From owner-freebsd-emulation  Wed Sep 10 18:16:50 1997
Return-Path: <owner-freebsd-emulation>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.7/8.8.7) id SAA12492
          for emulation-outgoing; Wed, 10 Sep 1997 18:16:50 -0700 (PDT)
Received: from kithrup.com (kithrup.com [205.179.156.40])
          by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id SAA12486
          for <emulation@FreeBSD.ORG>; Wed, 10 Sep 1997 18:16:46 -0700 (PDT)
Received: (from sef@localhost) by kithrup.com (8.8.5/8.6.6) id SAA09679; Wed, 10 Sep 1997 18:16:45 -0700 (PDT)
Date: Wed, 10 Sep 1997 18:16:45 -0700 (PDT)
From: Sean Eric Fagan <sef@Kithrup.COM>
Message-Id: <199709110116.SAA09679@kithrup.com>
To: grog@lemis.com
Subject: Re: Net posting: SCO gets Linux emulation
Cc: emulation@FreeBSD.ORG
Sender: owner-freebsd-emulation@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

>> I looked at the program; it's interesting, but not terribly exciting.
>> However, I might have done the iBCS2 emulation the same way, if I could
>> have -- however, trapping the system call vector in a user-mode program is
>> hard.  If 386BSD had used a different entry vector...
>It's a question of performance.  If I understand it correctly, you do
>a SIGSEGV for every system call.

That is a problem.  It takes something like three system calls for every
single Linux system call.  (I am counting signal delivery as equivalent to a
system call; in practice, it may be a bit higher.)

For syscall-intensive programs, this is a lot of overhead, and is a known
limitation.  However, not running at all is also quite a lot of overhead :).

Sean.