Hmm, do you think a proxy is required? Do you have a PC with a mainstream OS on that network to compare with?
Does ping find the IP address for google.com?
I assume you have "Automatic" mode enabled globally as well as for the interface in Network prefs?
For the problem in work, I'd recommend running some diagnostic commands.
First, to check if you're getting an IP address from DHCP: ifconfig net0
And for DNS (assuming pings aren't blocked on the network): ping google.com
Or try this in Odyssey: https://192.245.157.94/ (if you get a page that says "Sorry", it's working correctly)
OK I'll give that a try, but I realised today I'll need to use noacpi. I've decided to move over to my 64 bit machine for work, and I need to be able to connect it to external displays which is only possibly with noacpi as this allows me to use the screen switch button.
Talking of work, could I please ask for some advice on a recent issue? About a month ago I stopped being able to access websites in work on my 32 bit machine. According to the output from WirelessManager there is an authenticated connection to the netork, but for any website I try in OWB I get a timeout before the website is opened i.e. nothing is transferred at all. It looks like the DNS server is not being reached. Everthing was fine before about a month ago and I found today it's the same for 64 bit. IP address and DHCP are set to auto and everything is fine on my home network.
Any insights would be much appreciated. Of course, it could be a M$ update designed to make things even more difficult for users of alternative operating systems...
Cheers,
Nigel.
The buffer space errors might be a red herring then, as it's not too serious if it doesn't take the stack down (in other circumstances, those errors are often associated with driver bugs that bring networking to a halt).
So maybe you should just continue browsing with noioapic to see over time if it seems as good as noacpi. But if you want to try another ping test, I'd suggest leaving out the "-s 1024" option and see if the flood ping keeps going for say 30 minutes (if you still get buffer errors, note them but let the process continue). Then Ctrl-C it after 30 minutes and check if normal pings still work.
Be careful of spelling with the boot args: it should be noioapic rather than noioacpi. You can check if the IO-APIC is in use by looking at IRQ numbers in PCITool: if in use, high IRQ numbers will be seen like 18, 19, 20 etc.
Please try the 1024 test with noacpi, and with noioapic if you think it may have been misspelled on the initial test. If 1024 kills the connection (can't do a normal ping afterwards), the system (PC+OS) is unstable overall.
This test is with noioacpi set. I tried the with the addition of -s 1024 and very quickly (within about 11 seconds) got 'no buffer space available' and then lines with 'ret = -1' and the no buffer space error repeated. After stoping the program The final stats were:
9563 packets transmitted, 9267 packets received, +271 duplicates, 3% packet loss
round-trip min/avg/max = 1/14/83 ms
How would you like me to procceed - repeat with noacpi & nothing set or try without the -s 1024 parameter?
Cheers,
Nigel.
Thanks, it looks like pinging between two machines is the only way to do this test then in your environment. Did you have time to try the test between machines with the various boot options? You might need to leave the ping running for a long time (e.g. an hour) to get a good idea of reliability. You could also stress the driver further by adding the option "-s 1024" for bigger packets! Or doing a bidirectional flood ping, where each machine is pinging the other.
I have a Linux Mint installed on a hard drive that I can put into my 64 bit machine. Using that, I had the same result (99% packet loss). In every case only 11 packets were responded to by the hub.
Then I went back to AROS and did the ping from my 64 bit machine to my 32 bit machine. In this case there was 0% packet loss with round-trip min/avg/max = 0/2/35 ms. There were 62179 packets transmitted, 61923 packets received, +245 duplicates.
Cheers,
Nigel.
There's a possibility your router isn't cooperating with the flood ping. Could you try the same test from another machine with a different network driver that you suspect is more reliable? If the packet loss percentage is still high, you could try targeting another PC with the ping.
BTW, I get around 0% to 5% packet loss for the same test with atheros5000.
It's atheros5000.device.
Cheers,
Nigel.
Thanks for the tests Nigel. What network driver is this?
Just tried the ping with noacpi, noioacpi and neither set. In all cases when I stoped the test the packet loss was 99%. Ouch...
Cheers,
Nigel.
Edited by ntromans on 23-01-2026 15:51,
6 days ago