On Tue, Nov 05, 2019 at 12:44:06PM -0500, Brennan Vincent wrote:
> (Note: I have also posted this to the forums, but upon reading the forum 
> guidelines more carefully I realized the mailing list is probably a better 
> venue. So if you are also a forum reader, I apologize for the extra churn.)
> 
> Hello, I am running 13-CURRENT (compiled recently from source using default 
> build settings) and recently I have been kernel panics every so often. They 
> can happen at any time but seem to be more likely when the system is running 
> a graphical environment and is at high load (e.g., during `make -j64 
> buildworld`).
> 
> I have configured my system to collect core dumps: my swap partition is 50 GB 
> (large enough to contain any conceivable minidump), `dumpon` reports that it 
> is indeed configured as a dump partition, and I have `savecore_enable="YES"` 
> in /etc/rc.conf. I also have the sysctl `debug.debugger_on_panic` set to 0, 
> which seems to be necessary for core dumps to happen (instead of breaking 
> into the debugger on panic). Before, when that sysctl was set to 1, my 
> graphical environment would hand; now it reboots. The changing behavior 
> depending on the value of `debug.debugger_on_panic` is what makes me think 
> this really is a kernel panic, as opposed to some other possible issue that 
> could cause a crash.
> 
> The weird thing is that when I manually cause a panic via `sudo sysctl 
> debug.kdb.panic=1` , my system reboots as expected, and a core dump **does** 
> get generated and saved in /var/crash! So it's only the mysterious random 
> crashes that aren't causing core dumps.
> 
> Can anyone help me figure out why core dumps are not getting generated, and 
> how I can possibly debug what is going on?

Are you using one of the DRM graphics drivers?  I've found that setting
dev.drm.skip_ddb="1" in loader.conf is sometimes necessary.
_______________________________________________
[email protected] mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[email protected]"

Reply via email to