). So it's not a Level4 issue.
edit2 - Installed the Econet driver from the above thread and *Unplugged the one from the card ROM. No difference. Going to try the A3000 as a client, then as a server. Thinking this might be another issue with the A4000.

[BeebMaster]

I think that should work, although with real 1980s socket boxes the Clock is normally plugged in at "X" between two unlinked socket boxes as if it was two stations, rather than as a continuation of the (unlinked) linking cable between boxes, but I don't think that's likely to make a difference. Clock speed sounds OK, I've managed to get mine to 227KHz and working reliably with the RISC-PC.

I'm using an Econet Module Podule and RO 3.70. It's a Strong Arm so I think that patch fixes some SA problems, it may be needed for the Econet NIC as well.

[dandoore]

It's part of aun.zip but where that actual file came from I don't know, possibly somebody sent it me, but anyway here it is.

Awesome, cheers!

What's the deal with Advanced L4FS - is this just a newer version or actually a non-Acorn product all together?

Dan.

[philpem]

I think that should work, although with real 1980s socket boxes the Clock is normally plugged in at "X" between two unlinked socket boxes as if it was two stations, rather than as a continuation of the (unlinked) linking cable between boxes, but I don't think that's likely to make a difference. Clock speed sounds OK, I've managed to get mine to 227KHz and working reliably with the RISC-PC.

I think electrically that amounts to the same thing. Here is one of the socket boxes:

The sockets on the left and right are intended to be the equivalents of the punchdown blocks in the Acorn and SJ boxes. Only this way, I can patch things together with DIN leads.

The cable on the left goes to the Clockbox. The lower two are for stations, in this case the RISC PC and one open socket. The one on the right goes to the next box in line -- or in this case a terminator.

The box on the other end is the same except it's hooked up to the A4000 and A3000.

I'm using an Econet Module Podule and RO 3.70. It's a Strong Arm so I think that patch fixes some SA problems, it may be needed for the Econet NIC as well.

I'm seeing the same "no reply from network station 123" from the A3000. If I run Level 4 v1.24 on the A3000, it'll happily netboot the A4000 and serve applications to it. I get data aborts in ROM space when I shut down Level4 but I think this may be a bug in L4 v1.24.

RISC PC wise, I'm still seeing heap corruption ("Not a Heap block" errors) with the new module.

I think my next step is going to be installing an ARM610 in the RISC PC. After that, it'll be booting off a clean CF card with only the RO4 boot structure, Econet module and Level 4 on it.

edit - Realised I had Level 4 on a 1.6MB floppy. Did *Configure NoBoot, power cycled and started the new Econet module and Level4 from said floppy. Still playing up. It'll send directory listings but asking for file data (especially Iconsprites) causes it to fail. Going to have some lunch then try the ARM610.

Does anyone mind me having this discussion here or would you prefer to take the troubleshooting to 32bit Hardware?

[philpem]

What's the deal with Advanced L4FS - is this just a newer version or actually a non-Acorn product all together?

According to https://heyrick.eu/econet/intro/servers.html it's an enhanced version of Level 4 which was released around 1997. It was developed by Network Solutions.

There's a bit more about it here: http://web.archive.org/web/199902181850 ... atext.html

There's a demo version here: http://web.archive.org/web/199902182046 ... /demo.html

And some old updates here: http://web.archive.org/web/199902220612 ... nload.html

I get the impression this used to be on MDFS.net (there's a Level 4 v2.11 ZIP link there), but the file now 404s: http://mdfs.net/Apps/Networking/FServers/

[BeebMaster]

I'm seeing the same "no reply from network station 123" from the A3000. If I run Level 4 v1.24 on the A3000, it'll happily netboot the A4000 and serve applications to it. I get data aborts in ROM space when I shut down Level4 but I think this may be a bug in L4 v1.24.

RISC PC wise, I'm still seeing heap corruption ("Not a Heap block" errors) with the new module.

I think my next step is going to be installing an ARM610 in the RISC PC. After that, it'll be booting off a clean CF card with only the RO4 boot structure, Econet module and Level 4 on it.

edit - Realised I had Level 4 on a 1.6MB floppy. Did *Configure NoBoot, power cycled and started the new Econet module and Level4 from said floppy. Still playing up. It'll send directory listings but asking for file data (especially Iconsprites) causes it to fail. Going to have some lunch then try the ARM610.

It does seem to be a RISC PC issue then. You can turn on the log file from the L4 icon menu, that might help a bit. Also it might be worth trying in single-tasking mode using !Run2 inside the application ("SA" in there meaning Stand Alone and not Strong ARM!)

