From mboxrd@z Thu Jan  1 00:00:00 1970
From: Gleb Natapov <gleb@redhat.com>
Subject: Re: [PATCH 1/2] kvm-unit-tests: Add a func to run instruction in
 emulator
Date: Tue, 18 Jun 2013 19:09:51 +0300
Message-ID: <20130618160951.GA22555@redhat.com>
References: <1370532278-22063-1-git-send-email-yzt356@gmail.com>
 <51B8DF0E.3060105@redhat.com>
 <CABpY8M+TNnGyqdQU10DTVj4gGuYNam4_uvTquR2oa==61KrHjw@mail.gmail.com>
 <CABpY8MK+dip6tUmvVbfGDVOxuJkiTq75xH2r_ogLYoECZ6_FQA@mail.gmail.com>
 <20130618124517.GA16557@redhat.com>
 <CABpY8MKSh_fFmqn-5oHYRBA_GQJAuMG7699wLnMaZ4EjUPgtrA@mail.gmail.com>
 <20130618154754.GE21032@redhat.com>
 <CABpY8M+WSG9LvXvPhJ7QytCRbAc33k2WTV936uxoc-Fdgr0DYg@mail.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: QUOTED-PRINTABLE
Cc: Paolo Bonzini <pbonzini@redhat.com>, kvm <kvm@vger.kernel.org>
To: =?utf-8?B?5p2O5pil5aWHIDxBcnRodXIgQ2h1bnFpIExpPg==?=
	<yzt356@gmail.com>
Return-path: <kvm-owner@vger.kernel.org>
Received: from mx1.redhat.com ([209.132.183.28]:56410 "EHLO mx1.redhat.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1754090Ab3FRQJy convert rfc822-to-8bit (ORCPT
	<rfc822;kvm@vger.kernel.org>); Tue, 18 Jun 2013 12:09:54 -0400
Content-Disposition: inline
In-Reply-To: <CABpY8M+WSG9LvXvPhJ7QytCRbAc33k2WTV936uxoc-Fdgr0DYg@mail.gmail.com>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

On Tue, Jun 18, 2013 at 11:56:24PM +0800, =E6=9D=8E=E6=98=A5=E5=A5=87 <=
Arthur Chunqi Li> wrote:
> On Tue, Jun 18, 2013 at 11:47 PM, Gleb Natapov <gleb@redhat.com> wrot=
e:
> > On Tue, Jun 18, 2013 at 10:28:59PM +0800, =C3=8A??=C3=8A?=E2=80=A2=C3=
=82=E2=80=A2? <Arthur Chunqi Li> wrote:
> >> On Tue, Jun 18, 2013 at 8:45 PM, Gleb Natapov <gleb@redhat.com> wr=
ote:
> >> > On Thu, Jun 13, 2013 at 05:30:03PM +0800, =C3=8A=C3=B9=C3=A9=C3=8A=
=C3=B2=E2=80=A2=C3=82=E2=80=A2=C3=A1 <Arthur Chunqi Li> wrote:
> >> >> Hi Gleb,
> >> >> I'm trying to solve these problems in the past days and meet ma=
ny
> >> >> difficulties. You want to save all the general registers in cal=
ling
> >> >> insn_page, so registers should be saved to (save) in insn_page.
> >> >> Because all the instructions should be generated outside and co=
py to
> >> >> insn_page, and the instructions generated outside is RIP-relati=
ve, so
> >> >> inside insn_page (save) will be wrong pointed with RIP-relative=
 code.
> >> >>
> >> > They do not have to be generated outside. You can write code int=
o
> >> > insn_page directly. Something like this outside of any functions=
:
> >> >
> >> > asm(".align 4096\n\t"
> >> >     ".global insn_page\n\t"
> >> >     ".global insn_page_end\n\t"
> >> >     ".global test_insn\n\t"
> >> >     ".global test_insn_end\n\t"
> >> >     "insn_page:"
> >> >     "mov %%rax, outregs \n\t"
> >> >     ...
> >> >     "test_insn:\n\t"
> >> >     "in (%ds), %al\n\t"
> >> >     ". =3D . + 31\n\t"
> >> >     "test_insn_end:\n\t"
> >> >     "mov outregs, %%rax\n\t"
> >> >     ...
> >> >     "ret\n\t"
> >> >     ".align 4096\n\t"
> >> >     "insn_page_end:\n\t");
> >> >
> >> > Now you copy that into alt_insn_page, put instruction you want t=
o test
> >> > into test_insn offset and remap alt_insn_page into "insn_page" v=
irtual address.
> >> I used such codes:
> >>
> >> invlpg((void *)virt_to_phys(insn_page));
> > virt_to_phys?
> This is a mistake, I changed it to "invlpg(insn_page)" but the result
> is the same.
> >
> >> asm volatile("call *%0" : : "r"(insn_page));
> >> install_page(cr3, virt_to_phys(alt_insn_page), insn_page);
> >> asm volatile("call *%0": : "r"(insn_page+1));
> > +1?
> Here I put "ret" on the first byte of insn_page, so the first call of
> "insn_page" can just return, and the second call of "insn_page+1=E2=80=
=9C will
> directly call the second byte, which is the real content of insn_page=
=2E
Send the code.

--
			Gleb.