On Sat, Jan 28, 2006 at 11:51:19AM +0000, Darren Reed wrote: > I'm quite dismayed that it is so hard to get information about a panic > out of NetBSD. Writing a crash dump doesn't appear to work (where the > dump space is the same as swap) and nearly every time there is a panic, > the system faults again in ddb, either locking ddb and the system up > or just preventing you from getting what you need. I'm sorry to say but i have to agree with Darren here. I'm using and have used several architectures running NetBSD but the i386 is on of the worst on crash information recovering. This is part due to the stated lack of messagebuffer survival and the wierd kernel coredump saving in the swap space. If the swap space is not twice the amount of memory in the machine dumping a kernel core is impossible. What would be better is a mechanism that either revovers kernel coredumps _before_ the swap is turned on or a mechanism that allows the swapper to recognize kernel coredumps so it can spare the dump as long as possible. Can't the message buffer be saved in swap space too on each boot (voluntary or not) ? I'd like to see a 3 types of objects in the current setup in swap space: the first few blocks of swap space for message buffer, followed by the kernel coredump and the rest if nessisary for swapping upto recovery is done. On an other note, I also found in -current that ddb can print out correct stack traces but gdb can't on the kernel coredump.... we ought to get remote debugging working again. Thoughts?
Attachment:
pgpF8bh1AcFbg.pgp
Description: PGP signature