Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 23 Apr 2002 15:48:18 -0700 (PDT)
From:      Hiten Pandya <hitmaster2k@yahoo.com>
To:        Garance A Drosihn <drosih@rpi.edu>, Poul-Henning Kamp <phk@FreeBSD.org>, cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org
Cc:        David Cross <crossd@cs.rpi.edu>
Subject:   Re: cvs commit: src/sys/kern kern_malloc.c
Message-ID:  <20020423224818.41884.qmail@web21109.mail.yahoo.com>
In-Reply-To: <p0511170bb8eb8f6394ef@[128.113.24.47]>

next in thread | previous in thread | raw e-mail | index | archive | help
[ multilated text left for easy reference. thanks ]

--- Garance A Drosihn <drosih@rpi.edu> wrote:
> >Does this solve the bug which I told you about caused by
> >md devices, and especially when doing mdconfig -d -u x the
> >second / third time round.. or am I talking nuts again? :)
> 
> Are you running freebsd-stable, or freebsd-current?
> Dave Cross came across some bug with MD that is fixed in
> current, but which hasn't been MFC'ed yet.  I don't know if
> that's the bug you're seeing, though.

Hi Garance,

I use FreeBSD-CURRENT as of 2 weeks ago.  I think I was very quick in
describing the bug / defect / problem.  Here goes... :)

One:
First of all, I started playing around with md disks as I found them a
versatile way of "emulating" real disks, which is for the purpose of creating
temporary JFS memory volumes for testing my ongoing jfsutils port.

Two:
For doing the above, I make my JFS md disks in the following way:

  # mdconfig -a -t vnode -f /tmp/jfs.vn.2 -u 2

Lets say if /tmp/jfs.vn.2 was not available, mdconfig (or the driver) would
still allocate to make the /dev/md2 entry even though the -f entry has been
invalid.  (this is the first bug)

When I try to remake the md device with the right file:

  # mdconfig -a -t vnode -f /tmp/jfs.vn -u 2

I get an error saying that "unit 2" is busy i.e. already allocated for some
other foo purpose, which is truly a bug.  When I destroy the device inorder
to make it happen:

  # mdconfig -d -u 2

It works the first time round.  Once, I thought I will just redo the command
again to see if it has any effects (such as dropping into the debugger), and
nothing happened; the third time I did it, it dropped straight into the
debugger (i.e. DDB).  My DDB 'trace' showed that it ended up in mddetach, but
because I don't have a serial console, I am not able to produce a trace for
attachment.

I don't know if someone experiences the same problem like me, but I think
the problem I have produced above is pretty much straightforward, but I cant
find a suitable fix.  This problem was also noticed by my friend and fellow
porter for JFS4BSD , Martin Faxer (e with excent).

I think the problem is either in 'mdconfig' or in the mddetach function call
in src/sys/dev/md/md.c, but I am not sure. :) I tried to find it, but didn't
get anywhere, as I was hooked with other things to do with my school studies.

I would be very greatful if someone can reproduce this problem.  It will
save me from fsck'ing my disks everytime I get this problem, hehe, and drop
into DDB!. :D

Regards,

  -- Hiten

__________________________________________________
Do You Yahoo!?
Yahoo! Games - play chess, backgammon, pool and more
http://games.yahoo.com/

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20020423224818.41884.qmail>