To the device that contains them. This will allow for referencing
a GPIO input from it's canonical path (exciting for dynamic machine
generation!)
Reviewed-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NPeter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

Only allow a GPIO name to be one or the other. Inputs and outputs are
functionally different and should be in different namespaces. Prepares
support for the QOMification of IRQs as Links or Child objects.

The alternative is to munge names .e.g. with "-in" or "-out" suffixes
when giving QOM names. But that reduces clarity and if there are cases
out there where users want I and O with same name they can manually add
their own suffixes.
Reviewed-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NPeter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

At present, this function doesn't have partial cleanup implemented,
which will cause resource leaks in some scenarios.

Example:

1. Assume that "dc->realize(dev, &local_err)" executes successful
   and local_err == NULL;
2. device hotplug in hotplug_handler_plug() executes but fails
   (it is prone to occur). Then local_err != NULL;
3. error_propagate(errp, local_err) and return. But the resources
   which have been allocated in dc->realize() will be leaked.
Simple backtrace:
  dc->realize()
   |->device_realize
            |->pci_qdev_init()
                |->do_pci_register_device()
                |->etc.

Add fuller cleanup logic which assures that function can
goto appropriate error label as local_err population is
detected at each relevant point.
Signed-off-by: NGonglei <arei.gonglei@huawei.com>
Reviewed-by: NPeter Crosthwaite <peter.crosthwaite@xilinx.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: NAndreas Färber <afaerber@suse.de>

Forcefully unrealize all children regardless of errors in earlier
iterations (if any). We should keep going with cleanup operation
rather than report an error immediately. Therefore store the first
child unrealization failure and propagate it at the end. We also
forcefully unregister vmsd and unrealize actual object, too.
Signed-off-by: NGonglei <arei.gonglei@huawei.com>
Reviewed-by: NPeter Crosthwaite <peter.crosthwaite@xilinx.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: NAndreas Färber <afaerber@suse.de>

This error can not happen normally. If it happens, it indicates
something very wrong, we should abort QEMU. Moreover, the
user can only refer to /machine/peripheral or /objects, not
/machine/unattached.

While at it, remove superfluous check about local_err.
Signed-off-by: NGonglei <arei.gonglei@huawei.com>
Reviewed-by: NPeter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

It would be much better if we didn't terminate QEMU inside
device_post_init(), but at least exiting cleanly is better than aborting
and dumping core.

Before this patch:

    $ qemu-system-x86_64 -global cpu.xxx=y
    qemu-system-x86_64: Property '.xxx' not found
    Aborted (core dumped)

After this patch:

    $ qemu-system-x86_64 -global cpu.xxx=y
    qemu-system-x86_64: Property '.xxx' not found
Reviewed-by: NMichael S. Tsirkin <mst@redhat.com>
Reviewed-By: NIgor Mammedov <imammedo@redhat.com>
Signed-off-by: NEduardo Habkost <ehabkost@redhat.com>
Reviewed-by: NMichael S. Tsirkin <mst@redhat.com>
Signed-off-by: NMichael S. Tsirkin <mst@redhat.com>
Reviewed-by: NMichael S. Tsirkin <mst@redhat.com>
Signed-off-by: NEduardo Habkost <ehabkost@redhat.com>

When a device is unparented (i.e. made completely hidden from management)
we want to send a DEVICE_DELETED event only if the device actually was
realized.  This avoids raising DEVICE_DELETED events when device_add
fails.

However, this does not work right for recursively-deleted
devices: the whole tree is _first_ unrealized, _then_ unparented.
Then device_unparent sees realized==false and fails to trigger
the event.  The solution is simply to move have_realized into
the DeviceState struct.  If device_add fails, we never set the
new field to true and DEVICE_DELETED is not sent.

Fixes qemu-iotests testcase 067 (broken by commit 5942a190, though that
commit in turn fixed a possible segfault in the same test).
Reported-by: NMarkus Armbruster <armbru@redhat.com>
Reviewed-by: NMarkus Armbruster <armbru@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

The qdev_alias_all_properties() function creates QOM alias properties
for each qdev property on a DeviceState.  This is useful for parent
objects that wish to forward property accesses to their children.
Signed-off-by: NStefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: NPeter Crosthwaite <peter.crosthwaite@xilinx.com>

Allocate each IRQ individually on array allocations. This prepares for
QOMification of IRQs, where pointers to individual IRQs may be taken
and handed around for usage as QOM Links. The g_renew() scheme used here
is too fragile and would break all existing links should an IRQ list
be extended.

