OSDN Git Service

qmiga/qemu.git
10 years agolinux-headers: update linux headers to kvm/next
Alexander Graf [Wed, 4 Jun 2014 10:02:00 +0000 (12:02 +0200)]
linux-headers: update linux headers to kvm/next

This updates the kvm headers to commit 820b3fcd in kvm/next.

Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agolinux-headers: include psci.h
Alexander Graf [Wed, 4 Jun 2014 10:03:03 +0000 (12:03 +0200)]
linux-headers: include psci.h

The kvm headers now have a dependency on psci.h, sync it into our linux
header copy as well.

Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoPPC: SPE: Fix high-bits bitmask
Alexander Graf [Wed, 4 Jun 2014 00:01:10 +0000 (02:01 +0200)]
PPC: SPE: Fix high-bits bitmask

The SPE emulation code wants to access the highest 32bits of a 64bit register
and uses the andi TCG instruction for that. Unfortunately it masked with the
wrong mask. Fix the mask to actually cover the upper 32 bits.

This fixes simple multiplication tests with SPE guests for me.

Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoPPC: e500: Fix TLB lookup for 32bit CPUs
Alexander Graf [Tue, 3 Jun 2014 22:27:31 +0000 (00:27 +0200)]
PPC: e500: Fix TLB lookup for 32bit CPUs

When we run 32bit guest CPUs (or 32bit guest code on 64bit CPUs) on
qemu-system-ppc64 the TLB lookup will use the full effective address
as pointer.

However, only the first 32bits are valid when MSR.CM = 0. Check for
that condition.

This makes QEMU boot an e500v2 guest with more than 1G of RAM for me.

Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agohw/pci-host/ppce500: Fix typo in vmstate definition
Peter Maydell [Thu, 29 May 2014 11:05:34 +0000 (12:05 +0100)]
hw/pci-host/ppce500: Fix typo in vmstate definition

Fix a typo in the ppce500_pci vmstate definition which meant that
we were migrating the struct pci_inbound using the vmstate for
pci_outbound. Fortunately the two structures have exactly the same
format at the moment (four uint32_ts) so this was harmless, and
we can correcting the typo without a migration compatibility
break because the vmstate name doesn't go out on the wire.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Store Quadword Conditional Drops Size Bit
Tom Musta [Thu, 29 May 2014 14:12:24 +0000 (09:12 -0500)]
target-ppc: Store Quadword Conditional Drops Size Bit

The size and register information are encoded into the reserve_info field
of CPU state in the store conditional translation code.  Specifically, the
size is shifted left by 5 bits (see target-ppc/translate.c gen_conditional_store).

The user-mode store conditional code erroneously extracts the size by ANDing
with a 4 bit mask; this breaks if size >= 16.

Eliminate the mask to make the extraction of size mirror its encoding.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Confirm That .bss Pages Are Valid
Tom Musta [Thu, 29 May 2014 14:12:23 +0000 (09:12 -0500)]
target-ppc: Confirm That .bss Pages Are Valid

The existing code does a check to ensure that a .bss region is properly
mmap'd.  When additional mmap is required, the (guest) pages are also
validated.  However, this code has a bug: when host page size is larger
than target page size, it is possible for the .bss pages to already be
(host) mapped but the guest .bss pages may not be valid.

The check to mmap additional space is separated from the flagging of the
target (guest) pages, thus ensuring that both aspects are done properly.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Support VSX in PPC User Mode
Tom Musta [Thu, 29 May 2014 14:12:22 +0000 (09:12 -0500)]
target-ppc: Support VSX in PPC User Mode

Some modern tool chains use VSX instructions.  Therefore attempt to enable the VSX MSR
bit by default, just like similar bits (FP, VEC, SPE, etc.).

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Add a new user mode target for little-endian PPC64.
Doug Kwan [Thu, 29 May 2014 14:12:21 +0000 (09:12 -0500)]
target-ppc: Add a new user mode target for little-endian PPC64.

