Bug #1093
closed
Problems with em(4)
0%
Description
Some time after Dec 2, something has changed in the build. Now, when I attempt to upgrade to newer builds, or install any packages, the system freezes while attempting to download the packages, and the terminal shows a general protection fault. I have to hard reset to recover. I've done this multiple times, and can repeat it on any recent builds.
The last build I have that doesn't show this issue is:
Built On: Thu Dec 2 16:05:10 UTC 2010
for AMD64.
Updated by Jim Pingle over 13 years ago
- Status changed from New to Feedback
Needs a lot more detail.
I'm running snapshots later than that on amd64 and they do not crash. We need at the very least the exact text of the error message. Ideally, you could install a Developer's kernel (check the doc wiki for info on switching kernels) and get a backtrace.
Updated by Shannon McMahon over 13 years ago
I just updated to the latest build. I then initiated an install of squid as an example. shortly after starting the squid download, it froze again. I can't seem to find the info in the logs, but I have the following screenshot from the console.
http://www.mcmahonweb.com/wp-content/uploads/2010/12/gpf1.jpg
Updated by Chris Buechler over 13 years ago
- Project changed from pfSense Packages to pfSense
Updated by Chris Buechler over 13 years ago
- Category set to Operating System
- Status changed from Feedback to New
- Target version set to 2.0
- Affected Version set to 2.0
this sounds like the same thing I hit last night on em(4) as well, any traffic initiated by or destined to the host itself would quickly cause panics. Traffic through the system not an issue. I have a setup that should be able to replicate easily, will see if I can get a back trace. Suspect but not sure that the last em(4) update from RELENG_8 is at fault.
Updated by Shannon McMahon over 13 years ago
I think your hunch was on the money. I looked into known issues with em(4), and subsequently yanked the dual GB Intel PCI Express NIC. I am now using the internal Broadcom NIC's on the board, and the problem seems to be gone. I just updated to the latest build with no problem. Not a pfSense issue. Thanks for the help.
Updated by Chris Buechler over 13 years ago
Thanks for confirming, it is a pfSense issue though as we changed the driver, need to change back to what we had previously.
Updated by Chris Buechler over 13 years ago
- Subject changed from General Protection Fault to Problems with em(4)
- Priority changed from Normal to High
Updated by Shannon McMahon over 13 years ago
Ah, interesting. I assumed it was the BSD 8 default driver.
Updated by Shannon McMahon over 13 years ago
FYI: Let me know when you change the driver. I can throw the Intel NIC back in and test at that point.
Updated by Ermal Luçi over 13 years ago
- Status changed from New to Feedback
New driver is on new snapshots now.
Updated by Shannon McMahon over 13 years ago
I just reinstalled the Intel GB NIC. Loaded a build earlier today, and then just upgraded the build this evening. Also installed a package. Everything worked perfectly, don't see any errors so far.
Updated by Shannon McMahon over 13 years ago
Arrggghhh!! Problem came back. page fault related to the intel nic.
Updated by Jon Gerdes over 13 years ago
I am using 23rd Dec BETA4 amd64 on vmware with em driver.
I have been using this for some time and have never had a problem.
I notice that nowhere in this thread is there mention of the exact model of NIC involved. Perhaps Shannon could try another Intel NIC in its stead although I understand that multi port NICs are a bit pricey.
Intel NICs are quite common and I'd be surprised if the driver itself was at fault.
Whilst writing this I just installed ntop.
Updated by Max Riedel over 13 years ago
I'm on i386 full latest and never had any problems either. I got 2 dual GBit intel pci-x nics using em drivers..
Updated by Shannon McMahon over 13 years ago
Then close it. I am not buying another NIC to test it. I can say that the PCI Express NIC I have never had any trouble when this machine was used as a file server, and I've now put it in another machine, and it has no issues.
Updated by Chris Buechler over 13 years ago
- Status changed from Feedback to New
This is obviously not entirely fixed, though the bulk of the issues are gone. Something bad in the RELENG_8 driver still. May want to revert back to what we had prior to switching to RELENG_8's driver as I don't recall that having any issues.
Updated by Ermal Luçi about 13 years ago
I committed the driver from STABLE so it should be on new snapshots.
Please anybody with the issue test that snapshot.
Updated by Ermal Luçi about 13 years ago
- Status changed from New to Feedback
The problems should be fixed in latest snapshot that comes out.
Updated by Chris Buechler about 13 years ago
- Status changed from Feedback to Resolved
this seems to be fine.