The value is not actually live across basic blocks, so there's no
need for the local property.  This eliminates storing the temporary
to its home location at the branch.
Signed-off-by: NRichard Henderson <rth@twiddle.net>
Reviewed-by: NPeter Maydell <peter.maydell@linaro.org>
Signed-off-by: NAurelien Jarno <aurelien@aurel32.net>

What 23bf49b5 really seemed to try to
fix is that Makefile could recurse into *-softmmu/ subdirectories before
a new *-softmmu/config-devices.mak was generated from pci.mak.

Fix this by adding a dependency on config-all-devices.mak, which in turn
has dependencies on *-softmmu/config-devices.mak.
Reported-by: NGerhard Wiesinger <lists@wiesinger.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NBlue Swirl <blauwirbel@gmail.com>

Placing the config-devices.mak.d file alongside the config-devices.mak
file in *-softmmu/ lead to it getting included into through
*-softmmu/Makefile in addition to ./Makefile, leading to confusion.

Instead, emit it to ./%-config-devices.mak.d, where it is included.
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NBlue Swirl <blauwirbel@gmail.com>

This reverts commit 23bf49b5.

While *-softmmu/config-devices.mak.d is included through *.d pattern via
Makefile.target, the make_devices_config.sh call these dependencies are
for is in ./Makefile. Therefore revert to original behavior.

This should unbreak pci.mak dependencies not propagating.
Reported-by: NPeter Maydell <peter.maydell@linaro.org>
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NBlue Swirl <blauwirbel@gmail.com>

* 's390-for-upstream' of git://github.com/agraf/qemu:
  s390/virtio-ccw: remove redundant call to blockdev_mark_auto_del
  s390/css: Fix subchannel detection
  Allow virtio-net features for legacy s390 virtio bus
  s390: virtio-ccw maintainer
  s390: simplify kvm cpu init

* 'ppc-for-upstream' of git://github.com/agraf/qemu: (66 commits)
  pseries: Add compatible property to root of device tree
  target-ppc: Move CPU aliases out of translate_init.c
  target-ppc: Report CPU aliases for QMP
  target-ppc: List alias names alongside CPU models
  target-ppc: Make host CPU a subclass of the host's CPU model
  PPC: xnu kernel expects FLUSH to be cleared on STOP
  PPC: Fix dma interrupt
  target-ppc: Fix PPC_DUMP_SPR_ACCESS build
  target-ppc: Synchronize FPU state with KVM
  target-ppc: Add mechanism for synchronizing SPRs with KVM
  Save memory allocation in the elf loader
  pseries: Implement h_read hcall
  target-ppc: Change "POWER7" CPU alias
  target-ppc: Fix remaining microcontroller typos among models
  target-ppc: Split model definitions out of translate_init.c
  target-ppc: Update Coding Style for CPU models
  target-ppc: Turn descriptive CPU model comments into device descriptions
  target-ppc: Turn descriptive CPU family comments into device descriptions
  target-ppc: Set remaining fields on CPU family classes
  target-ppc: Register all types for TARGET_PPCEMB
  ...

blockdev_mark_auto_del is already called in virtio-blk-exit.
Remove the redundant call.
Signed-off-by: NChristian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: NJens Freimann <jfrei@linux.vnet.ibm.com>
Signed-off-by: NAlexander Graf <agraf@suse.de>

We have to consider the m bit to find the real channel subsystem when
determining the last subchannel.

