From: Paolo Bonzini <pbonzini@redhat.com>
To: Richard Henderson <richard.henderson@linaro.org>
Cc: Eduardo Habkost <ehabkost@redhat.com>, kvm <kvm@vger.kernel.org>,
Michael Roth <michael.roth@amd.com>,
Marcelo Tosatti <mtosatti@redhat.com>,
qemu-devel <qemu-devel@nongnu.org>,
Cameron Esfahani <dirty@apple.com>,
David Edmondson <david.edmondson@oracle.com>,
Babu Moger <babu.moger@amd.com>,
Roman Bolshakov <r.bolshakov@yadro.com>
Subject: Re: [RFC PATCH 8/8] target/i386: Move X86XSaveArea into TCG
Date: Wed, 7 Jul 2021 08:51:08 +0200 [thread overview]
Message-ID: <CABgObfZ4aczr5HVEAOUgjrWWfT3m-NzLyXG2mX-86kf1Gd_Gqw@mail.gmail.com> (raw)
In-Reply-To: <0d75c3ab-926b-d4cd-244a-8c8b603535f9@linaro.org>
[-- Attachment #1: Type: text/plain, Size: 8166 bytes --]
Migration from KVM to TCG is broken anyway. The changing offsets do break
migration of a KVM guest from Intel to AMD or vice versa, because of the
difference in CPUID. That however is not changed by this patch.
Paolo
Il mer 7 lug 2021, 03:09 Richard Henderson <richard.henderson@linaro.org>
ha scritto:
> On 7/5/21 3:46 AM, David Edmondson wrote:
> > Given that TCG is now the only consumer of X86XSaveArea, move the
> > structure definition and associated offset declarations and checks to a
> > TCG specific header.
> >
> > Signed-off-by: David Edmondson <david.edmondson@oracle.com>
> > ---
> > target/i386/cpu.h | 57 ------------------------------------
> > target/i386/tcg/fpu_helper.c | 1 +
> > target/i386/tcg/tcg-cpu.h | 57 ++++++++++++++++++++++++++++++++++++
> > 3 files changed, 58 insertions(+), 57 deletions(-)
> >
> > diff --git a/target/i386/cpu.h b/target/i386/cpu.h
> > index 96b672f8bd..0f7ddbfeae 100644
> > --- a/target/i386/cpu.h
> > +++ b/target/i386/cpu.h
> > @@ -1305,48 +1305,6 @@ typedef struct XSavePKRU {
> > uint32_t padding;
> > } XSavePKRU;
> >
> > -#define XSAVE_FCW_FSW_OFFSET 0x000
> > -#define XSAVE_FTW_FOP_OFFSET 0x004
> > -#define XSAVE_CWD_RIP_OFFSET 0x008
> > -#define XSAVE_CWD_RDP_OFFSET 0x010
> > -#define XSAVE_MXCSR_OFFSET 0x018
> > -#define XSAVE_ST_SPACE_OFFSET 0x020
> > -#define XSAVE_XMM_SPACE_OFFSET 0x0a0
> > -#define XSAVE_XSTATE_BV_OFFSET 0x200
> > -#define XSAVE_AVX_OFFSET 0x240
> > -#define XSAVE_BNDREG_OFFSET 0x3c0
> > -#define XSAVE_BNDCSR_OFFSET 0x400
> > -#define XSAVE_OPMASK_OFFSET 0x440
> > -#define XSAVE_ZMM_HI256_OFFSET 0x480
> > -#define XSAVE_HI16_ZMM_OFFSET 0x680
> > -#define XSAVE_PKRU_OFFSET 0xa80
> > -
> > -typedef struct X86XSaveArea {
> > - X86LegacyXSaveArea legacy;
> > - X86XSaveHeader header;
> > -
> > - /* Extended save areas: */
> > -
> > - /* AVX State: */
> > - XSaveAVX avx_state;
> > -
> > - /* Ensure that XSaveBNDREG is properly aligned. */
> > - uint8_t padding[XSAVE_BNDREG_OFFSET
> > - - sizeof(X86LegacyXSaveArea)
> > - - sizeof(X86XSaveHeader)
> > - - sizeof(XSaveAVX)];
> > -
> > - /* MPX State: */
> > - XSaveBNDREG bndreg_state;
> > - XSaveBNDCSR bndcsr_state;
> > - /* AVX-512 State: */
> > - XSaveOpmask opmask_state;
> > - XSaveZMM_Hi256 zmm_hi256_state;
> > - XSaveHi16_ZMM hi16_zmm_state;
> > - /* PKRU State: */
> > - XSavePKRU pkru_state;
> > -} X86XSaveArea;
> > -
> > QEMU_BUILD_BUG_ON(sizeof(XSaveAVX) != 0x100);
> > QEMU_BUILD_BUG_ON(sizeof(XSaveBNDREG) != 0x40);
> > QEMU_BUILD_BUG_ON(sizeof(XSaveBNDCSR) != 0x40);
> > @@ -1355,21 +1313,6 @@ QEMU_BUILD_BUG_ON(sizeof(XSaveZMM_Hi256) !=
> 0x200);
> > QEMU_BUILD_BUG_ON(sizeof(XSaveHi16_ZMM) != 0x400);
> > QEMU_BUILD_BUG_ON(sizeof(XSavePKRU) != 0x8);
> >
> > -QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, legacy.fcw) !=
> XSAVE_FCW_FSW_OFFSET);
> > -QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, legacy.ftw) !=
