All of lore.kernel.org
 help / color / mirror / Atom feed
From: Jeremy Fitzhardinge <jeremy@goop.org>
To: Keir Fraser <Keir.Fraser@cl.cam.ac.uk>
Cc: Glauber de Oliveira Costa <gcosta@redhat.com>,
	lguest@ozlabs.org, kvm-devel@lists.sourceforge.net,
	linux-kernel@vger.kernel.org, ak@suse.de, chrisw@sous-sol.org,
	tglx@linutronix.de, anthony@codemonkey.ws,
	akpm@linux-foundation.org,
	virtualization@lists.linux-foundation.org, mingo@elte.hu
Subject: Re: [PATCH 3/16] read/write_crX, clts and wbinvd for 64-bit paravirt
Date: Thu, 01 Nov 2007 10:41:19 -0700	[thread overview]
Message-ID: <472A0FBF.6040907@goop.org> (raw)
In-Reply-To: <C34FAA4C.17CA9%Keir.Fraser@cl.cam.ac.uk>

Keir Fraser wrote:
> volatile prevents the asm from being 'moved significantly', according to the
> gcc manual. I take that to mean that reordering is not allowed.
>   

That phrase doesn't appear in the gcc manual; in fact, it specifically
says that reordering can happen:

    The `volatile' keyword indicates that the instruction has important
    side-effects.  GCC will not delete a volatile `asm' if it is reachable.
    (The instruction can still be deleted if GCC can prove that
    control-flow will never reach the location of the instruction.)  Note
    that even a volatile `asm' instruction can be moved relative to other
    code, including across jump instructions.  For example, on many targets
    there is a system register which can be set to control the rounding
    mode of floating point operations.  You might try setting it with a
    volatile `asm', like this PowerPC example:

                asm volatile("mtfsf 255,%0" : : "f" (fpenv));
                sum = x + y;

    This will not work reliably, as the compiler may move the addition back
    before the volatile `asm'.  To make it work you need to add an
    artificial dependency to the `asm' referencing a variable in the code
    you don't want moved, for example:

             asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
             sum = x + y;

I take from this that it is not a good idea to assume "asm volatile" has
any ordering effects at all.

    J

WARNING: multiple messages have this Message-ID (diff)
From: Jeremy Fitzhardinge <jeremy-TSDbQ3PG+2Y@public.gmane.org>
To: Keir Fraser <Keir.Fraser-kDbDZe0LBGWFxr2TtlUqVg@public.gmane.org>
Cc: lguest-mnsaURCQ41sdnm+yROfE0A@public.gmane.org,
	kvm-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org,
	mingo-X9Un+BFzKDI@public.gmane.org,
	ak-l3A5Bk7waGM@public.gmane.org,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org,
	anthony-rdkfGonbjUSkNkDKm+mE6A@public.gmane.org,
	tglx-hfZtesqFncYOwBW4kG4KsQ@public.gmane.org,
	Glauber de Oliveira Costa
	<gcosta-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>,
	virtualization-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org,
	akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org
Subject: Re: [PATCH 3/16] read/write_crX, clts and wbinvd for 64-bit paravirt
Date: Thu, 01 Nov 2007 10:41:19 -0700	[thread overview]
Message-ID: <472A0FBF.6040907@goop.org> (raw)
In-Reply-To: <C34FAA4C.17CA9%Keir.Fraser-kDbDZe0LBGWFxr2TtlUqVg@public.gmane.org>

Keir Fraser wrote:
> volatile prevents the asm from being 'moved significantly', according to the
> gcc manual. I take that to mean that reordering is not allowed.
>   

That phrase doesn't appear in the gcc manual; in fact, it specifically
says that reordering can happen:

    The `volatile' keyword indicates that the instruction has important
    side-effects.  GCC will not delete a volatile `asm' if it is reachable.
    (The instruction can still be deleted if GCC can prove that
    control-flow will never reach the location of the instruction.)  Note
    that even a volatile `asm' instruction can be moved relative to other
    code, including across jump instructions.  For example, on many targets
    there is a system register which can be set to control the rounding
    mode of floating point operations.  You might try setting it with a
    volatile `asm', like this PowerPC example:

                asm volatile("mtfsf 255,%0" : : "f" (fpenv));
                sum = x + y;

    This will not work reliably, as the compiler may move the addition back
    before the volatile `asm'.  To make it work you need to add an
    artificial dependency to the `asm' referencing a variable in the code
    you don't want moved, for example:

             asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
             sum = x + y;

I take from this that it is not a good idea to assume "asm volatile" has
any ordering effects at all.

