From owner-freebsd-hackers  Sun Jun 15 21:04:35 1997
Return-Path: <owner-hackers>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.5/8.8.5) id VAA13919
          for hackers-outgoing; Sun, 15 Jun 1997 21:04:35 -0700 (PDT)
Received: from agora.rdrop.com (root@agora.rdrop.com [199.2.210.241])
          by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id VAA13914
          for <hackers@FreeBSD.ORG>; Sun, 15 Jun 1997 21:04:33 -0700 (PDT)
Received: from hq.icb.chel.su (hq.icb.chel.su [193.125.10.33])
	by agora.rdrop.com (8.8.5/8.8.5) with ESMTP id VAA06625
	for <hackers@FreeBSD.ORG>; Sun, 15 Jun 1997 21:03:17 -0700 (PDT)
Received: (babkin@localhost) by hq.icb.chel.su (8.8.3/8.6.5) id JAA24558; Mon, 16 Jun 1997 09:30:12 +0600 (ESD)
From: "Serge A. Babkin" <babkin@hq.icb.chel.su>
Message-Id: <199706160330.JAA24558@hq.icb.chel.su>
Subject: Re: C optimizer bug ?
To: terry@lambert.org (Terry Lambert)
Date: Mon, 16 Jun 1997 09:30:11 +0600 (ESD)
Cc: jkh@time.cdrom.com, hackers@FreeBSD.ORG
In-Reply-To: <199706151839.LAA16658@phaeton.artisoft.com> from "Terry Lambert" at Jun 15, 97 11:39:15 am
X-Mailer: ELM [version 2.4 PL23]
Content-Type: text
Sender: owner-hackers@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

> 
> > Okay, here is the description: The function scsi_done() must check
> > whether xs->bp is zero and then depending on it call biodone()
> 
> Declare this variable volitile.  You are probably suffering from
> overambitious ANSI register optimization, where the value being
> tested is loaded into a register and never reloaded (there's
> actually no excuse for this, given that the compiler can determine
> loop scoping and see from that whether or not the optimization
> would fail).  Sun's ANSI C and othes have similar problems.
> 
> Probably the success depends upon whether the value is initially
> zero or not.

xs->bp does not change during the operation, it is initialized before
calling the SCSI code. I tried to compile the kernel with -C -fvolatile
that must declare all values accessed by pointers as volatile
but this given no positive effect. I tried -fvolatile-global
flag too but the kernel refuses to compile with it at all.
The GCC version is 2.6.3. I'm going to study the produced code.

-SB