Highway on AmigaOS 4 classic

Hi everybody....

Anyone who's been reading the OS 4 installation thread on Amigaworld.net will know that I've been suffering instability on classic OS 4.

I think I've diagnosed the problem now as being my Highway, which has always worked perfectly under OS 3.9. If I take the card out I can format a 2Gig partition repeatedly (done it about 8 times) but it never succeeds with the Highway in, even if nothing's connected to it. It usually hangs the machine rock solid after about 10% of a 2 gig partition, but sometimes it'll go up to about 60%.

Has anyone else had this problem? Has anyone else been using a Highway for rigorous testing? Can I do anything about it? Use of my original keyboard mouse isn't much of an option, I'd rather have the instability but I'd prefer no problems at all!

Help? :)

@Spirantho

Today i have been trying a highway on an A4000 with CSPPC and it worked quite well...

Do you have serial output ?

@SimplePPC

How thoroughly did you test it?

Can you try doing a lot of writing to hard disks, as that's when things go kablooie on mine?

Formatting a disk will crash every time (that's a long format, 2Gig disk) unless I disable the Highway, in which case it works perfectly.
Doesn't matter what device it is (on-board IDE, UW SCSI, whatever filesystem)....

I've not got anything connected to the serial device... I wouldn't expect anything out of it as I'm not running a debug kernel or anything. Unless you know differently?

Thanks for your input!

@Spirantho

Do you have any other hardware attached ? I have noticed myself the the Prometheus doesnt always work here with the NIC card for example.

@SimplePPC

Tried with and without other hardware. Just standard AGA and the Highway crashed it.

The only way to make it stable is to remove the highwayusb driver or the card itself.

Very annoying as my A4000 keyboard has some dodgy keys so I need my trusty Wireless keyboard!

I might try a different PSU just in case... but I tried with no extra hard disks and the like and no difference. Plus it's always a freeze, never a reboot as you'd expect if it were power related.

Very frustrating and means I can't really use OS 4 on my 4000 properly.

@Spirantho

You can always comment out the usb lines in the kicklayout file and install posiedon.

I guess you can rule out power that way.

I use posiedon 3.8 here as the subway is not supported by the OS4 final CD.

@Jurassicc

Poseidon doesn't work right, because you can't patch the input.device (which is now in ROM and changed rather).

It'll run, but when it finds the keyboard, the keypresses never auto-repeat (makes life tricky) and the mouse isn't detected at all.

The Highway and Poseidon both run perfectly on my OS 3.9 installation.

@Spirantho

Quote:

I have a setup of A4000T + HIGHWAY running here flawless.

But: if Poseidon + OS3.9 + HIGHWAY work fine, while Sirion + OS4.0 + HIGHWAY do some bad things, then I must assume that there might be a bug inside the Sirion Highway driver. From what you tell here I would bet on a buig inside the interrupt handling.
We have loads of HIGHWAYs out at users for years now, and noone reported such a bug up to now.

I never noticed such behaviour as you described here.

Please report this failure to the AOS4 team, they may be able to reproduce it and maybe fix it.

Michael

@mboehmer

That's what I reckon - the interrupts. It's exactly the same as I got using a Bt8x8 card on my A1.

Is there any updatable firmware on the Highway that could be causing this? Could I be using old drivers somehow? More than one version of the Highway hardware?

If I can help any to fix it I will - I need my USB card for my keyboard and mouse! It's completely reproducible on my machine.. but I don't know how to officially report the problems or how to go about helping to fix it....

All I've heard from people so far is a couple of people saying "works fine here"... which of course isn't much help to me!

For the record, I tried installing Poseidon v3.8 this morning.

The keyboard still doesn't detect some combinations of keypresses, so for instance Shift+Backspace does a single backspace.
The mouse works fine.

...

But it crashes the system just like the OS 4 mative driver does!

This should tell the developers something, though! I'm sure it didn't crash with the version of Poseidon I was using before (v1.2) but I could be mistaken.

Rebooting with no HighwayUSB in the KickLayout and not running Poseidon, all seemed stable in the short tests I could do with it (but which usually break it)

@Jurassicc

Why wouldnt the subway be supported under OS4 ? I think it does...

@Spirantho

When you try the Highway is that with our without the usb keyboard/mouse whatever that isnt clear to me...

There should be a kernel.debug in Kickstart..

@SimplePPC

Just tried formatting the 2GB disk partition (how I test it) and it got to 20% and then froze solid, as I expected. This was with nothing at all connected to the Highway.

Can I debug this? Use the debug kernel? And a serial port ... or...?

@Spirantho

Change the line in your kicklayout-AXXXX

EXEC Kickstart/kernel

To

EXEC Kickstart/kernel.debug

Edit the startupo-sequence to:-

C:Bootloader COMMANDLINE "debuglevel=10" serial baudrate=9600"

or

C:Bootloader VERBOSE COMMANDLINE "debuglevel=10" serial baudrate=9600"

Then save the changes to startup-sequence and either hardreset or type reboot in the shell

C:Bootloader VERBOSE is handy for kickstart problems or if you just want to watch the kickstart being loaded.

Not only will boot take much longer but expect to hold the mouse buttons for quite a while when doing boot with no startup-sequence etc. You'll observe the power light flickering which is a sign both mouse buttons can be released.