> XSAVE_FTW_FOP_OFFSET);
> > -QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, legacy.fpip) !=
> XSAVE_CWD_RIP_OFFSET);
> > -QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, legacy.fpdp) !=
> XSAVE_CWD_RDP_OFFSET);
> > -QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, legacy.mxcsr) !=
> XSAVE_MXCSR_OFFSET);
> > -QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, legacy.fpregs) !=
> XSAVE_ST_SPACE_OFFSET);
> > -QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, legacy.xmm_regs) !=
> XSAVE_XMM_SPACE_OFFSET);
> > -QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, avx_state) !=
> XSAVE_AVX_OFFSET);
> > -QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, bndreg_state) !=
> XSAVE_BNDREG_OFFSET);
> > -QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, bndcsr_state) !=
> XSAVE_BNDCSR_OFFSET);
> > -QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, opmask_state) !=
> XSAVE_OPMASK_OFFSET);
> > -QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, zmm_hi256_state) !=
> XSAVE_ZMM_HI256_OFFSET);
> > -QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, hi16_zmm_state) !=
> XSAVE_HI16_ZMM_OFFSET);
> > -QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, pkru_state) !=
> XSAVE_PKRU_OFFSET);
> > -
> > typedef struct ExtSaveArea {
> > uint32_t feature, bits;
> > uint32_t offset, size;
> > diff --git a/target/i386/tcg/fpu_helper.c b/target/i386/tcg/fpu_helper.c
> > index 4e11965067..74bbe94b80 100644
> > --- a/target/i386/tcg/fpu_helper.c
> > +++ b/target/i386/tcg/fpu_helper.c
> > @@ -20,6 +20,7 @@
> > #include "qemu/osdep.h"
> > #include <math.h>
> > #include "cpu.h"
> > +#include "tcg-cpu.h"
> > #include "exec/helper-proto.h"
> > #include "fpu/softfloat.h"
> > #include "fpu/softfloat-macros.h"
> > diff --git a/target/i386/tcg/tcg-cpu.h b/target/i386/tcg/tcg-cpu.h
> > index 36bd300af0..53a8494455 100644
> > --- a/target/i386/tcg/tcg-cpu.h
> > +++ b/target/i386/tcg/tcg-cpu.h
> > @@ -19,6 +19,63 @@
> > #ifndef TCG_CPU_H
> > #define TCG_CPU_H
> >
> > +#define XSAVE_FCW_FSW_OFFSET 0x000
> > +#define XSAVE_FTW_FOP_OFFSET 0x004
> > +#define XSAVE_CWD_RIP_OFFSET 0x008
> > +#define XSAVE_CWD_RDP_OFFSET 0x010
> > +#define XSAVE_MXCSR_OFFSET 0x018
> > +#define XSAVE_ST_SPACE_OFFSET 0x020
> > +#define XSAVE_XMM_SPACE_OFFSET 0x0a0
> > +#define XSAVE_XSTATE_BV_OFFSET 0x200
> > +#define XSAVE_AVX_OFFSET 0x240
> > +#define XSAVE_BNDREG_OFFSET 0x3c0
> > +#define XSAVE_BNDCSR_OFFSET 0x400
> > +#define XSAVE_OPMASK_OFFSET 0x440
> > +#define XSAVE_ZMM_HI256_OFFSET 0x480
> > +#define XSAVE_HI16_ZMM_OFFSET 0x680
> > +#define XSAVE_PKRU_OFFSET 0xa80
> > +
> > +typedef struct X86XSaveArea {
> > + X86LegacyXSaveArea legacy;
> > + X86XSaveHeader header;
> > +
> > + /* Extended save areas: */
> > +
> > + /* AVX State: */
> > + XSaveAVX avx_state;
> > +
> > + /* Ensure that XSaveBNDREG is properly aligned. */
> > + uint8_t padding[XSAVE_BNDREG_OFFSET
> > + - sizeof(X86LegacyXSaveArea)
> > + - sizeof(X86XSaveHeader)
> > + - sizeof(XSaveAVX)];
> > +
> > + /* MPX State: */
> > + XSaveBNDREG bndreg_state;
> > + XSaveBNDCSR bndcsr_state;
> > + /* AVX-512 State: */
> > + XSaveOpmask opmask_state;
> > + XSaveZMM_Hi256 zmm_hi256_state;
> > + XSaveHi16_ZMM hi16_zmm_state;
> > + /* PKRU State: */
> > + XSavePKRU pkru_state;
> > +} X86XSaveArea;
> > +
> > +QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, legacy.fcw) !=
> XSAVE_FCW_FSW_OFFSET);
> > +QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, legacy.ftw) !=
> XSAVE_FTW_FOP_OFFSET);
> > +QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, legacy.fpip) !=
> XSAVE_CWD_RIP_OFFSET);