    J

  parent reply	other threads:[~2007-11-01 17:41 UTC|newest]

Thread overview: 83+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2007-10-31 19:14 [PATCH 0/7] (Re-)introducing pvops for x86_64 - Real pvops work part Glauber de Oliveira Costa
2007-10-31 19:14 ` Glauber de Oliveira Costa
2007-10-31 19:14 ` [PATCH 1/16] Wipe out traditional opt from x86_64 Makefile Glauber de Oliveira Costa
2007-10-31 19:14 ` Glauber de Oliveira Costa
2007-10-31 19:14   ` Glauber de Oliveira Costa
2007-10-31 19:14   ` [PATCH 2/16] paravirt hooks at entry functions Glauber de Oliveira Costa
2007-10-31 19:14     ` Glauber de Oliveira Costa
2007-10-31 19:14     ` [PATCH 3/16] read/write_crX, clts and wbinvd for 64-bit paravirt Glauber de Oliveira Costa
2007-10-31 19:14     ` Glauber de Oliveira Costa
2007-10-31 19:14       ` Glauber de Oliveira Costa
2007-10-31 19:14       ` [PATCH 4/16] provide native irq initialization function Glauber de Oliveira Costa
2007-10-31 19:14         ` Glauber de Oliveira Costa
2007-10-31 19:14         ` [PATCH 5/16] report ring kernel is running without paravirt Glauber de Oliveira Costa
2007-10-31 19:14           ` Glauber de Oliveira Costa
2007-10-31 19:14           ` [PATCH 6/16] export math_state_restore Glauber de Oliveira Costa
2007-10-31 19:14             ` Glauber de Oliveira Costa
2007-10-31 19:14             ` [PATCH 7/16] native versions for set pagetables Glauber de Oliveira Costa
2007-10-31 19:14             ` Glauber de Oliveira Costa
2007-10-31 19:14               ` Glauber de Oliveira Costa
2007-10-31 19:14               ` [PATCH 8/16] add native functions for descriptors handling Glauber de Oliveira Costa
2007-10-31 19:14                 ` Glauber de Oliveira Costa
2007-10-31 19:14                 ` [PATCH 9/16] This patch add provisions for time related functions so they Glauber de Oliveira Costa
2007-10-31 19:14                 ` Glauber de Oliveira Costa
2007-10-31 19:14                   ` Glauber de Oliveira Costa
2007-10-31 19:14                   ` [PATCH 10/16] export cpu_gdt_descr Glauber de Oliveira Costa
2007-10-31 19:14                     ` Glauber de Oliveira Costa
2007-10-31 19:14                     ` [PATCH 11/16] turn priviled operation into a macro in head_64.S Glauber de Oliveira Costa
2007-10-31 19:14                     ` Glauber de Oliveira Costa
2007-10-31 19:14                       ` Glauber de Oliveira Costa
2007-10-31 19:14                       ` [PATCH 12/16] tweak io_64.h for paravirt Glauber de Oliveira Costa
2007-10-31 19:14                       ` Glauber de Oliveira Costa
2007-10-31 19:14                         ` Glauber de Oliveira Costa
2007-10-31 19:14                         ` [PATCH 13/16] native versions for page table entries values Glauber de Oliveira Costa
2007-10-31 19:14                         ` Glauber de Oliveira Costa
2007-10-31 19:14                           ` Glauber de Oliveira Costa
2007-10-31 19:14                           ` [PATCH 14/16] prepare x86_64 architecture initialization for paravirt Glauber de Oliveira Costa
2007-10-31 19:14                             ` Glauber de Oliveira Costa
2007-10-31 19:15                             ` [PATCH 15/16] consolidation of paravirt for 32 and 64 bits Glauber de Oliveira Costa
2007-10-31 19:15                             ` Glauber de Oliveira Costa
2007-10-31 19:15                               ` Glauber de Oliveira Costa
2007-10-31 19:15                               ` [PATCH 16/16] make vsmp a paravirt client Glauber de Oliveira Costa
