Re: [gentoo-user] Die, process! Die!

[Richard Fish]

On 11/3/06, Alan McKinnon <alan@××××××××××××××××.za> wrote:
> On Friday 03 November 2006 06:44, Walter Dnes wrote:
> > On Mon, Oct 30, 2006 at 07:48:58PM -0700, Richard Fish wrote
> >
> > > If -9 doesn't work, it means your kernel is hungup, and yeah,
> > > you'll have to reboot to fix.
> >
> >   Where does kill -15 fit in?
>
> signal 15 is SIGTERM, and the default for kill. The thread is about
> unkillable processes, meaning those that don't go away with kill or
> kill -15

Just to expand on this a bit...

SIGTERM can be caught, blocked, or ignored by a process.  It is
basically asking the _process_ to "quit now!".

SIGKILL cannot be caught, blocked, or ignored by a process.  In fact,
no user-space code is even executed for SIGKILL.  It is basically a
request to the kernel to "wipe this thing from memory!".

So if SIGKILL doesn't work, that usually means that the process has
allocated some resource in the kernel that now cannot be freed.  An
example would be files open on an NFS server (mounted with the 'hard'
option), with dirty buffers needing to be flushed, but the NFS server
cannot be reached.  SIGKILL would attempt to close those files, which
would attempt to flush out those buffers, which would not work.
Another example would be a buggy driver and a hung device (seen this
with ipw3945d on my own system!)

-Richard
-- 
gentoo-user@g.o mailing list