Why page-in a SIGKILL-ed process?
mi+kde at aldan.algebra.com
Sat Oct 22 09:20:25 PDT 2005
On Saturday 22 October 2005 08:24 am, Peter Jeremy wrote:
= >17850 mi 1 -16 0 4158M 1118M wdrain 1 0:06 6.10% vim
= >The question is: Why bother with paged-out parts of the process, when
= >it is already doomed by SIGKILL?
= wdrain appears to be associated with file I/O rather than paging
= (though I may be wrong here). Is it possible that vim had started
= core-dumping before you SIGKILL'd it? I've seen problems on other
= OS's where core-dumping processes couldn't be killed and caused
= significant performance degradation if they were very large.
Well, indeed, there was a core-dump too. The reason I thought this was
swap-related is because prior to settling on `wdrain', the process was
in `pfault' for a few moments... You are, probably, right -- it was
dumping the vim's core, when I started killing it.
As for the performance degradation during a core-dump, yes, this
definetely is not a FreeBSD-specific problem... Can't this be
interrupted, though, by SIGKILL-ing the dying process?
More information about the freebsd-current