We now have to pass the IRQ count to qemu_free_irqs(). We have so few
call sites however, so this change is reasonably trivial.

Cc: agarcia@igalia.com
Cc: mst@redhat.com
Reviewed-by: NPeter Maydell <peter.maydell@linaro.org>
Acked-by: NAlberto Garcia <agarcia@igalia.com>
Signed-off-by: NPeter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

Signed-off-by: NWenchao Xia <wenchaoqemu@gmail.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Reviewed-by: NEric Blake <eblake@redhat.com>
Signed-off-by: NLuiz Capitulino <lcapitulino@redhat.com>

When the patch was posted that became 5c21ce77 (qdev: Realize buses
on device realization, 2014-03-12), it included recursive realization
and unrealization of devices when the bus's "realized" property
was toggled.

However, due to the same old worries about recursive realization
and prerequisites not being realized yet, those hunks were dropped when
committing the patch.  Unfortunately, this causes a use-after-free bug
(easily reproduced by a PCI hot-unplug action).

Before the patch, device_unparent behaved as follows:

   for each child bus
     unparent bus ----------------------------.
     | for each child device                  |
     |   unparent device ---------------.     |
     |   | unrealize device             |     |
     |   | call dc->unparent            |     |
     |   '-------------------------------     |
     '----------------------------------------'
   unrealize device

After the patch, it behaves as follows instead:

   unrealize device --------------------.
   | for each child bus                 |
   |   unrealize bus               (A)  |
   '------------------------------------'
   for each child bus
     unparent bus ----------------------.
     | for each child device            |
     |   unrealize device          (B)  |
     |   call dc->unparent              |
     '----------------------------------'

At the step marked (B) the device might use data from the bus that is
not available anymore due to step (A).

To fix this, we need to unrealize devices before step (A).  To sidestep
concerns about recursive realization, only do recursive unrealization
and leave the "value && !bus->realized" case as it is.

The resulting flow is:

   for each child bus
     unrealize bus ---------------------.
     | for each child device            |
     |   unrealize device          (B)  |
     | call bc->unrealize          (A)  |
     '----------------------------------'
   unrealize device
   for each child bus
     unparent bus ----------------------.
     | for each child device            |
     |   unparent device                |
     '----------------------------------'

where everything is "powered down" before it is unassembled.

Cc: qemu-stable@nongnu.org
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Tested-by: NMichael S. Tsirkin <mst@redhat.com>
Reviewed-by: NMichael S. Tsirkin <mst@redhat.com>
Signed-off-by: NMichael S. Tsirkin <mst@redhat.com>
Reviewed-by: NAndreas Färber <afaerber@suse.de>

No semantic change.

Cc: qemu-stable@nongnu.org
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Tested-by: NMichael S. Tsirkin <mst@redhat.com>
Reviewed-by: NMichael S. Tsirkin <mst@redhat.com>
Signed-off-by: NMichael S. Tsirkin <mst@redhat.com>
Reviewed-by: NAndreas Färber <afaerber@suse.de>

so that management could detect via QOM interface if device was
hotplugged
Signed-off-by: NIgor Mammedov <imammedo@redhat.com>
Acked-by: NPeter Crosthwaite <peter.crosthwaite@xilinx.com>
Reviewed-by: NMichael S. Tsirkin <mst@redhat.com>
Signed-off-by: NMichael S. Tsirkin <mst@redhat.com>

Add get_hotplug_handler() method to machine, and
make bus-less device use it during hotplug
as a means to discover a hotplug handler controller.
The returned controller is used to perform hotplug
actions.
Signed-off-by: NIgor Mammedov <imammedo@redhat.com>
Acked-by: NPeter Crosthwaite <peter.crosthwaite@xilinx.com>
Reviewed-by: NMichael S. Tsirkin <mst@redhat.com>
Signed-off-by: NMichael S. Tsirkin <mst@redhat.com>

Implement named GPIOs on the Device layer. Listifies the existing GPIOs
stuff using string keys. Legacy un-named GPIOs are preserved by using
a NULL name string - they are just a single matchable element in the
name list.
Reviewed-by: NPeter Maydell <peter.maydell@linaro.org>
Signed-off-by: NPeter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

Signed-off-by: NStefan Weil <sw@weilnetz.de>
Signed-off-by: NMichael Tokarev <mjt@tls.msk.ru>

Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

QEMU supports firmware names for all devices in the QEMU tree but
some architectures expect some parts of firmware path names in different
format.

