From mboxrd@z Thu Jan  1 00:00:00 1970
From: Huang Ying <ying.huang@intel.com>
Subject: Re: [PATCH -v4] QEMU-KVM: MCE: Relay UCR MCE to guest
Date: Tue, 22 Sep 2009 09:12:54 +0800
Message-ID: <1253581974.15717.726.camel@yhuang-dev.sh.intel.com>
References: <1253501005.15717.548.camel@yhuang-dev.sh.intel.com>
	 <4AB750A6.1090000@redhat.com>
Mime-Version: 1.0
Content-Type: text/plain
Content-Transfer-Encoding: 7bit
Cc: Marcelo Tosatti <mtosatti@redhat.com>,
	Andi Kleen <andi@firstfloor.org>,
	Anthony Liguori <aliguori@us.ibm.com>,
	"kvm@vger.kernel.org" <kvm@vger.kernel.org>
To: Avi Kivity <avi@redhat.com>
Return-path: <kvm-owner@vger.kernel.org>
Received: from mga09.intel.com ([134.134.136.24]:44619 "EHLO mga09.intel.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1750783AbZIVBMw (ORCPT <rfc822;kvm@vger.kernel.org>);
	Mon, 21 Sep 2009 21:12:52 -0400
In-Reply-To: <4AB750A6.1090000@redhat.com>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

On Mon, 2009-09-21 at 18:08 +0800, Avi Kivity wrote: 
> On 09/21/2009 05:43 AM, Huang Ying wrote:
> > UCR (uncorrected recovery) MCE is supported in recent Intel CPUs,
> > where some hardware error such as some memory error can be reported
> > without PCC (processor context corrupted). To recover from such MCE,
> > the corresponding memory will be unmapped, and all processes accessing
> > the memory will be killed via SIGBUS.
> >
> > For KVM, if QEMU/KVM is killed, all guest processes will be killed
> > too. So we relay SIGBUS from host OS to guest system via a UCR MCE
> > injection. Then guest OS can isolate corresponding memory and kill
> > necessary guest processes only. SIGBUS sent to main thread (not VCPU
> > threads) will be broadcast to all VCPU threads as UCR MCE.
> >
> >
> >
> > --- a/qemu-kvm.c
> > +++ b/qemu-kvm.c
> > @@ -27,10 +27,23 @@
> >   #include<sys/mman.h>
> >   #include<sys/ioctl.h>
> >   #include<signal.h>
> > +#include<sys/signalfd.h>
> >    
> 
> This causes a build failure, since not all hosts have <sys/signalfd.h>, 
> but more importantly:

Maybe we can just add necessary fields to struct qemu_signalfd_siginfo.
But this may be not portable.

> > +
> > +static void sigbus_handler(int n, struct signalfd_siginfo *siginfo, void *ctx)
> > +{
> >    
> 
> Here you accept signalfd_siginfo, while
> 
> > +
> > +    memset(&action, 0, sizeof(action));
> > +    action.sa_flags = SA_SIGINFO;
> > +    action.sa_sigaction = (void (*)(int, siginfo_t*, void*))sigbus_handler;
> > +    sigaction(SIGBUS,&action, NULL);
> > +    prctl(PR_MCE_KILL, 1, 1);
> >       return 0;
> >    
> 
> here you arm the function with something that will send it a siginfo_t.  
> So it looks like this is broken if a signal is ever received directly?  
> But can this happen due to signalfd?

Because SIGBUS is blocked, I think the signal handler will not be called
directly, but from sigfd_handler.

> >   }
> >
> > @@ -1962,7 +2116,10 @@ static void sigfd_handler(void *opaque)
> >           }
> >
> >           sigaction(info.ssi_signo, NULL,&action);
> > -        if (action.sa_handler)
> > +        if ((action.sa_flags&  SA_SIGINFO)&&  action.sa_sigaction)
> > +            action.sa_sigaction(info.ssi_signo,
> > +                                (siginfo_t *)&info, NULL);
> > +	else if (action.sa_handler)
> >               action.sa_handler(info.ssi_signo);
> >    
> 
> The whole "extract handler from sigaction and call it" was a hack.

The "hack" above (signalfd_siginfo vs siginfo_t) is for "extract handler
from sigaction and call it" too. So I suggest to replace it with calling
handler directly.

Best Regards,
Huang Ying