NAME

dump − directions for forcing system physical memory dump

DESCRIPTION

On rare occasions, one may desire to secure a snapshot of physical memory. This can, for instance, be of use to systems analysts when a system appears to be hung.  In order to force a dump, perform the following steps:

- Type "<CR>~i"
at the console terminal.  Note that <CR> stands for "carriage return".  This HALTS the machine (without syncing the disks)! - Enter the CP command:  wb (%pc) ff. 
The parentheses are required. This writes the opcode for a BPT (Break Point Trap) instruction (or invalid instruction) in the address at which the CPU will next execute. - Enter the CP command, r
This tells the CPU to run, and execute the BPT.  This will cause a "panic", which will eventually dump as much of physical memory as possible to the configured "dump device" (see config(1M)).  Note: The system will not panic and consequently will not be dumped unless halted within the kernel.  - On the next reboot to multi-user mode, the /etc/rc script will execute savecore, which will copy the dump of physical memory from the dump device to a file system file, vmcore.XX.  If space on the file system that will hold the dump is at a premium, you may want to boot first to single-user mode, run fsck on that file system, and make room for the file that will be created (see savecore(1M)). 

Many utilities (kadb, ps, pstat, vmstat, etc.) can then be used to extract valuable information from the vmcore.XX file. 

SEE ALSO

savecore(1M), kadb(1), ps(1), pstat(1M), vmstat(1), brc(1M), config(1M), shutdown(1M), fdump(1M), dump(1)

CX/UX Administrator’s Reference