DragonFly BSD
DragonFly bugs List (threaded) for 2004-10
[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]

Re: file weirdness

From: Matthew Dillon <dillon@xxxxxxxxxxxxxxxxxxxx>
Date: Mon, 18 Oct 2004 11:29:45 -0700 (PDT)

:This is probably not terribly useful information, but -
:I bit the bullet and updated to HEAD two days ago.  Trying the system
:out today, the first thing I noticed was that it seemed to start out a
:lot faster - small applications (like 'dillo') would seem to load nearly
:instantaneously, and such - but that might have been an illusion after
:getting used to using NetBSD on a slow drive for a side project.  Then
:after some intensive use, DragonFly would return to a normal-seeming
:speed, and after yet more use, drag down to a crawl.
:Specifically I've been doing 'make realquickrel', and it starts out
:fine, but by the time it gets to 'makewhatis', it's only dragging along.
:Opening another rxvt (so that I could run 'top' and see what's going on)
:took about a minute in total.  'top' does not show 'makewhatis' using
:much CPU - rather, it gets stuck in 'vlruwk' state for ages at a time.
:Hmmm, it just finished.  By the looks of it, the other programs like
:'make' and 'install' are sticking in 'vlruwk' quite badly too.
:I haven't run into any problems along the lines of "file reported
:missing when it's actually there" yet.

    There is definitely a problem with vlruwk... actually, it's more likely
    a problem with a vnode ref count leak somewhere.  We have one good bug
    report from Pawel Biernacki already.

    Could you compile up the ncptrace.c and vnodeinfo.c from
    /usr/src/test/debug ?  See the cc line in the comment at the top of 
    each file for how.

    Then run both after your system has gone into the 'slowed down' state 
    and put the results on leaf somewhere.  

    Hopefully this will help me track down the latest issue.

					Matthew Dillon 

[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]