SLUG Mailing List Archives
Re: [SLUG] vmscan.c killing processes
- To: johna@xxxxxxxxxxxxxxxxxx
- Subject: Re: [SLUG] vmscan.c killing processes
- From: Conrad Parker <conradp@xxxxxxxxxxxxxxx>
- Date: Sun Sep 24 03:11:14 2000
- Cc: slug@xxxxxxxxxxx
- User-agent: Mutt/1.2i
On Sat, Sep 23, 2000 at 05:53:18PM +1100, johna@xxxxxxxxxxxxxxxxxx wrote:
> What does :
> if (vma->vm_ops->swapout(vma, page_map))
> kill_proc(pid, SIGBUS, 1);
> do, and why ?
it attempts to swap the page out, and if that fails (eg. the write
to the swap device fails) it kills the process.
nb. the comment above that about a "really dirty page" just means a
page which is actually dirty, ie. not clean, ie. has been modified
at all since it was last in swap. If the system has to swap it out
and can't then it has no choice but to kill the process, as its
memory is in an undefined state.
> Are there any ways around this functionality ?
avoiding playing 3d games on low memory machines which swap over nfs
is probably a good start ;)