From owner-freebsd-hackers  Tue Jul 29 18:46:21 1997
Return-Path: <owner-freebsd-hackers>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.5/8.8.5) id SAA28117
          for hackers-outgoing; Tue, 29 Jul 1997 18:46:21 -0700 (PDT)
Received: from implode.root.com (implode.root.com [198.145.90.17])
          by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id SAA28107
          for <freebsd-hackers@FreeBSD.ORG>; Tue, 29 Jul 1997 18:46:17 -0700 (PDT)
Received: from implode.root.com (localhost [127.0.0.1])
	by implode.root.com (8.8.5/8.8.5) with ESMTP id SAA11272;
	Tue, 29 Jul 1997 18:46:09 -0700 (PDT)
Message-Id: <199707300146.SAA11272@implode.root.com>
To: Terry Lambert <terry@lambert.org>
cc: freebsd-hackers@FreeBSD.ORG
Subject: Re: Location of copyin() and copyout().. 
In-reply-to: Your message of "Tue, 29 Jul 1997 17:30:46 PDT."
             <199707300030.RAA03691@phaeton.artisoft.com> 
From: David Greenman <dg@root.com>
Reply-To: dg@root.com
Date: Tue, 29 Jul 1997 18:46:08 -0700
Sender: owner-freebsd-hackers@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

>> It is generally considered _extremely_ bad to use copyin/copyout, as it
>> precludes your functions from being called from within kernel space.
>> 
>> FWIW, this is what eventually stumped me when it came to getting Linux
>> WABI running on FreeBSD.
>
>This is a good point, considering that kernel preemption is one
>of my own goals.
>
>The issue is recursion with a paging operation in progress.
>
>I think with the limited cases I gave (the page is filled, the
>call made, and the page is still mapped) are OK.
>
>copying out is more likely to cause problems than copying in,
>since you copy in pages after they are touched, but you potentially
>touch on copyout.
>
>Probably there needs to be _copyin/_copyout for the internal usage,
>and a "safe" copyin/copyout for usage which isn't safed against
>reentrancy.

   There is no problem with using copyout() in the kernel to copy data out to
a user process, assuming that the currently running process is the intended
target of the copy. That is the sole purpose of the function. copyout()
correctly handles all issues of COW/ZF/page faults.
   uiomove() is escentially a wrapper for copyin/copyout that has an
optimization for the case of kernel-to-kernel copies (in which case it
uses bcopy instead).

-DG

David Greenman
Core-team/Principal Architect, The FreeBSD Project