This introduces a firmware-pathname-change interface definition.
If some machines needs to redefine the firmware path format, it has
to add the TYPE_FW_PATH_PROVIDER interface to an object that is above
the device on the QOM tree (typically /machine).
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NAlexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

There are currently three types of object_property_add_link() callers:

1. The link property may be set at any time.
2. The link property of a DeviceState instance may only be set before
   realize.
3. The link property may never be set, it is read-only.

Something similar can already be achieved with
object_property_add_str()'s set() argument.  Follow its example and add
a check() argument to object_property_add_link().

Also provide default check() functions for case #1 and #2.  Case #3 is
covered by passing a NULL function pointer.

Cc: Peter Crosthwaite <peter.crosthwaite@petalogix.com>
Cc: Alexander Graf <agraf@suse.de>
Cc: Anthony Liguori <aliguori@amazon.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Signed-off-by: NStefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: NPaolo Bonzini <pbonzini@redhat.com>
[AF: Tweaked documentation comment]
Signed-off-by: NAndreas Färber <afaerber@suse.de>

Some object_property_add_link() callers expect property deletion to
unref the link property object.  Other callers expect to manage the
refcount themselves.  The former are currently broken and therefore leak
the link property object.

This patch adds a flags argument to object_property_add_link() so the
caller can specify which refcount behavior they require.  The new
OBJ_PROP_LINK_UNREF_ON_RELEASE flag causes the link pointer to be
unreferenced when the property is deleted.

This fixes refcount leaks in qdev.c, xilinx_axidma.c, xilinx_axienet.c,
s390-virtio-bus.c, virtio-pci.c, virtio-rng.c, and ui/console.c.

Rationale for refcount behavior:

 * hw/core/qdev.c
   - bus children are explicitly unreferenced, don't interfere
   - parent_bus is essentially a read-only property that doesn't hold a
     refcount, don't unref
   - hotplug_handler is leaked, do unref

 * hw/dma/xilinx_axidma.c
   - rx stream "dma" links are set using set_link, therefore they
     need unref
   - tx streams are set using set_link, therefore they need unref

 * hw/net/xilinx_axienet.c
   - same reasoning as hw/dma/xilinx_axidma.c

 * hw/pcmcia/pxa2xx.c
   - pxa2xx bypasses set_link and therefore does not use refcounts

 * hw/s390x/s390-virtio-bus.c
 * hw/virtio/virtio-pci.c
 * hw/virtio/virtio-rng.c
 * ui/console.c
   - set_link is used and there is no explicit unref, do unref

Cc: Peter Crosthwaite <peter.crosthwaite@petalogix.com>
Cc: Alexander Graf <agraf@suse.de>
Cc: Anthony Liguori <aliguori@amazon.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Signed-off-by: NStefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

Integrate (un)realization of child buses with realization/unrealization
of the device hosting them. Code in device_unparent() is reordered for
unrealization of buses to work as part of device unrealization.

That way no changes need to be made to bus instantiation.
Signed-off-by: NBandan Das <bsd@redhat.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

Add a "realized" property calling realize/unrealize hooks as for devices.
Signed-off-by: NBandan Das <bsd@redhat.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

Move setting DeviceClass::hotpluggable default from device's
class_base_init() to device's class_init().
Reported-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NIgor Mammedov <imammedo@redhat.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

Commit 1a37eca1 (qdev: add
"hotpluggable" property to Device) added a property "hotpluggable" to
each device, with its getter accessing parent_bus->allow_hotplug.

Add a NULL check.

Cc: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

When we have 2 separate qdev devices that both create a qbus of the
same type without specifying a bus name or device name, we end up
with two buses of the same name, such as ide.0 on the Mac machines:

  dev: macio-ide, id ""
    bus: ide.0
      type IDE
  dev: macio-ide, id ""
    bus: ide.0
      type IDE

If we now spawn a device that connects to a ide.0 the last created
bus gets the device, with the first created bus inaccessible to the
command line.

After some discussion on IRC we concluded that the best quick fix way
forward for this is to make automated bus-class type based allocation
count a global counter. That's what this patch implements. With this
we instead get

  dev: macio-ide, id ""
    bus: ide.1
      type IDE
  dev: macio-ide, id ""
    bus: ide.0
      type IDE

on the example mentioned above.

