On Sat, 06 Dec 2003 02:39:55 +0000,
"Roland Postle" <[EMAIL PROTECTED]> wrote:
>Here's a quick patch against kdb-v4.3-2.4.22-i386-1 to prevent a null
>dereference when using ptex on user space addresses of a process which
>doesn't have an mm structure (ie. swapper).
Thanks, added to my kdb tree.
>Incidentaly kdb acts a little strange elsewhere with the swapper
>process. If it's an active process it'll list it in ps and begin with
>it selected, but if it's not you won't see it in ps and can't select it
>(with the pid command) because for_each_task() doesn't enumerate it. I
>guess it's something worth debugging, so this should be fixed?
kdb v4.4 adds a new process state, 'I' for idle. The swapper[*] task
will have this state if it is truly idle, i.e. it is not servicing an
interrupt. Both ps and btc will not show tasks in state 'I' by
default, these tasks contain no useful data and they fill up the output
on large machines. Instead kdb v4.4 will print a line like this
idle cpus, use ps I to see them: 0, 2-4, 7-21, 22-255
SGI have machines with large cpu counts.
[*] I assume you know that the swapper task does not really swap. The
name is a hangover from 2.0 kernels where it handled swapping
directly. Now the idle task just spins and may service interrupts.
Use http://oss.sgi.com/ecartis to modify your settings or to unsubscribe.