If we fail to take this into account, removal of a subchannel in
the middle of a big list of devices will stop device detection after
a reboot.
Signed-off-by: NChristian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: NJens Freimann <jfrei@linux.vnet.ibm.com>
Reviewed-by: NCornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Enable all virtio-net features for the legacy s390 virtio bus. This also fixes
kernel BUG at /usr/src/packages/BUILD/kernel-default-3.0.58/linux-3.0/drivers/s390/kvm/kvm_virtio.c:121!
Signed-off-by: NChristian Borntraeger <borntraeger@de.ibm.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: NAlexander Graf <agraf@suse.de>

Add myself as maintainer for virtio-ccw and the s390-ccw-virtio machine.
Signed-off-by: NCornelia Huck <cornelia.huck@de.ibm.com>
[agraf: add myself for virtio-ccw machine]
Signed-off-by: NAlexander Graf <agraf@suse.de>

There is no special code right now and the reset ioctl is done later
on in the the reset handler anyway. Lets simplify the cpu init.
Signed-off-by: NChristian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Currently, for the pseries machine the device tree supplied by qemu to SLOF
and from there to the guest does not include a 'compatible property' at the
root level.  Usually that works fine, since in this case the compatible
property doesn't really give any information not already found in the
'device_type' or 'model' properties.

However, the lack of 'compatible' confuses the bootloader install in the
SLES11 SP2 and SLES11 SP3 installers.  This patch therefore adds a token
'compatible' property to work around that.
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>
Cc: qemu-stable@nongnu.org
Signed-off-by: NAlexander Graf <agraf@suse.de>

Move array of CPU aliases to cpu-models.c, alongside model definitions.
This requires to zero-terminate the aliases array since ARRAY_SIZE() can
no longer be used in translate_init.c then.
Suggested-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

The QMP query-cpu-definitions implementation iterated over CPU classes
only, which were getting less and less as aliases were extracted.

Keep them in QMP as valid -cpu arguments even if not guaranteed stable.
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Revert adding a separate -cpu ? output section for aliases and list them
per CPU subclass.
Requested-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

This avoids assigning individual class fields and contributors
forgetting to add field assignments in KVM-only code.

ppc_cpu_class_find_by_pvr() requires the CPU model classes to be
registered, so defer host CPU type registration to kvm_arch_init().

Only register the host CPU type if there is a class with matching PVR.
This lets us drop error handling from instance_init.
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

otherwise it gets stuck in a loop
so clear it when unsetting run when flush is set

void
IODBDMAStop( volatile IODBDMAChannelRegisters *registers)
{

	IOSetDBDMAChannelControl( registers,
	IOClearDBDMAChannelControlBits( kdbdmaRun )
		| IOSetDBDMAChannelControlBits(  kdbdmaFlush ));

DBDMA: writel 0x0000000000000b00 <= 0xa0002000
DBDMA: channel 0x16 reg 0x0
DBDMA:     status 0x00002000

	while( IOGetDBDMAChannelStatus( registers) & (
			kdbdmaActive | kdbdmaFlush))
		eieio();

DBDMA: readl 0x0000000000000b04 => 0x00002000
DBDMA: channel 0x16 reg 0x1
DBDMA: readl 0x0000000000000b04 => 0x00002000
DBDMA: channel 0x16 reg 0x1
DBDMA: readl 0x0000000000000b04 => 0x00002000
DBDMA: channel 0x16 reg 0x1
DBDMA: readl 0x0000000000000b04 => 0x00002000
DBDMA: channel 0x16 reg 0x1
it continues to get printed

}
Signed-off-by: NAmadeusz Sławiński <amade@asmblr.net>
[agraf: replace tabs with spaces]
Signed-off-by: NAlexander Graf <agraf@suse.de>

In openbios (drivers/ide.c) they are set to

0000000d 00000000 00000002 00000000
0000000e 00000000 00000003 00000000
0000000f 00000000 00000004 00000000
(The last one seems to be not implemented in qemu)

It follows convention of how they are set on real machines,
both ide and dma ones are increased

Real machine one:
http://web.archive.org/web/20090107151044/http://penguinppc.org/historical/dev-trees-html/g4_agp_500_2.html
00000013 00000001 0000000b 00000000
00000014 00000001 0000000c 00000000
00000015 00000001 0000000d 00000000
Signed-off-by: NAmadeusz Sławiński <amade@asmblr.net>
Signed-off-by: NAlexander Graf <agraf@suse.de>

A victim of the d523dd00 AREG0
conversion, insert the missing cpu_env arguments.
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Currently qemu does not get and put the state of the floating point and
vector registers to KVM.  This is obviously a problem for savevm, as well
as possibly being problematic for debugging of FP-using guests.