> > +QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, legacy.fpdp) !=
> XSAVE_CWD_RDP_OFFSET);
> > +QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, legacy.mxcsr) !=
> XSAVE_MXCSR_OFFSET);
> > +QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, legacy.fpregs) !=
> XSAVE_ST_SPACE_OFFSET);
> > +QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, legacy.xmm_regs) !=
> XSAVE_XMM_SPACE_OFFSET);
> > +QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, avx_state) !=
> XSAVE_AVX_OFFSET);
> > +QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, bndreg_state) !=
> XSAVE_BNDREG_OFFSET);
> > +QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, bndcsr_state) !=
> XSAVE_BNDCSR_OFFSET);
> > +QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, opmask_state) !=
> XSAVE_OPMASK_OFFSET);
> > +QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, zmm_hi256_state) !=
> XSAVE_ZMM_HI256_OFFSET);
> > +QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, hi16_zmm_state) !=
> XSAVE_HI16_ZMM_OFFSET);
> > +QEMU_BUILD_BUG_ON(offsetof(X86XSaveArea, pkru_state) !=
> XSAVE_PKRU_OFFSET);
>
> My only quibble is that these offsets are otherwise unused. This just
> becomes validation
> of compiler layout.
>
> I presume that XSAVE_BNDREG_OFFSET is not merely
> ROUND_UP(offsetof(avx_state) +
> sizeof(avx_state), some_pow2)?
>
> Do these offsets need to be migrated? Otherwise, how can one start a vm
> with kvm and then
> migrate to tcg? I presume the offsets above are constant for a given cpu,
> and that
> whatever cpu provides different offsets is not supported by tcg? Given
> the lack of avx,
> that's trivial these days...
>
>
> r~
>
>
[-- Attachment #2: Type: text/html, Size: 9874 bytes --]
next prev parent reply other threads:[~2021-07-07 6:53 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-07-05 10:46 [RFC PATCH 0/8] Derive XSAVE state component offsets from CPUID leaf 0xd where possible David Edmondson
2021-07-05 10:46 ` [RFC PATCH 1/8] target/i386: Declare constants for XSAVE offsets David Edmondson
2021-07-05 10:46 ` [RFC PATCH 2/8] target/i386: Consolidate the X86XSaveArea offset checks David Edmondson
2021-07-05 10:46 ` [RFC PATCH 3/8] target/i386: Clarify the padding requirements of X86XSaveArea David Edmondson
2021-07-05 10:46 ` [RFC PATCH 4/8] target/i386: Pass buffer and length to XSAVE helper David Edmondson
2021-07-05 10:46 ` [RFC PATCH 5/8] target/i386: Make x86_ext_save_areas visible outside cpu.c David Edmondson
2021-07-05 10:46 ` [RFC PATCH 6/8] target/i386: Observe XSAVE state area offsets David Edmondson
2021-07-05 10:46 ` [RFC PATCH 7/8] target/i386: Populate x86_ext_save_areas offsets using cpuid where possible David Edmondson
2021-07-05 10:46 ` [RFC PATCH 8/8] target/i386: Move X86XSaveArea into TCG David Edmondson
2021-07-07 1:09 ` Richard Henderson
2021-07-07 6:51 ` Paolo Bonzini [this message]
2021-07-07 10:10 ` David Edmondson
2021-07-08 7:45 ` David Edmondson
2021-07-08 15:22 ` Richard Henderson
2021-07-08 16:13 ` David Edmondson
2021-07-05 16:57 ` [RFC PATCH 0/8] Derive XSAVE state component offsets from CPUID leaf 0xd where possible Paolo Bonzini
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CABgObfZ4aczr5HVEAOUgjrWWfT3m-NzLyXG2mX-86kf1Gd_Gqw@mail.gmail.com \
--to=pbonzini@redhat.com \
--cc=babu.moger@amd.com \
--cc=david.edmondson@oracle.com \
--cc=dirty@apple.com \
--cc=ehabkost@redhat.com \
--cc=kvm@vger.kernel.org \
--cc=michael.roth@amd.com \
--cc=mtosatti@redhat.com \
--cc=qemu-devel@nongnu.org \
--cc=r.bolshakov@yadro.com \
--cc=richard.henderson@linaro.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).