In message <20050511213341.GA15017@home.chello.upc.cz>, Jaromir Dolecek writes:
>On Wed, May 11, 2005 at 11:14:25AM -0400, Steven M. Bellovin wrote:
>> One of my machines has been crashing overnight every few days.  It 
>> seemed to be a hang; I finally realized that I should set ddb.onpanic 
>> to 0, since I wouldn't see a panic message while I was in X.  The 
>> backtrace indicates that the problem is in the ntfs code; I have
>> /windows mounted read-only, so I suspect it's dying during the daily 
>> job.  Is anyone else seeing anything like that?  I've built and booted 
>> a debugging kernel so I can get better diagnostic info (and I'll sendpr 
>> when I do), but I'm perplexed because I don't see this on any of my 
>> other similarly-configured -current machines.  (Btw, I don't know what 
>> the panic is; gdb didn't display it, and I haven't had the chance to 
>> figure out yet why not.)
>
>Perry had similar problem, I haven't observed that particular problem
>myself tho.
>
>Would it be possible to switch to text console for night, leave
>ddb.onpanic back to 1, and get the full ddb traceback?
>

Not for a couple of weeks; I'm traveling too much and sometimes need to 
log in to that machine remotely.  But wouldn't a gdb backtrace from a 
debugging kernel have the same traceback?

		--Prof. Steven M. Bellovin, http://www.cs.columbia.edu/~smb