This patch fixes this by using new extensions to the ONE_REG interface to
synchronize the qemu floating point state with KVM.
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Currently when runing under KVM on ppc, we synchronize a certain number of
vital SPRs to KVM through the SET_SREGS call.  This leaves out quite a lot
of important SPRs which are maintained in KVM.  It would be helpful to
have their contents in qemu for debugging purposes, and when we implement
migration it will be vital, since they include important guest state that
will need to be restored on the target.

This patch sets up for synchronization of any registers supported by the
KVM ONE_REG calls.  A new variant on spr_register() allows a ONE_REG id to
be stored with the SPR information.  When we set/get information to KVM
we also synchronize any SPRs so registered.

For now we set this mechanism up to synchronize a handful of important
registers that already have ONE_REG IDs, notably the DAR and DSISR.
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>
Signed-off-by: NAlexander Graf <agraf@suse.de>

The current elf loader uses too much memory. For example, I have a
executable with a bss section of 400 MB and I set the ram size to 512
MB. Qemu uses about 780MB of RAM (which is fine), but there's a peak at
1.6 GB during initialization (this is not fine).

This patch fixes two things:
 1) do not allocate each elf program twice.
 2) do not allocate memory for areas that are only zeros.

For this we need a new field in Rom: "datasize" which is the size of the
allocated data. If datasize is less than romsize, it means that the area
from datasize to romsize is filled with zeros.
Signed-off-by: NFabien Chouteau <chouteau@adacore.com>
Signed-off-by: NAlexander Graf <agraf@suse.de>

This h_call is useful for DLPAR in future amongst other things. Given an index
it fetches the corresponding PTE stored in the htab.
Signed-off-by: NErlon Cruz <erlon.cruz@br.flextronics.com>
Acked-by: NDavid Gibson <david@gibson.dropbear.id.au>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Let it resolve to v2.3 rather than v2.0.
Suggested-by: NDavid Gibson <david@gibson.dropbear.id.au>
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

controler -> controller
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Now that model definitions only reference their parent type, model
definitions are independent of the family definitions and can be
compiled independently of TCG translation.

Keep all #if defined(TODO) code local to cpu-models.c.
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Drop the space in #if defined (TODO).
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Fix microcontroller typo while at it.
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

This gets rid of some more overly long comments that have lost most of
their purpose now that in most cases there's only two functions left per
CPU family.

The class field is inherited by the actual CPU models, so override it.
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Now POWERPC_DEF_SVR() no longer sets family-specific fields itself.
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Don't attempt to suppress registration of CPU types, since the criteria
is actually a property of the class and should thus become a field.
Since we can't check a field set in a class_init function before
registering the type that leads to execution of that function, guard the
-cpu class lookup instead and suppress exposing these classes in -cpu ?
and in QMP.

In case someone tries to hot-add an incompatible CPU via device_add,
error out in realize.
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Instead of assigning *_<family> constants, set .parent to a family type.

Introduce a POWERPC_FAMILY() macro to keep type registration close to
its implementation. This macro will need tweaking later.
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Turn the array of model definitions into a set of self-registering QOM
types with their own class_init. Unique identifiers are obtained from
the combination of PVR, SVR and family identifiers; this requires all
alias #defines to be removed from the list. Possibly there are some more
left after this commit that are not currently being compiled.

Prepares for introducing abstract intermediate CPU types for families.

Keep the right-aligned macro line breaks within 78 chars to aid
three-way merges.
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

We are about to drop the redundant name field along with ppc_def_t.
Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Signed-off-by: NAndreas Färber <afaerber@suse.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>