Signed-off-by: Doug Kwan <dougkwan@google.com>
Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Allow little-endian user mode.
Doug Kwan [Thu, 29 May 2014 14:12:20 +0000 (09:12 -0500)]
target-ppc: Allow little-endian user mode.

This allows running PPC64 little-endian in user mode if target is configured
that way.  In PPC64 LE user mode we set MSR.LE during initialization.

Signed-off-by: Doug Kwan <dougkwan@google.com>
Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Support little-endian PPC64 in user mode.
Doug Kwan [Thu, 29 May 2014 14:12:19 +0000 (09:12 -0500)]
target-ppc: Support little-endian PPC64 in user mode.

Look at ELF header to determine ABI version on PPC64.  This is required
for executing the first instruction correctly.  Also print correct machine
name in uname() system call.

Signed-off-by: Doug Kwan <dougkwan@google.com>
Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoPPC: e500: Fix MMUCSR0 emulation
Alex Zuepke [Wed, 28 May 2014 17:25:36 +0000 (19:25 +0200)]
PPC: e500: Fix MMUCSR0 emulation

A  "mtspr SPRMMUCSR0, reg"  always flushed TLB0,
because it passed the SPR number 0x3f4 to the flush routine.
But we want to flush either TLB0 or TBL1 depending on the GPR value.

Signed-off-by: Alex Zuepke <alexander.zuepke@hs-rm.de>
[agraf: change subject line, fix TCGv size mismatch]
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr_iommu: Introduce bus_offset in sPAPRTCETable
Alexey Kardashevskiy [Tue, 27 May 2014 05:36:37 +0000 (15:36 +1000)]
spapr_iommu: Introduce bus_offset in sPAPRTCETable

This adds @bus_offset into sPAPRTCETable to tell where TCE table starts
from. It is set to 0 for emulated devices. Dynamic DMA windows will use
other offset.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr_iommu: Introduce page_shift in sPAPRTCETable
Alexey Kardashevskiy [Tue, 27 May 2014 05:36:36 +0000 (15:36 +1000)]
spapr_iommu: Introduce page_shift in sPAPRTCETable

At the moment only 4K pages are supported by sPAPRTCETable. Since sPAPR
spec allows other page sizes and we are going to implement them, we need
page size to be configrable.

This adds @page_shift into sPAPRTCETable and replaces SPAPR_TCE_PAGE_SHIFT
with it where it is possible.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr_iommu: Get rid of window_size in sPAPRTCETable
Alexey Kardashevskiy [Tue, 27 May 2014 05:36:35 +0000 (15:36 +1000)]
spapr_iommu: Get rid of window_size in sPAPRTCETable

This removes window_size as it is basically a copy of nb_table
shifted by SPAPR_TCE_PAGE_SHIFT. As new dynamic DMA windows are
going to support windows as big as the entire RAM and this number
will be bigger that 32 capacity, we will have to do something
about @window_size anyway and removal seems to be the right way to go.

This removes dma_window_start/dma_window_size from sPAPRPHBState as
they are no longer used.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr_iommu: Convert old qdev_init_nofail() to object_property_set_bool
Alexey Kardashevskiy [Tue, 27 May 2014 05:36:34 +0000 (15:36 +1000)]
spapr_iommu: Convert old qdev_init_nofail() to object_property_set_bool

qdev_init_nofail() was replaced by object_property_set_bool("realized")
all over the QEMU so do we.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr_pci: Allow multiple TCE tables per PHB
Alexey Kardashevskiy [Tue, 27 May 2014 05:36:33 +0000 (15:36 +1000)]
spapr_pci: Allow multiple TCE tables per PHB

At the moment sPAPRPHBState contains a @tcet pointer to the only
TCE table. However sPAPR spec allows having more than one DMA window.

Since the TCE object is already a child of SPAPR PHB object, there is
no need to keep an additional pointer to it in sPAPRPHBState so remove it.

This changes the way sPAPRPHBState::reset performs reset of sPAPRTCETable
objects.

