RE: PowerPC page faults - Liu Yu-B13201

From: "Liu Yu-B13201" <Yu.Liu@freescale.com>
To: "Anthony Liguori" <anthony@codemonkey.ws>,
	"Hollis Blanchard" <hollisb@us.ibm.com>
Cc: "Gregory Haskins" <ghaskins@novell.com>,
	"Avi Kivity" <avi@redhat.com>,
	"Chris Wright" <chrisw@sous-sol.org>,
	"Gregory Haskins" <gregory.haskins@gmail.com>,
	<linux-kernel@vger.kernel.org>, <kvm@vger.kernel.org>
Subject: RE: PowerPC page faults
Date: Tue, 12 May 2009 13:46:25 +0800	[thread overview]
Message-ID: <0A1FE637C2C7E148B9573BB60CC630E52565C7@zch01exm26.fsl.freescale.net> (raw)
In-Reply-To: <4A08A411.9060104@codemonkey.ws>

> -----Original Message-----
> From: kvm-owner@vger.kernel.org 
> [mailto:kvm-owner@vger.kernel.org] On Behalf Of Anthony Liguori
> Sent: Tuesday, May 12, 2009 6:18 AM
> To: Hollis Blanchard
> Cc: Gregory Haskins; Avi Kivity; Chris Wright; Gregory 
> Haskins; linux-kernel@vger.kernel.org; kvm@vger.kernel.org
> Subject: Re: PowerPC page faults
> 
> Hollis Blanchard wrote:
> > On Mon, 2009-05-11 at 12:54 -0500, Anthony Liguori wrote:
> >   
> >> For future ppcemb's, do you know if there is an equivalent 
> of a PF exit 
> >> type?  Does the hardware squirrel away the faulting 
> address somewhere 
> >> and set PC to the start of the instruction?  If so, no 
> guest memory load 
> >> should be required.
> >>     
> >
> > Ahhh... you're saying that the address itself (or offset 
> within a page)
> > is the hypercall token, totally separate from IO emulation, 
> and so we
> > could ignore the access size.
> 
> No, I'm not being nearly that clever.
> 
> I was suggesting that hardware virtualization support in future PPC 
> systems might contain a mechanism to intercept a guest-mode 
> TLB miss.  
> If it did, it would be useful if that guest-mode TLB miss "exit" 
> contained extra information somewhere that included the PC of the 
> faulting instruction, the address response for the fault, and enough 
> information to handle the fault without instruction decoding.

I don't think this can come true if it is software managed TLB.
For guest will never know the content of its TLB without the help of software.

> 
> I assume all MMIO comes from the same set of instructions in PPC?  
> Something like ld/st instructions?  Presumably all you need 
> to know from 
> instruction decoding is the destination register and whether it was a 
> read or write?

Yes, but more there is a search in guest TLB to see if it is a guest TLB miss currently.
A big overhead, especially for a large TLB without being organized by sth. like rbtree.