This also means that if you did -device ...,bus=ide.0 you got a device
on the first bus (the last created one) before this patch and get that
device on the second one (the first created one) now.  Breaks
migration unless you change bus=ide.0 to bus=ide.1 on the destination.

This is intended and makes the bus enumeration work as expected.

As per review request follows a list of otherwise affected boards and
the reasoning for the conclusion that they are ok:

   target      machine         bus id              times
   ------      -------         ------              -----

   aarch64     n800            i2c-bus.0           2
   aarch64     n810            i2c-bus.0           2
   arm         n800            i2c-bus.0           2
   arm         n810            i2c-bus.0           2

-> Devices are only created explicitly on one of the two buses, using
   s->mpu->i2c[0], so no change to the guest.

   aarch64     vexpress-a15    virtio-mmio-bus.0   4
   aarch64     vexpress-a9     virtio-mmio-bus.0   4
   aarch64     virt            virtio-mmio-bus.0   32
   arm         vexpress-a15    virtio-mmio-bus.0   4
   arm         vexpress-a9     virtio-mmio-bus.0   4
   arm         virt            virtio-mmio-bus.0   32

-> Makes -device bus= work for all virtio-mmio buses.  Breaks
   migration.  Workaround for migration from old to new: specify
   virtio-mmio-bus.4 or .32 respectively rather than .0 on the
   destination.

   aarch64     xilinx-zynq-a9  usb-bus.0           2
   arm         xilinx-zynq-a9  usb-bus.0           2
   mips64el    fulong2e        usb-bus.0           2

-> Normal USB operation not affected. Migration driver needs command
   line to use the other bus.

   i386        isapc           ide.0               2
   x86_64      isapc           ide.0               2
   mips        mips            ide.0               2
   mips64      mips            ide.0               2
   mips64el    mips            ide.0               2
   mipsel      mips            ide.0               2
   ppc         g3beige         ide.0               2
   ppc         mac99           ide.0               2
   ppc         prep            ide.0               2
   ppc64       g3beige         ide.0               2
   ppc64       mac99           ide.0               2
   ppc64       prep            ide.0               2

-> Makes -device bus= work for all IDE buses.  Breaks migration.
   Workaround for migration from old to new: specify ide.1 rather than
   ide.0 on the destination.
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
Reviewed-by: NAndreas Faerber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

prop->info->legacy_name is still used by "-device foo,?".
Reviewed-by: NIgor Mammedov <imammedo@redhat.com>
Reviewed-by: NEric Blake <eblake@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

Reviewed-by: NIgor Mammedov <imammedo@redhat.com>
Reviewed-by: NEric Blake <eblake@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

make qdev_unplug()/device_set_realized() to call hotplug handler's
plug/unplug methods if available and remove not needed anymore
hot(un)plug handling from PCIDevice.

In case if hotplug handler is not available, revert to the legacy
hotplug method for compatibility with not yet converted buses.
Signed-off-by: NIgor Mammedov <imammedo@redhat.com>
Reviewed-by: NMichael S. Tsirkin <mst@redhat.com>
Signed-off-by: NMichael S. Tsirkin <mst@redhat.com>

Currently it's possible to make PCIDevice not hotpluggable
by using no_hotplug field of PCIDeviceClass. However it
limits this only to PCI devices and prevents from
generalizing hotplug code.

So add similar field to DeviceClass so it could be reused
with other Devices and would allow to replace PCI specific
hotplug callbacks with generic implementation. Following
patches will replace PCIDeviceClass.no_hotplug with this
new property.

In addition expose field as "hotpluggable" readonly property,
to make it possible to read its value via QOM interface.

Make DeviceClass hotpluggable by default as it was assumed
before.
Signed-off-by: NIgor Mammedov <imammedo@redhat.com>
Reviewed-by: NMichael S. Tsirkin <mst@redhat.com>
Signed-off-by: NMichael S. Tsirkin <mst@redhat.com>

It will allow to reuse field with different BUSes,
reducing code duplication. Field is intended for
replacing 'hotplug_qdev' field in PCIBus and also
will allow to avoid adding equivalent field to
DimmBus with possiblitity to refactor other BUSes
to use it instead of custom field.
In addition once all users of allow_hotplug field
are converted to new API, link could replace
allow_hotplug field in qdev hotplug code.
Signed-off-by: NIgor Mammedov <imammedo@redhat.com>
Reviewed-by: NMichael S. Tsirkin <mst@redhat.com>
Signed-off-by: NMichael S. Tsirkin <mst@redhat.com>

