X-Sylpheed-Account-Id:1
S:andy.sharp@onstor.com
SCF:#mh/Mailbox/sent
X-Sylpheed-Sign:0
X-Sylpheed-Encrypt:0
X-Sylpheed-Privacy-System:
X-Sylpheed-End-Special-Headers: 1
Date: Wed, 25 Jul 2007 13:26:04 -0700
From: Andrew Sharp <andy.sharp@onstor.com>
To: "Larry Scheer" <larry.scheer@onstor.com>
Cc: Tim Gardner <tim.gardner@onstor.com>, Jay Michlin
 <jay.michlin@onstor.com>
Subject: Re: No simple solution
Message-ID: <20070725132604.5d10dcd6@ripper.onstor.net>
References: <20070109101845.632e4947@ripper.onstor.net>
 <BB375AF679D4A34E9CA8DFA650E2B04E0A9465@onstor-exch02.onstor.net>
Organization: Onstor
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit

OK,

I've looked at Larry's test case and the code to install, and it seems
to stongly suggest a kernel bug in Openbsd vis-a-vis mmap and OOM or
low memory conditions.  In the short term, I recommend that we switch
to using 'cp' instead of 'install', but I'm hearing that this bug has
been biting us in other places, like corruption in the cluster database
when in low memory conditions and so on, so I guess at some point we
might want to consider expending resources to fix it.  Sigh.

I just did an upgrade myself and had occassion to note that *after*
untarring the upgrade release into the ramdisk, the system had 35MiB
free, but as it progressed with the upgrade, it consumed another 25MiB,
which is huge.  I can't think of any reason, other than a memory leak,
why it would have used such a huge amount of memory.  I recommend that
we also audit the code a bit for memory leaks.

Cheers,

a

 On Tue, 9 Jan 2007 11:06:01 -0800 "Larry Scheer"
<larry.scheer@onstor.com> wrote:

> I should have included that in my previous email. My apologies for my
> thoughtlessness.
> Like all of openbsd source it is kept in an obvious place ;-)
> 
> openbsd/src/usr.bin/xinstall/xinstall.c
> 
> Yea, it took me a while to find it too.
> 
> Larry
> 
> 
> 
> -----Original Message-----
> From: Andrew Sharp [mailto:andy.sharp@onstor.com] 
> Sent: Tuesday, January 09, 2007 10:19 AM
> To: Tim Gardner
> Cc: Larry Scheer; Jay Michlin
> Subject: Re: No simple solution
> 
> Where are we keeping the source for install?  I can't find it under
> openbsd.  Or anywhere else.
> 
> a
> 
> On Mon, 8 Jan 2007 21:05:35 -0800 "Tim Gardner"
> <tim.gardner@onstor.com> wrote:
> 
> > Andy,
> >  
> > Please take a look at the source code for install.
> > Lets find out if it is using mmap. Also look at the code for cp. We
> > need to find out what is different. That will help us understand
> > what is causing the corruption. 
> > Tim
> >  
> > 
> > ________________________________
> > 
> > From: Larry Scheer
> > Sent: Mon 1/8/2007 7:32 PM
> > To: Jay Michlin; Andy Sharp; Tim Gardner
> > Cc: Larry Scheer
> > Subject: No simple solution
> > 
> > 
> > 
> > I downloaded, built, and installed the latest revision
> > of /usr/bin/install from OpenBSD 4.0. I started my test suites and I
> > am still seeing file corruption. (/usr/local/agile/lib/libucdmibs.so
> > for instance)
> > 
> > I am still seeing the corruption when running /usr/bin/install
> > standalone (in a script) or with our upgrade program.
> > 
> > I think this indicates it is something deep with BSD, perhaps in
> > libc?
> > 
> > I can provide a map of install so we can see all of the functions it
> > uses. If that would help.
> > 
> > Larry
> > 
> > 
> > 
