From owner-freebsd-hackers  Wed Jun 12 15:28:41 1996
Return-Path: owner-hackers
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.5/8.7.3) id PAA09458
          for hackers-outgoing; Wed, 12 Jun 1996 15:28:41 -0700 (PDT)
Received: from paris.CS.Berkeley.EDU (paris.CS.Berkeley.EDU [128.32.34.47])
          by freefall.freebsd.org (8.7.5/8.7.3) with SMTP id PAA09451
          for <freebsd-hackers@freebsd.org>; Wed, 12 Jun 1996 15:28:39 -0700 (PDT)
Received: from paris.CS.Berkeley.EDU (localhost.Berkeley.EDU [127.0.0.1]) by paris.CS.Berkeley.EDU (8.6.11/8.6.9) with ESMTP id PAA21366; Wed, 12 Jun 1996 15:27:42 -0700
From: Josh MacDonald <jmacd@CS.Berkeley.EDU>
Message-Id: <199606122227.PAA21366@paris.CS.Berkeley.EDU>
To: "John S. Dyson" <toor@dyson.iquest.net>
cc: freebsd-hackers@freebsd.org
Subject: Re: strange mmap programming bug 
In-reply-to: Your message of "Wed, 12 Jun 1996 09:35:23 CDT."
             <199606121435.JAA01198@dyson.iquest.net> 
Date: Wed, 12 Jun 1996 15:27:41 -0700
Sender: owner-hackers@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

> > 
> > I'm having this problem.  I don't know if its only on FreeBSD or not,
> > as I have only experienced it on FreeBSD and its hard to reproduce.
> > 
> > Occasionally, when a program of mine exits abnormally, it seems that
> > small files which were opened O_RDONLY and then mmaped are being
> > modified.  I am mmaping the entire file.  I have only noticed this
> > for files smaller than PAGESIZE.  The file's length is changed to
> > be exactly 4096 bytes with zero's padding the original data.
> > 
> More questions:
> 	What filesystems were you using?
> 	How much paging was going on?
> 	Were you ever doing ANY read/write style I/O
> 	to that file?  (even in another process.)
> 
> 	Tell me all you can...  There has been a problem
> 	(that hasn'nt been seen in -current) with mmap.
> 	If you can run -current, try that also.
> 
> 	Thanks...
> 	John

Sorry for not providing enough information.

This is on a UFS filesystem, SCSI, 1G Seagate 1080N.  The machine
is

FreeBSD deceit.xcf.berkeley.edu 2.2-960323-SNAP FreeBSD 2.2-960323-SNAP #0: Mon May  6 19:57:34 PDT 1996     root@:/home3/2.2-960323-SNAP-src/sys/compile/DECEIT  i386

The machine has one other local filesystem mounted which is also
a SCSI disk.  

I don't know how much paging was going on, I have 32 megs of ram and the
machine was not too loaded.  The process may have been paging due to 
it basically mmaps each of a (possibly long--100 in this case)
list of files to compute checksum and then replace RCS-style keywords.

Thinking about it further, I guess the file may have been the last one
open before forking an RCS CI command on the file.  I don't know how
RCS opens files.  I will try to reproduce this again.

-josh