[philpem]

It does seem to be a RISC PC issue then. You can turn on the log file from the L4 icon menu, that might help a bit. Also it might be worth trying in single-tasking mode using !Run2 inside the application ("SA" in there meaning Stand Alone and not Strong ARM!)

Sadly running it single-tasking doesn't help either

I've also tried --

• Disabling the ViewFinder card -- no effect
• Switching to an ARM610 processor (Viewfinder on and off) -- no effect
• Using the RISC PC as a client for the Econet, with Level 4 on the A3000 -- works fine.
• Using awServer on the RISC PC instead of Level 4 -- same issue.

If the RISC PC is the server, I can log on, *Cat, anything handled by executing a command on the fileserver.
When I try to write a new file, that (usually) works too.
If I try to read anything (and that includes the Filer going through the Filer_Boot/Iconsprites process for apps in a directory) then it'll throw "Station not responding".

I've had a play with the Advanced Level 4 demo and it does the same thing

[philpem]

It looks like there's an update for L4FS 1.30 to bring it to version 1.33. It used to be on the Acorn FTP site, which is now on 4corn:
http://www.4corn.co.uk/aview.php?sPath= ... ing/level4

The README says it's to upgrade L4FS Issue 3 -- if you try to upgrade v1.24, it crashes with an Address Exception on start up.

So it appears the versioning scheme is --

• Less than v1.20: Issue 1
• V1.20 to V1.29: Issue 2
• V1.30+: Issue 3

[richw]

Back in 1994/95 I remember my dad trying to run Level 4 on his Risc PC with Econet. It didn't work. Acorn said they needed to modify the Econet board to make it work 'as a server'. It was returned and then worked fine. Don't know what they did to it. Carl Elkins of Acorn was suggesting that we didn't bother, as a Risc PC would be no better than an A5000. In the end, though, he arranged for the card to be modified.

So it worked as an Econet client, but not as a server until the mod.

[BeebMaster]

That makes it sound like there was a bug preventing the RISC PC receiving transmissions on the file server port. I wonder if it was a hardware fault or a bug in the ROM code.

[richw]

Makes sense. I would have thought it wasn't a simp!e ROM issue, as you would think that could be resolved with a soft load driver/patch (and Risc PCs all had hard discs). I suspect a hardware fault.

[philpem]

I've been doing some digging into this on Google Groups. I found a post from 1997 about having to hassle Mr Elkins about the bug (https://groups.google.com/d/msg/comp.sy ... GzbTDl5AkJ) then I noticed who posted it -- Richard Walker!
A later thread in 2002 (https://groups.google.com/d/msg/comp.sy ... IDCIMooqQJ) describes the same issue in a bit more detail, but no detail on the hardware mods. *Cat and writing files to the RPC works, but reading gets you a "no reply" or "not responding".

I'm curious what might cause this issue. There isn't a lot on the card - address decoder/cycle-stretcher PAL, EPROM, 16MHz clock and the usual Econet ADLC+line driver circuitry. It'd be very interesting to find a modified card and take a look at it.
I wonder if it's some kind of interrupt or timing issue as putting the Advanced Level 4 demo into Trace gets it to at least send a couple of files.

RichW -- do you still have yours? Could you possibly take some photos of the board?


Also speaking of ROMs -- there's apparently a newer, StrongARM-compatible version of the Econet and BBCEconet modules in the RO3.71 ROM. That said there's also a patched and bugfixed version elsewhere on Stardot...

[acorndave]

Carl Elkins hangs out on Github...

https://github.com/rce1086

You never know he might be able to help

[BeebMaster]

Not sure if this is a clue but Chris Whytehead's card is unmodified:



whereas mine has a bodge, a replacement IC4 in a socket, and an earlier serial number:



It's the same part in both cases - 26LS30 - maybe early production run boards originally used a different part, or a duff batch of chips?

Although, the line driver chips are prone to blowing, so it could just be a coincidence that mine has had its replaced.

[philpem]

Although, the line driver chips are prone to blowing, so it could just be a coincidence that mine has had its replaced.

I think there are a few possibilities here -- your card was updated and the bodge is under the line driver, or the card isn't modified at all and the line driver had failed and been replaced at some point.

Mine has a 26LS30 line driver too, soldered onto the PCB. That's not too surprising (To me anyway) as it was brand-new in box.

Serial number wise, mine is 72-AEH60-1000394. Yours is -- correct me if I'm wrong -- 72-AEH60-1000362. Chris Whytehead's is 72-AEH60-1000485, which is slightly newer. So it seems we're all in the same kind of ball-park.

I don't see a lot which could go wrong with the Econet card - thought I don't have a schematic. The ROM interface stuff is a standard functional block from the network card spec (https://www.4corn.co.uk/archive/docs/Ne ... on-opt.pdf), the rest seems to be bus cycle decoding, buffering and cycle stretching (with the READY line). I can't think of many things that could go wrong -- especially when they cause such a weird failure mode.

