Hi,First of all, thank you for the hint. Savecore -c worked like a charm. I've also taken your suggestion, and generated a kernel with debugging (without re0). In short, this does not change anything so I still have no coredump to debug (see point 3 and 4 below)
Second, it appears as if the problems are related to the fact that I was running NetBSD in a (KVM) virtual machine. On real hardware (which took some time to get, hence my late response), it's simply gone.
Now, having said that, I'm still very interested in running NetBSD from KVM (not Xen for other reasons).
Running NetBSD (ffv1 / and /tmp) from a KVM (with bridged network configuration) now causes me serious trouble:
1) (Unrelated) The virtual re* network interface has "re0: watchdog timeout" problems. A workaround seems to be compiling a kernel without "re*", which causes the rtk* driver to attach to the interface. This works mostly, however, I am now seeing occasional "rtk0: watchdog timeout" messages, after which the network becomes unusable, and I have also been able to trigger a kernel crash by (putatively) external network traffic and an unresponsive external NFS server (NetBSD crashed within seconds of the external NFS server, having mounted only a non-system directory).
2) fssconfig immediately causes a kernel crash and core dump. This is unrelated to my removal of the re* driver, because I get the same problem with a stock GENERIC kernel.
3) <ctrl>-<alt>-<escape>, followed by "sync" generates a memory dump. After this memory dump I'm back in the same repetitive core-saving cycle, and I have to "savecore -c" again.
4) savecore (during reboot) fails with an "savecore: (null): Bad address", and while I end up with a "netbsd.N.core.gz", the accompanying "netbsd.N.gz" file is simply empty. My suspicion is that this is what is causing 3) to happen.
Apparently there is a problem with savecore. If I uncompres the memory dump manually, it is indeed the size of the memory (1GB). I have looked inside (hexdump), but found no clue of where to find and extract what gdb needs.
What can I do now? with kind regards, Theo Juergen Hannken-Illjes wrote:
On Fri, Feb 11, 2011 at 01:08:58PM +0100, theo borm wrote:Hi,After upgrading (actually a new install) to NetBSD 5.1 I find myself unable to use filesystem snapshots. "fssconfig -c -x fss0 / /tmp/backingstore" reliably results in lots of green text on the console and attached (first part) log entries.What type file system type is / and /tmp? Do you have 'makeoptions DEBUG="-g"' in your kernel config? Are you able to debug the coredump with gdb and get a backtrace? [snip]How do I stop this save-core-ing?savecore -c ?