Replace assert_no_error() usages with the error_abort system.
&error_abort is passed into API calls to signal to the Error sub-system
that any errors are fatal. Removes need for caller assertions.
Signed-off-by: NPeter Crosthwaite <peter.crosthwaite@xilinx.com>
Reviewed-by: NMarkus Armbruster <armbru@redhat.com>
Signed-off-by: NLuiz Capitulino <lcapitulino@redhat.com>

This is unreachable code, as it's already asserted that no errors have
occurred. Delete.
Signed-off-by: NPeter Crosthwaite <peter.crosthwaite@xilinx.com>
Reviewed-by: NMarkus Armbruster <armbru@redhat.com>
Signed-off-by: NLuiz Capitulino <lcapitulino@redhat.com>

Same reasoning as commit 02a5c4c9
("qdev: Drop misleading qdev_free() function").  The qbus_free()
function removes the child from the namespace and decrements the
reference count.  It does not, however, guarantee to free the child
since the refcount may still be held.

Just call object_unparent() directly.
Suggested-by: NMarkus Armbruster <armbru@redhat.com>
Signed-off-by: NStefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

Post-order is the only sensible direction for the reset signals.
For example, suppose pre-order is used and the parent has some data
structures that cache children state (for example a list of active
requests).  When the reset method is invoked on the parent, these caches
could be in any state.

If post-order is used, on the other hand, these will be in a known state
when the reset method is invoked on the parent.

This change means that it is no longer possible to block the visit of
the devices, so the callback is changed to return void.  This is not
a problem, because PCI was returning 1 exactly in order to achieve the
same ordering that this patch implements.

PCI can then rely on the qdev core having sent a "reset signal" (whatever
that means) to the device, and only do the PCI-specific initialization
with pci_do_device_reset.

MST: fixed up virtio-ccw
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NMichael S. Tsirkin <mst@redhat.com>

Resetting should be done in post-order, not pre-order.  However,
qdev_walk_children and qbus_walk_children do not allow this.  Fix
it by adding two extra arguments to the functions.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NMichael S. Tsirkin <mst@redhat.com>

The qdev_free() function name is misleading since all the function does
is unlink the device from its parent.  The device is not necessarily
freed.

The device will be freed when its QObject refcount reaches zero.  It is
usual for the parent (bus) to hold the final reference but there are
cases where something else holds a reference so "free" is a misleading
name.

Call object_unparent(obj) directly instead of having a qdev wrapper
function.
Signed-off-by: NStefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: NEric Blake <eblake@redhat.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

To be passed on to object_initialize_with_type().

Acked-by: Cornelia Huck <cornelia.huck@de.ibm.com> (virtio-ccw)
Signed-off-by: NAndreas Färber <afaerber@suse.de>

To be passed to object_initialize().

Since commit 39355c38 the argument is
void*, so drop some superfluous (BusState *) casts or direct parent
field usages.
Signed-off-by: NAndreas Färber <afaerber@suse.de>

This way, properties registered in the instance_init function of
child classes will be handled properly by qdev_prop_set_globals(), too.

Includes a unit test for the new functionality.
Signed-off-by: NEduardo Habkost <ehabkost@redhat.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

Recent virtio refactoring in QEMU made virtio-bus become the parent bus
of scsi-bus, and virtio-bus doesn't have get_fw_dev_path implementation,
typename will be added to fw_dev_path by default, the new fw_dev_path
could not be identified by seabios. It causes that bootindex parameter
of scsi device doesn't work.

This patch implements get_fw_dev_path() in BusClass, it will be called
if bus doesn't implement the method, tyename will be added to
fw_dev_path. If the implemented method returns NULL, nothing will be
added to fw_dev_path.

It also implements virtio_bus_get_fw_dev_path() to return NULL. Then
QEMU will still pass original style of fw_dev_path to seabios.
Signed-off-by: NAmos Kong <akong@redhat.com>
Reviewed-by: NPaolo Bonzini <pbonzini@redhat.com>
Reviewed-by: NLaszlo Ersek <lersek@redhat.com>
Message-id: 1369814202-10346-1-git-send-email-akong@redhat.com
--
v2: only add nothing to fw_dev_path when get_fw_dev_path() is
    implemented and returns NULL. then it will not effect other devices
    don't have get_fw_dev_path() implementation.
v3: implement default get_fw_dev_path() in BusClass
Signed-off-by: NAnthony Liguori <aliguori@us.ibm.com>