OS4 Final Stability and Network problems

I've been having problems with stability under OS4 Final. My MicroA1 will freeze up while using IBrowse OEM, AWEB and other network applications. My system was very stable under Update 4 with uptime measured in weeks and months.

Before anyone asks, yes I did a clean install. The only thing I have added to the core OS is the prism2.device for my wireless card and the MUI libs I needed for Simplemail.

(Phase 1) - I started to get the feeling that the instability was caused by my wireless card or the prism driver. I removed the prism driver and the associated network interface from the machine and rebooted. The system was rock solid but obviously with no internet.

(Phase 2) - I hooked up my A1 directly to my DLink DI-624 Wireless rounter and did the basic DHCP configuration for the A1's onboard network card. BAM! I could barely boot up the system. Hard freezes all over the place. Usually the system would freeze just after Workbench loaded but sometimes it would freeze during the black boot screen.

(Phase 3) - Checked my battery. It was fine. Tried a new one just in case. Still the same issue. I couldn't use the machine for more than 60 seconds before it would freeze.

(Phase 4) - Disconnected my network cable. WOW! The machine boots after a long delay. Runs perfectly stable but with no network. On a whim I configure a fixed IP address, DNS server and Default Gateway on the onboard network card. Reboot. TOTALLY STABLE. I ran this thing for hours at a time banging multiple browsers with NO issues. Rebooted and again Rock solid with a fixed address. I was really hard on this machine and I couldn't get it to crash.

(Phase 5) - Back to DHCP. Again totally unstable. Hard freezes that can only be resolved by powering off the machine.

(Phase 6) - Removed the config for the internal card and reinstalled the wireless card. This time I setup the wireless card with a fixed address. Reboot. So fare after an hours worth of testing and 1 test reboot the system is rock solid.

CONCLUSION - OS4 Final hates my D-Link wireless router when using DHCP. I was very careful in testing to make sure I wasn't doing anything stupid. It looks like there are issues with Roadshow or the drivers.

HELP!

EDIT: Another odd "feature" that i forgot to report. When the machine is configured with DHCP on either network card I was getting packet loss of 20% or more when I pinged my router. With a fixed address I get 0% packet loss. I have 2 Macs and 2 Windows XP machines that have no issues with packet loss or DHCP.

@BobW

Quote:

If you connected your Amiga to uplink port then you need a cross over cable, you can?t use normal Ethernet cable, the black boot screen is most likely due long DHCP timeouts, adding run <>nil: to addinterfaces in startup sequence will most likely make your computer boot.

@LiveForIt

Quote:

I'm aware of the uplink port. I've worked in IT for 13 years.

Even if I add the run <>nil thing the machine will become unstable after the network connection is up. I don't mind the long wait for DHCP. What bothers me are the crashes after

@BobW

After a crash you can type "C:DumpDebugBuffer" in shell you can even do it after warm reboot, you can even get debug information provided by Grims and DebugPrintF, I think there is way to increases debug level in OS4, have look at kickstartlayout in update4 or the documentations on the OS4 install final CD

You can also try changing to best pio mode on your hard disks to see if that helps, if your getting GRIMS from AWEB / IB place post the stack trace part and the version number of programs and library?s in the stack trace.

@LiveForIt

Unfortunately I have to do a cold reboot after these crashes so I doubt I'll get any debug info. The system is totally frozen.

Edit: Also I'm not blaming Aweb or Ibrowse. Seems like network activity in general is causing problems.

@BobW
Quote:
I have some questions/suggestions:
- Do you have any Grim Reaper reports?
- Do you have a serial connection?
- Did you try the debug kernel (kernel.debug) and adjust the debug level?
- Did you try using the DEBUG option in your interface config file?
- Did you try using tcpdump to view the DHCP data exchange?

There are some docs in the Documentation drawer on the OS4 CD which may help as well.

@ssolie

Quote:

I don't see Grim when this happens. It's like a hardware freeze.

Quote:

No but I'll see if I can scare up a cable. Micro's do have the pins for serial on the board right?

I haven't done anything with debug levels yet. I'll have to dig into the docs when I get home.

@BobW

I have a d-link624 router and it works fine with OS4Final.

I had to set up an interface using the eth3com device in prefs/internet.
I also set up a host as 192.168.0.1 router router, and an entry under name resoloution of 192.168.0.1.

My A1 is wire connected to the Dlink to port lan1, I have no problems whatsoever.

I have not the technical expertise to give you any other help, except that your set up works for me.

I wonder if you have some settings that were required for the wireless card and not for direct link.

@Swoop

Quote:

Are you using DHCP or did you use a static IP? DHCP should provide the default router.

I'd also be interested to know what the firmware version is for your router. You can find out by going into the web interface at http:// 192.168.0.1

For problems with your routers, try this:

Edit your connection (in Devs:NetInterfaces/ETH3COM) and modify the line "mtu" this way:
mtu=1500

Then create the directories ENVARC:RoadShow/tcp

Open NotePad and type 1460, save your text in ENVARC:RoadShow/tcp naming it "mssdflt" (without the quotes of course). Reboot.

