On Sunday 04 March 2007 15:38, Jeremie Le Hen wrote: > Hi, Eric, > > On Sun, Mar 04, 2007 at 02:21:54PM -0600, Eric Anderson wrote: > > On 03/02/07 15:04, Jeremie Le Hen wrote: > > >I've just changed my window manager, and I'm experiencing a few issues > > >with it. This may be a panic but I can't be sure for now. Indeed I > > >don't have a serial port on my laptop. > > > > > >I'm sure it is possible to ask the kernel to call doadump() > > >automatically on panic, but I can't remember how. Please, could you > > >give me the correct way to do it ? > > > > > > Any word on this? I'd love to set this up too. Would be nice to > > auto-dump+reset on servers.. > > Robert Watson replied me privately. Hereafter you will find my reply: > > On Sat, Mar 03, 2007 at 03:12:53PM +0100, Jeremie Le Hen wrote: > > On Sat, Mar 03, 2007 at 12:35:43PM +0000, Robert Watson wrote: > > > You may be able to set the "unattended" sysctl or kernel compile option to > > > cause it to panic and reboot rather than dropping into DDB. However, I'm > > > not in front of a box I can test that on right now, so I'm replying > > > privately in case I'm completely wrong :-). > > > > Thank you for your reply. I've checked the code in kern_shutdown.c and > > indeed KDB_UNATTENDED sets the default value of debug.debugger_on_panic > > to 0. In case we didn't drop to the debugger, doadump() is called, > > this is what I want. Actually turning off the sysctl manually should > > also do it. > > I hope this wille help. Yes, KDB_UNATTENDED will work, but there is a bug in that the system should not enter DDB when you are in X, it should just drop a dump and then reboot. This used to work but is broken in 5.x and later I believe. IWBN to get this fixed again. -- John BaldwinReceived on Tue Mar 06 2007 - 16:04:13 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:06 UTC