[BeebMaster]

Do you have an A7000 you could try it in? Or have, or build, an Econet-Module-Podule for the RISC PC?

[philpem]

Do you have an A7000 you could try it in? Or have, or build, an Econet-Module-Podule for the RISC PC?

Sadly I have neither. I could make an Econet-module-Podule (probably with a proper ECId and cycle-stretching too...) but after spending £125 on the AEH60 I was hoping I could get it working!

[BeebMaster]

A couple of other things to try to rule out a problem with the Econet NIC...

Try reducing the clock speed a bit in case it's just too fast for your setup. Maybe also make a very limited network with the RPC and one other station connected together using the clock only and see if the RPC will serve files that way.

Check that there isn't a problem with the shared Level 4 area in the Exports file, e.g. non-existent location, broken directory, disc error, very long pathname, characters supported by ADFS but not NFS, etc. Maybe use a temporary "one-line" Exports file sharing a new top-level directory such as "$.Net" and see if the user station can save/load using that.

Perhaps other people with an Econet NIC could post pictures of theirs, see if we can spot any mods? I don't have mine here at present otherwise I would take out IC4 and the ROM to see if anything has been re-worked underneath.

[richw]

Sadly I don't have the hardware any more, so could not oblige with a photo.

I thought maybe one of the Econet boffins on here who understood the hardware could maybe compare the circuits...

[philpem]

A couple of other things to try to rule out a problem with the Econet NIC...

Try reducing the clock speed a bit in case it's just too fast for your setup. Maybe also make a very limited network with the RPC and one other station connected together using the clock only and see if the RPC will serve files that way.

Check that there isn't a problem with the shared Level 4 area in the Exports file, e.g. non-existent location, broken directory, disc error, very long pathname, characters supported by ADFS but not NFS, etc. Maybe use a temporary "one-line" Exports file sharing a new top-level directory such as "$.Net" and see if the user station can save/load using that.

Perhaps other people with an Econet NIC could post pictures of theirs, see if we can spot any mods? I don't have mine here at present otherwise I would take out IC4 and the ROM to see if anything has been re-worked underneath.

I've tried a slower clock -- I flipped the highest "period" switch on, which I think would be about a 12us period with 1us mark. No dice. I haven't tried 50/50 mark/space at that speed, I didn't think it'd have any effect.

I've also tried the thing with exporting the root -- I've tried exporting the IDEFS hard drive and the RAM drive. Same problem. It wasn't a one-line Exports file but everything else was commented out.

I could try to pull the StrongARM-compatible Econet modules from the RO3.71 ROM and give them a shot (RO4.02 doesn't have Econet in ROM). Sadly I can't fish them out with an emulator because they won't RMLoad without the hardware. I think there's a tool somewhere on Stardot to split ROM images...

Unfortunately I don't know Econet well enough to be able to figure out what types of hardware failure would cause the file data transfer to fail. It did work (briefly) on Advanced Level 4 when I started playing with the delays, so perhaps it's some kind of timing issue.

I've dropped a post on ROOL's forums too -- https://www.riscosopen.org/forum/forums/11/topics/15417 -- hopefully someone can find Carl Elkins and see if he remembers what the fix was.

[BeebMaster]

Here's a read-out of the module versions I have, including some which aren't Econet specific which I know were troublesome (because they were missing or old versions) when I first started with L4 FS:

[BeebMaster]

Another idea reading some earlier posts...make sure the net number and station number of the file server is set correctly on the user station.

If the file server is station 123 do *FS 0.123 on the user station. Sometimes omitting the net number (0, meaning local net) can result in a junk number being set as the network number. For example if you do *FS 123 and read it back with just *FS it may say "File server is 47.123" or something like that.

Also try logging on specifying net and station number of the file server - *I AM 0.123 SYST etc.

I'm not sure if there's a version of NetMon for 32-bit stations, using that will show the packets being transmitted and may pinpoint where the transmissions are being lost.

[dandoore]

According to https://heyrick.eu/econet/intro/servers.html it's an enhanced version of Level 4 which was released around 1997. It was developed by Network Solutions.

There's a bit more about it here: http://web.archive.org/web/199902181850 ... atext.html

There's a demo version here: http://web.archive.org/web/199902182046 ... /demo.html

And some old updates here: http://web.archive.org/web/199902220612 ... nload.html

I get the impression this used to be on MDFS.net (there's a Level 4 v2.11 ZIP link there), but the file now 404s: http://mdfs.net/Apps/Networking/FServers/

Cool, cheers - I have V2.12 up and running on the Arch running the demo version patched

Dan.