qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Eduardo Habkost <ehabkost@redhat.com>
To: qemu-devel@nongnu.org
Cc: "Michael Mueller" <mimu@linux.vnet.ibm.com>,
	"Marcel Apfelbaum" <marcel.a@redhat.com>,
	"Alexander Graf" <agraf@suse.de>,
	"Christian Borntraeger" <borntraeger@de.ibm.com>,
	"Jason J. Herne" <jjherne@linux.vnet.ibm.com>,
	"Paolo Bonzini" <pbonzini@redhat.com>,
	"Andreas Färber" <afaerber@suse.de>
Subject: [Qemu-devel] [RFC 00/25] QOMify accelerator code
Date: Wed,  9 Jul 2014 19:03:57 -0300	[thread overview]
Message-ID: <1404943462-711-1-git-send-email-ehabkost@redhat.com> (raw)

Git tree:
  https://github.com/ehabkost/qemu-hacks/tree/accel-qom.v1

This is an attempt to convert the accel initialization and registration code
to be QOM-based. I see two use cases for this:

 * KVM-specific CPU compat bits;
 * Probing for accelerator-specific CPU model "runnable" information.

= KVM-specific compat bits =

Currently, the PC compat code that deals with KVM-specific CPUID compat bits
can't use compat_props to control compatibility, because it needs to affect the
X86CPU CPUID initialization behavior only when KVM is enabled. The solution I
propose is to have a "x86-kvm-accel" class which would be responsible for the
KVM-specific CPU initialization behavior. Any compatibility properties for that
behavior could then be tuned using "x86-kvm-accel" compat properties.

To be able to implement x86-specific accelerator methods, a TYPE_X86_ACCEL
interface was introduced.

Patches 19-24 implements this (except for the actual compat properties).

= Accel-specific runnable information =

There are plans to eventually make the "query-cpu-definitions" QMP command
return "runnable" information, which is accelerator-specific (e.g. some CPU
models may be runnable when using KVM, but not when using TCG). The existing
accelerator code maintains lots of global state, and the QOM-based approach will
allow us to eventually make query-cpu-definitions instantiate short-lived
accelerator objects (without affecting any machine and/or global state) to query
for accelerator-specific information.

Cc: Michael Mueller <mimu@linux.vnet.ibm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Alexander Graf <agraf@suse.de>
Cc: "Jason J. Herne" <jjherne@linux.vnet.ibm.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Andreas Färber <afaerber@suse.de>
Cc: Marcel Apfelbaum <marcel.a@redhat.com>

Eduardo Habkost (25):
  vl.c: Small coding style fix
  accel: Move accel code to accel.c
  accel: Create struct AccelType
  accel: Simplify configure_accelerator() using AccelType *acc variable
  accel: Move accel name lookup to separate function
  accel: Use QOM classes for accel types
  accel: Make AccelClass.available() optional
  accel: Move KVM accel registration to kvm-all.c
  accel: Move Xen registration code to xen-common.c
  accel: Move qtest accel registration to qtest.c
  accel: Remove tcg_available() function
  accel: Move accel init/allowed code to separate function
  accel: Rename 'init' method to 'init_machine'
  accel: Pass MachineState object to accel init functions
  accel: Create accel object when initializing machine
  accel: Save AccelState on MachineState when initializing
  kvm: Make KVMState be the TYPE_KVM_ACCEL instance struct
  accel: Get target name as argument when initializing accelerator
  accel: Use target-specific accel class if available
  accel: TYPE_X86_ACCEL interface
  kvm: x86 KVM accelerator subclass
  target-i386: Add AccelState parameter to cpu_x86_create()
  target-i386: Move accelerator-specific code outside
    X86CPU.instance_init
  target-i386: Accept "host" as value for CPU vendor
  target-i386: Move KVM CPUID hacking to accelerator cpu_post_init hook