I did the changes you suggested and there were no changes, no useful debug output on crash. However, it did get me checking other things...

Now, I format a disk to see if it's working or not - it's the simplest way, and doesn't risk loss of data.

Without the Highwayusb.device.kmod in the Kicklayout, debug kernel:

Boot with no Startup-sequence
SYS:System/Format DRIVE hd1: name Test
Formatting cylinder 1, 2024 to go (0:00:45h remaining) [or similar]

This updates until cylinder 250. The timer counts down, but when it gets to cylinder 250, it says (<1 sec remaining) even thought there's another 1775 cylinders to go. This makes it repeat the message as it goes off the edge of the 80 column screen.
However, it carries on formatting.

Same thing with Highwayusb enabled:

Updates until cylinder 250 as before, repeats the message 10 times or whatever, says "Verifiying cylinder 250" etc.

and the crashes solid.
Just tried it again, got as far as 275 before crashing solid (which is of course the next chunk after 250).

Can someone else with a Highway do exactly the same test? My HD1: drive is a partition on the CSPPC device, but I've had the same problem on other drives (a CD-RW device).

@Spirantho

That sounds like a severe hardware problem. We (a friend of mine and myself) had no formatting problems at all with the Highway plugged in and using the Highway OS4 driver. I bet it would be possible to format a USB hard drive without any problem (if we would use FFS2/SFS). The A4000 is still running very fine (astonishing fast for 233MHz), no freezes or crashes that would be Highway related, so we have really no clue what your problem may be

Regards

@whose

But if it's a hardware problem how come

a) It's only when writing heavily to a disk
b) It's only under OS4.0

With the Highway installed, OS 4 wouldn't install - it'd crash about 75% of the way through the install, even with everything else removed.
With the Highway removed, OS4 installed perfectly.

The hardware is fine, just to prove it I'm playing a DVD (somewhat slowly of course!), it's been on for an hour with no problems, and DVD playing is very CPU intensive..... but if I write a large file then it freezes pretty much every time.

If it's hardware I'd expect it to be much more random, i.e. resets as well as hangs.

It really makes no sense to me unless there's something which is just about valid on most peoples' setups but not mine. Very strange and very annoying.

My CS-PPC is an early one - serial number 72 or something I think.. could that have anything to do with it? Maybe it's the Eagle tower backplane it's on, but it all seems a little too consistent...

Either way, though, if it works on OS 3.9 there must be something OS 4 is doing which is somehow more demanding...

@Spirantho

Somehow I got to many quote marks in there:-

C:Bootloader VERBOSE COMMANDLINE "debuglevel=10 serial baudrate=9600"

Should be the above but you've probably figured that out.

There is a usb.log in t: what does thay say ?

IIRC there was a usbinfo tool in the 51.22 sdk.

Do you have problems formatting IDE drives or Just scsi ?

@Spirantho

Well, if I could answer these questions I would have done it the first time All I can say is that there should be no problem using Highway with OS4 and its native driver, because it works absolutely flawless with my machine (and several others it seems).

We did the OS4 install with all cards plugged in (including the Highway) and had no problems at all. The machine has several other cards plugged in, Toccata (working very fine), IOBlix MultiIO (not tested yet), Ariadne2 ethernet (working very fine) and a GVP SCSI controller (likely not working because of the driver glitches with DMA thingie, but we didn't test it for now).

Later on the friend changed to a USB keyboard and is still very happy with my machine. Its an A4000 Elbox towered and the CSPPC has serial # 69, so it is very likely of the same charge as yours.

Maybe your Eagle backplane is the cause of your problems, but it is very difficult to say if the hardware isn't available for diagnostics. Maybe it is a really severe problem, maybe it is simply a capacitor dried up.

All I can say is, that it can't be a driver problem as there are many people out there using the A4000/Highway/OS4 combo without any glitches (sometimes even with Poseidon instead of Sirion and they still have no problems like yours).

Regarding the differences to OS3.9, you should count in that most likely some timings has changed and bus stressing is much higher with a PPC doing all the hardware related work than it is when the 68k is the "master" of the system.

Very frustrating problem, I know, and I wish I could help easily. But I can't, I'm sorry

Don't know if it helps any but I noticed that my faster hard disk is much more prone to crashing than the slower one (when the Highway is used, naturally).

I tried lowering the disk speed to 16-bit 1MHz symchronous transfers and it became very unstable.. Put it back up to default (20MHz or whatever) and it was back to normal.

It's interesting, though, that the faster the disk the more likely to crash. The Micropolis Stinger 4.3GB 4743WS is the more stable: I copied an 83MB file to it, no problem. I then did a join to itself to double the size ("join test.vob test.vob as test2.vob") and this worked. Did the same again and it crashed.

Tried the same on the Fujitsu (10K RPM 9GB) and it wouldn't even finish copying the file before crashing.

Copying the 83MB file to the DVD+RW in the first place was no problem at all.

I'm hoping the Deneb comes out and works soon so I can try that. This is very frustrating!

I wonder sometimes if the PSU is iffy but then I'd expect reboots not consistent hangs like this. That's what bugs me most - it's so consistent....