All shoud be ok then. It worked for me

@K-L

Quote:

mtu is for the device driver. Is the mssdflt for tcpip stack to set limit on transfer units?
The whole issure reminds me a problem I had with "level-one" router (non wireless) hooked to a dsl modem. PC defaulted to 1500, and the dsl modem had 1492 or something on lan side. It caused non-reliable connection, not dhcp or static connection but pppoe (handled by the router). Going down to 1490 on pc solved the issue.
The difference here is that when using static connection the mtu is explicilty set as the default one (probably 1500). when automatic the mtu can be set by the server. That leads to the dhcp client that doesn't support it well or bugged? Wrong mtu should result in bad connection, but not freeze (imho). So according to this logic it is bad mtu handling by dhcp client of Roadshow (overflow or something). Bad mtu negotiation can be blame of the router or Roadshow.

I would experiment with 1490 to see if it reliable (broblems should apper pretty fast if they exist).

Jack

@BobW

An impertinent question I know, but did you boot the Install CD and perform the installation from the Install system, or did you load the CD and run the Installation script from your old system?

@BobW

I know most of this thread is focussing on the net side of things but I notice you mentioned copying over mui stuff.

Please make sure that you install the mui ppc classes available for Ibrowse and ensure that you haven't contaminated your sys:libs directory with old mui stuff, I believe this is essential under Final.

There is another thread covering this here or on AW, start by looking in OSDepot for the classes. You may have more than one problem making diagnostics very hard.

cheers
ace

@acefnq

Yeah, but having incorrect or missing classes usually brings up an error message telling you that anyway.

@BobW

Quote:

Sorry about not getting back to you sooner!!

I use DHCP, and from the router itself the firmware is version 2.09, dated mon, 02 jan 2006, the subnet is 255.255.255.0, and DHCP server is enabled.

Hope that helps.

@Tonyw

Quote:

I formatted the drive then booted from CD. It's a totally clean install.

@acefnq

Quote:

I've been very careful with the MUI classes. I've only installed a couple and they are all the latest PPC libraries.

I don't think that has anything to do with it anyway since I get most of the lockups on the built in ethernet before Workbench even loads. Also the nature of the crash is different than one would expect from a bad MUI class.

Also, if it were a MUI problem you wouldn't expect the problem to go away by switching to a static IP address.

@Swoop

Thanks for the info. I may try updating the firmware on my router when I have time.

@All

Unfortunately I didn't have time last night to work with my A1 and tonight isn't looking much better. I have a bunch of things I want to try but no time

Edit: I did have time to check my mail and surf a little this morning. Since setting the IP to static my machine has been running and crunching Dnet with no issues. Uptime is at something like 34 hours now.

@BobW

Most OS4 user that have problems whit network, have copied there network configurations files over from there Update4 backup, did you do the same?

@BobW

Did you copy across your interface file from update#4?

I recreated on os4final using prefs/internet.

name = internet
device driver = DEVS:Networks/eth3com.device

selected - try dynamic configuration (DHCP)
selected - default IP packet size.


Also, I had one hell of a job getting two wireless devices to work on the router.

I have a desktop PC, my A1, and a HP photosmart 2710 connected direct to the router, and my laptop connected wirelessly.
The HP printer is an all-in-one network printer with a wireless adapter built in, and before I got the router my laptop printed wirelessly via that printer.

I won the printer in the raffle at BigBash4, and it was that which encouraged me to set up the router.

With the printer and the laptop connected wirelessly to the router the connection kept dropping and I had to reset the router to get the connection back. After several days phoneing the help line (based in Canada) and not really solving the problem, I had for thr umpteenth time uninstalled and was reinstalling the printer drivers, and configuring its connection (HP's own wizard not my expertise) when I was told there was a problem and I needed to connect the printer via a cable which would be removed later once my printer was configured. I decided to take the easy option and leave the printer connected direct. Since then my network has been rock solid. I assune the problem is to do with the HP network drivers conflicting either with the router, or the laptop. Anyway mu current setup works so I am leaving well alone.

@LiveForIt

Quote:

Nope, totally fresh config. I usually do everything the hard way and restore my old settings manually.

I haven't brought anything from Update 4 over to Final.

@BobW

Ok, I now have 2 workarounds:

1) I can set a static IP

2) Add "Run >NIL:" to AddNetInterface

Without either of those two workarounds my A1 is guarenteed to freeze. I'm pretty sure it has something to do with an incompatibility with my Dlink DI-624 Rev C3 (there are at least 4 different DI-624's). Before I tried anything tonight I flashed my Dlink with the latest bios. That actually made the problem worse! Not only did the system freeze but on the rare occasion where I could get into shell ping had 100% packet loss.

After flashing my router I also tried setting MTU to 1500 and mssdflt to 1460 with no apparent improvement.

I'm still trying to find a serial cable for my Micro and right now I don't have access to my OS4 Final CD so I'm done for now. Maybe this weekend I'll try some different debug options and possibly a sniffer to see whats happening.