 arch_init.c                |  11 +--
 hw/core/Makefile.objs      |   1 +
 hw/core/accel.c            | 171 +++++++++++++++++++++++++++++++++++++++++++++
 hw/i386/pc.c               |   3 +-
 include/hw/accel.h         |  60 ++++++++++++++++
 include/hw/boards.h        |   1 +
 include/hw/i386/accel.h    |  42 +++++++++++
 include/hw/xen/xen.h       |   1 -
 include/qemu/typedefs.h    |   5 ++
 include/sysemu/arch_init.h |   1 -
 include/sysemu/kvm.h       |   5 +-
 include/sysemu/qtest.h     |   1 -
 kvm-all.c                  |  40 +++++++++--
 kvm-stub.c                 |   5 --
 qtest.c                    |  27 ++++++-
 target-i386/Makefile.objs  |   1 +
 target-i386/accel.c        |  38 ++++++++++
 target-i386/cpu.c          | 101 +++++++++++---------------
 target-i386/cpu.h          |   2 +-
 target-i386/kvm.c          |  70 +++++++++++++++++++
 vl.c                       |  83 +---------------------
 xen-common-stub.c          |   6 --
 xen-common.c               |  25 ++++++-
 23 files changed, 528 insertions(+), 172 deletions(-)
 create mode 100644 hw/core/accel.c
 create mode 100644 include/hw/accel.h
 create mode 100644 include/hw/i386/accel.h
 create mode 100644 target-i386/accel.c

-- 
1.9.3

             reply	other threads:[~2014-07-09 22:04 UTC|newest]

Thread overview: 26+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-07-09 22:03 Eduardo Habkost [this message]
2014-07-09 22:03 ` [Qemu-devel] [RFC 01/25] vl.c: Small coding style fix Eduardo Habkost
2014-07-09 22:03 ` [Qemu-devel] [RFC 02/25] accel: Move accel code to accel.c Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 03/25] accel: Create struct AccelType Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 04/25] accel: Simplify configure_accelerator() using AccelType *acc variable Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 05/25] accel: Move accel name lookup to separate function Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 06/25] accel: Use QOM classes for accel types Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 07/25] accel: Make AccelClass.available() optional Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 08/25] accel: Move KVM accel registration to kvm-all.c Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 09/25] accel: Move Xen registration code to xen-common.c Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 10/25] accel: Move qtest accel registration to qtest.c Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 11/25] accel: Remove tcg_available() function Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 12/25] accel: Move accel init/allowed code to separate function Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 13/25] accel: Rename 'init' method to 'init_machine' Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 14/25] accel: Pass MachineState object to accel init functions Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 15/25] accel: Create accel object when initializing machine Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 16/25] accel: Save AccelState on MachineState when initializing Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 17/25] kvm: Make KVMState be the TYPE_KVM_ACCEL instance struct Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 18/25] accel: Get target name as argument when initializing accelerator Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 19/25] accel: Use target-specific accel class if available Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 20/25] accel: TYPE_X86_ACCEL interface Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 21/25] kvm: x86 KVM accelerator subclass Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 22/25] target-i386: Add AccelState parameter to cpu_x86_create() Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 23/25] target-i386: Move accelerator-specific code outside X86CPU.instance_init Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 24/25] target-i386: Accept "host" as value for CPU vendor Eduardo Habkost
2014-07-09 22:04 ` [Qemu-devel] [RFC 25/25] target-i386: Move KVM CPUID hacking to accelerator cpu_post_init hook Eduardo Habkost

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=1404943462-711-1-git-send-email-ehabkost@redhat.com \
    --to=ehabkost@redhat.com \
    --cc=afaerber@suse.de \
    --cc=agraf@suse.de \
    --cc=borntraeger@de.ibm.com \
    --cc=jjherne@linux.vnet.ibm.com \
    --cc=marcel.a@redhat.com \
    --cc=mimu@linux.vnet.ibm.com \
    --cc=pbonzini@redhat.com \
    --cc=qemu-devel@nongnu.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).