This changes the default DMA window properties calculation.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr_pci: spapr_iommu: Make DMA window a subregion
Alexey Kardashevskiy [Tue, 27 May 2014 05:36:32 +0000 (15:36 +1000)]
spapr_pci: spapr_iommu: Make DMA window a subregion

Currently the default DMA window is represented by a single MemoryRegion.
However there can be more than just one window so we need
a "root" memory region to be separated from the actual DMA window(s).

This introduces a "root" IOMMU memory region and adds a subregion for
the default DMA 32bit window. Following patches will add other
subregion(s).

This initializes a default DMA window subregion size to the guest RAM
size as this window can be switched into "bypass" mode which implements
direct DMA mapping.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr_pci: Introduce a finish_realize() callback
Alexey Kardashevskiy [Tue, 27 May 2014 05:36:31 +0000 (15:36 +1000)]
spapr_pci: Introduce a finish_realize() callback

The spapr-pci PHB initializes IOMMU for emulated devices only.
The upcoming VFIO support will do it different. However both emulated
and VFIO PHB types share most of the initialization code.
For the type specific things a new finish_realize() callback is
introduced.

This introduces sPAPRPHBClass derived from PCIHostBridgeClass and
adds the callback pointer.

This implements finish_realize() for emulated devices.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
[agraf: Fix compilation]
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr_iommu: Enable multiple TCE requests
Alexey Kardashevskiy [Tue, 27 May 2014 05:36:30 +0000 (15:36 +1000)]
spapr_iommu: Enable multiple TCE requests

Currently only single TCE entry per request is supported (H_PUT_TCE).
However PAPR+ specification allows multiple entry requests such as
H_PUT_TCE_INDIRECT and H_STUFF_TCE. Having less transitions to the host
kernel via ioctls, support of these calls can accelerate IOMMU operations.

This implements H_STUFF_TCE and H_PUT_TCE_INDIRECT.

This advertises "multi-tce" capability to the guest if the host kernel
supports it (KVM_CAP_SPAPR_MULTITCE) or guest is running in TCG mode.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr: Enable dynamic change of the supported hypercalls list
Alexey Kardashevskiy [Tue, 27 May 2014 05:36:29 +0000 (15:36 +1000)]
spapr: Enable dynamic change of the supported hypercalls list

At the moment the "ibm,hypertas-functions" list is fixed. However some
calls should be listed there if they are supported by QEMU or the host
kernel.

This enables hyperrtas_prop to grow on stack by adding
a SPAPR_HYPERRTAS_ADD macro. "qemu,hypertas-functions" is converted as well.

The first user of this is going to be a "multi-tce" property.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agomacio: Fix timer endianness
Alexander Graf [Mon, 26 May 2014 23:59:06 +0000 (01:59 +0200)]
macio: Fix timer endianness

The timer registers on our KeyLargo macio emulation are read as byte reversed
from the big endian guest, so we better expose them endian reversed as well.

This fixes initial hickups of booting Mac OS X with -M mac99 for me.

Signed-off-by: Alexander Graf <agraf@suse.de>
Tested-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
10 years agomacio ide: Do remainder access asynchronously
Alexander Graf [Mon, 26 May 2014 08:27:58 +0000 (10:27 +0200)]
macio ide: Do remainder access asynchronously

The macio IDE controller has some pretty nasty magic in its implementation to
allow for unaligned sector accesses. We used to handle these accesses
synchronously inside the IO callback handler.

However, the block infrastructure changed below our feet and now it's impossible
to call a synchronous block read/write from the aio callback handler of a
previous block access.

Work around that limitation by making the unaligned handling bits also go
through our asynchronous handler.

This fixes booting Mac OS X for me.

Reported-by: John Arbuckle <programmingkidx@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Fix popcntb Opcode Bug
Tom Musta [Wed, 14 May 2014 17:14:42 +0000 (12:14 -0500)]
target-ppc: Fix popcntb Opcode Bug

The popcntb instruction is erroneously encoded with opcode extension (opc1,opc2) = (0x03,0x03).
Bits 21-30 of popcntb are 122 = 0b00011-0b11010 and therefore this should be encoded
as (opc1,opc2) = (0x1A, 0x03).

