panic: kthread_create called too soon

Adrian Nida nida at musc.edu
Wed Nov 30 06:55:31 PST 2005


<Snip/>
1) Is your dumpdev set in rc.conf?
yep:

dumpdev="/dev/ad0s1b"

That's my swap device which is 1GB in size.  I only have ~384MB physical 
RAM in the machine.

2) Was the crash dump written to your dumpdev? I.e., did you get the 
"...[256] [255] [250]..." progress indication?
Hmmm, I don't remember seeing that.  Do I need to activate an option to 
make that work?

- Do you have enough space in /var/crash? I usually symlink it to 
something like /usr/var.crash or so.
I symlinked /var/crash to a directory on my biggest drive (music). 
There is 7.8GB free currently.  I would imagine that's big enough for this.

I use 'call dumpsys' (typed at the db> prompt) to get my crash dumps. 
Maybe that does the trick for you.
Oh, you have to manually generate them?  I didn't know that (there was 
no mention of it on the wiki).  I'll give that a shot tonight when I get 
back home.

I know a lot of *ix people don't like to hear about Microsoft.  However, 
I really do think they are (now) doing their error reporting right.  If 
a crash (application or OS) occurs, their version of a core file is 
automatically saved, and the user is then prompted to upload the file 
directly to Microsoft.  Now, the one thing I don't like is that this 
option defaults to no.

After transferring the file, the user is given an option to read more 
information about the crash, the error that caused it, possible 
solutions, its status in the bug fixing process, and whether an update 
that solves the issue (or "other issues that may/may not be related") is 
available.

I know the people doing this project are stretching themselves thin, but 
in my mind this is a really nice feature in the Operating System (hint, 
hint).

Thanks,

Adrian





More information about the Kernel mailing list