One specific speed test causes fatal double fault and crashes box
-
I just got pfSense 1.2-RC3 up and running with a dual WAN configuration and every thing seems to be working fine…
I go to various speed test sites and they show I am getting my full connection speed, 50/20 from FIOS. Then as soon as I go to the Java speed test on DSLReports (http://www.dslreports.com/speedtest?loc=5&premium=) pfSense crashes right at the end of the upload test. I rebooted the box and go back to the speed test and it crashed again at the same exact point in the test, the console showing a fatal double fault.
Where should I even start to attempt to figure this issue out? It is only this specific speed test, others max out my connection fine, and I have even downloaded at full 50 mbit for 20 minutes with no issue.
-
Seems you're the first person to ever run into anything like this. I can run the java tests on dslreports repeatedly, behind two pfSense boxes, and have no issues at all. I'm sure thousands of others have run those tests behind pfSense as well.
Only way to get any info on the cause would be to install an update and choose the developer kernel. Then make it crash, and you should get a db> prompt. Run all of the following and post the output. Clear, legible pictures of the screen is acceptable, ideally we would like text output from a serial console since this will have to go to FreeBSD developers.
bt
bt all
show allpcpu
show alllocks
show lock
ps
set $lines 0
allt
show lockedvnods
reset -
OK, I will try that. I have since narrowed it down to happening only when the client running the speed test is a Vista machine.
-
Also, try more than one speed test provider to see if there is a dependency on the server. My own speed test java server is down right now or I'd have you try it. I use it to test my pfSense box and get no errors. I have 100/100 symmetric (fiber) and I can test it up to about 95/89.
-
Well, I switched interfaces, making my LAN go to the Nvidia nve0 interface and my WAN go to my Realtek re1 interface and it doesn't crash anymore. It seems like it might have something to do with the dynamic window scaling on my vista box and one of those interfaces.