2007-10-31 19:15                                 ` Glauber de Oliveira Costa
2007-11-01  4:38                                 ` Jeremy Fitzhardinge
2007-11-01  4:38                                 ` Jeremy Fitzhardinge
2007-11-01  4:38                                   ` Jeremy Fitzhardinge
2007-10-31 19:15                               ` Glauber de Oliveira Costa
2007-10-31 19:14                           ` [PATCH 14/16] prepare x86_64 architecture initialization for paravirt Glauber de Oliveira Costa
2007-11-01  4:50                       ` [PATCH 11/16] turn priviled operation into a macro in head_64.S Jeremy Fitzhardinge
2007-11-01  4:50                       ` Jeremy Fitzhardinge
2007-11-01  4:50                         ` Jeremy Fitzhardinge
2007-11-01 13:50                         ` Glauber de Oliveira Costa
2007-11-01 13:50                           ` Glauber de Oliveira Costa
2007-11-01 13:50                         ` Glauber de Oliveira Costa
2007-10-31 19:14                   ` [PATCH 10/16] export cpu_gdt_descr Glauber de Oliveira Costa
2007-10-31 19:14               ` [PATCH 8/16] add native functions for descriptors handling Glauber de Oliveira Costa
2007-10-31 19:14           ` [PATCH 6/16] export math_state_restore Glauber de Oliveira Costa
2007-10-31 19:14         ` [PATCH 5/16] report ring kernel is running without paravirt Glauber de Oliveira Costa
2007-10-31 19:14       ` [PATCH 4/16] provide native irq initialization function Glauber de Oliveira Costa
2007-11-01  4:48       ` [PATCH 3/16] read/write_crX, clts and wbinvd for 64-bit paravirt Jeremy Fitzhardinge
2007-11-01  4:48       ` Jeremy Fitzhardinge
2007-11-01  4:48         ` Jeremy Fitzhardinge
2007-11-01 13:48         ` Glauber de Oliveira Costa
2007-11-01 13:48         ` Glauber de Oliveira Costa
2007-11-01 13:48           ` Glauber de Oliveira Costa
2007-11-01 15:30           ` Jeremy Fitzhardinge
2007-11-01 15:30             ` Jeremy Fitzhardinge
2007-11-01 16:07             ` Keir Fraser
2007-11-01 16:07               ` Keir Fraser
2007-11-01 16:13               ` Glauber de Oliveira Costa
2007-11-01 16:13                 ` Glauber de Oliveira Costa
2007-11-01 16:13               ` Glauber de Oliveira Costa
2007-11-01 17:41               ` Jeremy Fitzhardinge [this message]
2007-11-01 17:41                 ` Jeremy Fitzhardinge
2007-11-01 16:55                 ` [Lguest] " Zachary Amsden
2007-11-01 16:55                   ` Zachary Amsden
2007-11-02  1:21                   ` [Lguest] " Jeremy Fitzhardinge
2007-11-02  1:21                     ` Jeremy Fitzhardinge
2007-11-02  1:21                   ` Jeremy Fitzhardinge
2007-11-01 16:55                 ` Zachary Amsden
2007-11-01 17:41               ` Jeremy Fitzhardinge
2007-11-01 16:07             ` Keir Fraser
2007-11-01 15:30           ` Jeremy Fitzhardinge
2007-10-31 19:14   ` [PATCH 2/16] paravirt hooks at entry functions Glauber de Oliveira Costa

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=472A0FBF.6040907@goop.org \
    --to=jeremy@goop.org \
    --cc=Keir.Fraser@cl.cam.ac.uk \
    --cc=ak@suse.de \
    --cc=akpm@linux-foundation.org \
    --cc=anthony@codemonkey.ws \
    --cc=chrisw@sous-sol.org \
    --cc=gcosta@redhat.com \
    --cc=kvm-devel@lists.sourceforge.net \
    --cc=lguest@ozlabs.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@elte.hu \
    --cc=tglx@linutronix.de \
    --cc=virtualization@lists.linux-foundation.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.