From: "Pavel Dovgaluk" <Pavel.Dovgaluk@ispras.ru>
References: <20141107103123.6136.18545.stgit@PASHA-ISP>
	<20141107103223.6136.57870.stgit@PASHA-ISP>
	<545CAB18.3020503@redhat.com>
In-Reply-To: <545CAB18.3020503@redhat.com>
Date: Fri, 7 Nov 2014 14:39:12 +0300
Message-ID: <003c01cffa7f$73af90b0$5b0eb210$@Dovgaluk@ispras.ru>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Content-Language: ru
Subject: Re: [Qemu-devel] [RFC PATCH v4 10/25] i386: do not cross the pages
	boundaries in replay mode
To: 'Paolo Bonzini' <pbonzini@redhat.com>, qemu-devel@nongnu.org
Cc: peter.maydell@linaro.org, peter.crosthwaite@xilinx.com, alex.bennee@linaro.org, mark.burton@greensocs.com, real@ispras.ru, batuzovk@ispras.ru, maria.klimushenkova@ispras.ru, afaerber@suse.de, fred.konrad@greensocs.com

> From: Paolo Bonzini [mailto:pbonzini@redhat.com]
> On 07/11/2014 11:32, Pavel Dovgalyuk wrote:
> > This patch denies crossing the boundary of the pages in the replay =
mode,
> > because it can cause an exception. Do it only when boundary is
> > crossed by the first instruction in the block.
> > If current instruction already crossed the bound - it's ok,
> > because an exception hasn't stopped this code.
> >
> > Signed-off-by: Pavel Dovgalyuk <pavel.dovgaluk@ispras.ru>
> > ---
> >  target-i386/cpu.h       |    7 +++++++
> >  target-i386/translate.c |   14 ++++++++++++++
> >  2 files changed, 21 insertions(+), 0 deletions(-)
> >
> > diff --git a/target-i386/cpu.h b/target-i386/cpu.h
> > index 2968749..bc3f9f5 100644
> > --- a/target-i386/cpu.h
> > +++ b/target-i386/cpu.h
> > @@ -28,6 +28,13 @@
> >  #define TARGET_LONG_BITS 32
> >  #endif
> >
> > +/* Maximum instruction code size */
> > +#ifdef TARGET_X86_64
> > +#define TARGET_MAX_INSN_SIZE 16
> > +#else
> > +#define TARGET_MAX_INSN_SIZE 16
> > +#endif
> > +
> >  /* target supports implicit self modifying code */
> >  #define TARGET_HAS_SMC
> >  /* support for self modifying code even if the modified instruction =
is
> > diff --git a/target-i386/translate.c b/target-i386/translate.c
> > index 4d5dfb3..a264908 100644
> > --- a/target-i386/translate.c
> > +++ b/target-i386/translate.c
> > @@ -8035,6 +8035,20 @@ static inline void =
gen_intermediate_code_internal(X86CPU *cpu,
> >              gen_eob(dc);
> >              break;
> >          }
> > +        /* Do not cross the boundary of the pages in icount mode,
> > +           it can cause an exception. Do it only when boundary is
> > +           crossed by the first instruction in the block.
> > +           If current instruction already crossed the bound - it's =
ok,
> > +           because an exception hasn't stopped this code.
> > +         */
> > +        if (use_icount
> > +            && ((pc_ptr & TARGET_PAGE_MASK)
> > +                !=3D ((pc_ptr + TARGET_MAX_INSN_SIZE - 1) & =
TARGET_PAGE_MASK)
> > +                || (pc_ptr & ~TARGET_PAGE_MASK) =3D=3D 0)) {
> > +            gen_jmp_im(pc_ptr - dc->cs_base);
> > +            gen_eob(dc);
> > +            break;
> > +        }
> >          /* if too long translation, stop generation too */
> >          if (tcg_ctx.gen_opc_ptr >=3D gen_opc_end ||
> >              (pc_ptr - pc_start) >=3D (TARGET_PAGE_SIZE - 32) ||
> >
>=20
> Why only in icount mode?  Does it have a sensible performance problem?

Maybe. This is the same problem, which was discussed here for ARM: =
http://lists.gnu.org/archive/html/qemu-devel/2014-10/msg02232.html
But I haven't figured out what I have to do with it.

Pavel Dovgalyuk