This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
Home Products Support Community News


RE: [Xen-ia64-devel] RE: vcpu_translate issue

To: "Matt Chapman" <matthewc@xxxxxxxxxxxxxxx>, "Tian, Kevin" <kevin.tian@xxxxxxxxx>
Subject: RE: [Xen-ia64-devel] RE: vcpu_translate issue
From: "Magenheimer, Dan (HP Labs Fort Collins)" <dan.magenheimer@xxxxxx>
Date: Fri, 11 Nov 2005 07:48:07 -0800
Cc: xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
Delivery-date: Fri, 11 Nov 2005 15:48:03 +0000
Envelope-to: www-data@xxxxxxxxxxxxxxxxxxx
List-help: <mailto:xen-ia64-devel-request@lists.xensource.com?subject=help>
List-id: Discussion of the ia64 port of Xen <xen-ia64-devel.lists.xensource.com>
List-post: <mailto:xen-ia64-devel@lists.xensource.com>
List-subscribe: <http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-ia64-devel>, <mailto:xen-ia64-devel-request@lists.xensource.com?subject=subscribe>
List-unsubscribe: <http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-ia64-devel>, <mailto:xen-ia64-devel-request@lists.xensource.com?subject=unsubscribe>
Sender: xen-ia64-devel-bounces@xxxxxxxxxxxxxxxxxxx
Thread-index: AcXmhY0FrGhj7Q+DTECg3NXLHKUK3gAUU5uw
Thread-topic: [Xen-ia64-devel] RE: vcpu_translate issue
Good catch!  I'm guessing the solution will be for
vcpu_itc (and fast_itc, though it is turned off
by default) to handle this case by using the
correct (rr0) rid when in metaphysical mode.


> -----Original Message-----
> From: Matt Chapman [mailto:matthewc@xxxxxxxxxxxxxxx] 
> Sent: Thursday, November 10, 2005 11:02 PM
> To: Tian, Kevin
> Cc: Magenheimer, Dan (HP Labs Fort Collins); 
> xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
> Subject: Re: [Xen-ia64-devel] RE: vcpu_translate issue
> On Thu, Nov 10, 2005 at 05:02:52PM +0800, Tian, Kevin wrote:
> > 
> > >I think delivering a TLB miss is not sufficient.  Because the
> > >mapping for the zero page is indeed in Linux's pagetable, the TLB
> > >miss handler succeeds in finding and inserting it, and the page
> > >fault handler is not triggered.
> > 
> > Actually that depends. Only when short format vhpt entry 
> exists in TLB
> > which contains that mapping for zero page, low level TLB 
> miss handler
> > can hit it. Or else the flow still goes to C code page 
> fault handler,
> > right? 
> Yes, the C page fault handler is invoked if it's not in the pagetable.
> But as far as Linux is concerned, there is a valid read-only VMA at
> address 0, so the C page fault handler won't raise a fault on read, it
> will simply insert a read-only mapping in the page table.
> So, a write to the zero page will result in an access fault and
> termination of the process.  A read from the zero page will succeed
> (incorrectly).  But that doesn't explain why it now hangs there.
> The more important problem seems to be that itc inserts into the wrong
> region when we're metaphysical, so the execution flow is like this:
> 1. [rr0=X] guest accesses address 0 -> fault
> 2. [rr0=X] Xen accesses VHPT at 0x1fffff000000000 -> deliver 
> 3. guest switches to metaphysical to walk pagetable
> 4. [rr0=mprid] guest inserts translation for 0 and 0x1fffff0000000000
> 5. guest returns to virtual mode (rfi)
> 6. [rr0=X] goto 1
> Matt

Xen-ia64-devel mailing list

<Prev in Thread] Current Thread [Next in Thread>