From mboxrd@z Thu Jan  1 00:00:00 1970
From: compudj@krystal.dyndns.org (Mathieu Desnoyers)
Date: Mon, 15 Feb 2010 09:56:30 -0500
Subject: [ltt-dev] [PATCH 04/11] define sync_core for x86 PIC
In-Reply-To: <4B790166.70901@redhat.com>
References: <1266081392-8732-1-git-send-email-pbonzini@redhat.com>
	<1266081392-8732-5-git-send-email-pbonzini@redhat.com>
	<20100214143743.GE5871@Krystal> <4B790166.70901@redhat.com>
Message-ID: <20100215145630.GD13436@Krystal>

* Paolo Bonzini (pbonzini at redhat.com) wrote:
> On 02/14/2010 03:37 PM, Mathieu Desnoyers wrote:
>> Please add a note saying that it also works in the presence of signal
>> handlers, because we are not using the ebx value returned by cpuid.
>>
>> It would not be correct to use this with signals enabled if we cared
>> about the ebx value returned by cpuid.
>
> It would, because %ebx is callee-save and the signal handler would have  
> to reinitialize it (since each shared library for example has its own  
> GOT, it doesn't know if the caller had the same GOT).  In general,  
> functions that are not static or hidden, or that have their address  
> taken need to rebuild the GOT pointer.  Signal handlers fall obviously  
> in the latter category.

OK, that makes perfect sense. Thanks for the explanation. It could be
good to add a few words explaining this in your patch.

Mathieu

>
> Paolo
>

-- 
Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68