qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Paolo Bonzini <pbonzini@redhat.com>
To: Aurelien Jarno <aurelien@aurel32.net>
Cc: qemu-devel <qemu-devel@nongnu.org>
Subject: [Qemu-devel] Re: Regression: segfault on ARM host
Date: Mon, 01 Mar 2010 23:33:23 +0100	[thread overview]
Message-ID: <4B8C40B3.6030008@redhat.com> (raw)
In-Reply-To: <20100301213351.GN8952@hall.aurel32.net>

[-- Attachment #1: Type: text/plain, Size: 1013 bytes --]

On 03/01/2010 10:33 PM, Aurelien Jarno wrote:
> While trying to implement setcond on TCG ARM, I have discovered it does
> not work anymore. I have bisected this regression to:
>
> commit 6113d6d3169393c323ac4c82d756a850145a5e7a
> Author: Paolo Bonzini<pbonzini@redhat.com>
> Date:   Fri Jan 15 09:42:09 2010 +0100
>
>      change while to if
>
>      The while loop will be executed exactly 0 or 1 times, depending on
>      env->exit_request.
>
>      Signed-off-by: Paolo Bonzini<pbonzini@redhat.com>
>      Signed-off-by: Anthony Liguori<aliguori@us.ibm.com>
>
> The assertion is actually triggered. When the next patch removing the
> assertion is also applied it segfaults instead.

Looks like a race.  The only piece of logic that is changed by that 
commit is reverted in the attached patch, can you try it?  If it passes, 
I can resubmit with S-o-b.

If it doesn't pass, I wonder whether the while loop was there to trick 
the compiler into not optimizing something.  Seems a bit too clever though.

Paolo

[-- Attachment #2: qemu-race.patch --]
[-- Type: text/plain, Size: 1030 bytes --]

diff --git a/cpu-exec.c b/cpu-exec.c
index 5d6dd51..61b1c59 100644
--- a/cpu-exec.c
+++ b/cpu-exec.c
@@ -602,9 +602,15 @@ int cpu_exec(CPUState *env1)
                 /* cpu_interrupt might be called while translating the
                    TB, but before it is linked into a potentially
                    infinite loop and becomes env->current_tb. Avoid
-                   starting execution if there is a pending interrupt. */
-                if (!unlikely (env->exit_request)) {
-                    env->current_tb = tb;
+                   starting execution if there is a pending interrupt. 
+		   Doing it this way is necessary to avoid races with
+		   cpu_unlink_tb (called by cpu_exit).  */
+                env->current_tb = tb;
+                if (unlikely (env->exit_request)) {
+                    env->current_tb = NULL;
+		}
+
+		if (likely (env->current_tb)) {
                     tc_ptr = tb->tc_ptr;
                 /* execute the generated code */
 #if defined(__sparc__) && !defined(CONFIG_SOLARIS)

  reply	other threads:[~2010-03-01 22:33 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-03-01 21:33 [Qemu-devel] Regression: segfault on ARM host Aurelien Jarno
2010-03-01 22:33 ` Paolo Bonzini [this message]
2010-03-02  0:05   ` [Qemu-devel] " Aurelien Jarno
2010-03-05 20:03     ` Aurelien Jarno

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=4B8C40B3.6030008@redhat.com \
    --to=pbonzini@redhat.com \
    --cc=aurelien@aurel32.net \
    --cc=qemu-devel@nongnu.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).