Troubleshooting system-level problems

Examining a memory dump with crash(ADM)

The crash(ADM) command provides a great deal of information that is useful when implementing drivers or doing other kernel development, and requires a good understanding of UNIX system internals to take full advantage of the information provided. The notes in this section enable you to get some basic information about a system panic from the crash output.

To run crash on the memory dump on the /dev/swap device (after the system has been rebooted to single-user state but before it goes to multiuser state), issue a command similar to the following. This command assumes that the system had been booted on the /unix.test kernel image when it panicked, and saves all output to the /tmp/crash.out file:

   /etc/crash -d /dev/swap -n /unix.test -w /tmp/crash.out

For example, to run crash on the copy of the memory dump that was written to the /tmp/06May94 file with the ldsysdump(ADM) command using the /unix kernel image file, you would use the command:

   /etc/crash -d /tmp/06May94 -w /tmp/crash.out

For more information on ldsysdump, see ``Recovering from a system panic''.

WARNING: The argument to the -n option must be the name of the executable kernel file on which the system was last booted. If the -n option is not specified, crash assumes the /unix file.