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr_iommu: Replace @instance_id with LIOBN for migration
Alexey Kardashevskiy [Mon, 12 May 2014 08:46:32 +0000 (18:46 +1000)]
spapr_iommu: Replace @instance_id with LIOBN for migration

SPAPR IOMMU is a bus-less device and therefore its only ID in
migration stream is an instance id which is not reliable ID
as it depends on the command line parameters order. Since
libvirt may change the order, we need something better than that.

This removes VMSD descriptor from the class definitiion and
registers it with @liobn as an intance ID to let the destination
side find the right device to receive migration data.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoKVM: PPC: Enable compatibility mode
Alexey Kardashevskiy [Fri, 23 May 2014 02:26:58 +0000 (12:26 +1000)]
KVM: PPC: Enable compatibility mode

The host kernel implements a KVM_REG_PPC_ARCH_COMPAT register which
this uses to enable a compatibility mode if any chosen.

This sets the KVM_REG_PPC_ARCH_COMPAT register in KVM. ppc_set_compat()
signals the caller if the mode cannot be enabled by the host kernel.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
[agraf: fix TCG compat setting]
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr: Implement processor compatibility in ibm, client-architecture-support
Alexey Kardashevskiy [Fri, 23 May 2014 02:26:57 +0000 (12:26 +1000)]
spapr: Implement processor compatibility in ibm, client-architecture-support

Modern Linux kernels support last POWERPC CPUs so when a kernel boots,
in most cases it can find a matching cpu_spec in the kernel's cpu_specs
list. However if the kernel is quite old, it may be missing a definition
of the actual CPU. To provide an ability for old kernels to work on modern
hardware, a Processor Compatibility Mode has been introduced
by the PowerISA specification.

>From the hardware prospective, it is supported by the Processor
Compatibility Register (PCR) which is defined in PowerISA. The register
enables one of the compatibility modes (2.05/2.06/2.07).
Since PCR is a hypervisor privileged register and cannot be
directly accessed from the guest, the mode selection is done via
ibm,client-architecture-support (CAS) RTAS call using which the guest
specifies what "raw" and "architected" CPU versions it supports.
QEMU works out the best match, changes a "cpu-version" property of
every CPU and notifies the guest about the change by setting these
properties in the buffer passed as a response on a custom H_CAS hypercall.

This implements ibm,client-architecture-support parameters parsing
(now only for PVRs) and cooks the device tree diff with new values for
"cpu-version", "ibm,ppc-interrupt-server#s" and
"ibm,ppc-interrupt-server#s" properties.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr: Limit threads per core according to current compatibility mode
Alexey Kardashevskiy [Fri, 23 May 2014 02:26:56 +0000 (12:26 +1000)]
spapr: Limit threads per core according to current compatibility mode

This puts a limit to the number of threads per core based on the current
compatibility mode. Although PowerISA specs do not specify the maximum
threads per core number, the linux guest still expects that
PowerISA2.05-compatible CPU supports only 2 threads per core as this
is what POWER6 (2.05 compliant CPU) implements, the same is for
POWER7 (2.06, 4 threads) and POWER8 (2.07, 8 threads).

This calls spapr_fixup_cpu_smt_dt() with the maximum allowed number of
threads which affects ibm,ppc-interrupt-server#s and
ibm,ppc-interrupt-gserver#s properties.

The number of CPU nodesremains unchanged.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr: Rework spapr_fixup_cpu_dt()
Alexey Kardashevskiy [Fri, 23 May 2014 02:26:55 +0000 (12:26 +1000)]
spapr: Rework spapr_fixup_cpu_dt()

In PPC code we usually use the "cs" name for a CPUState* variables
and "cpu" for PowerPCCPU. So let's change spapr_fixup_cpu_dt() to
use same rules as spapr_create_fdt_skel() does.

This adds missing nodes creation if they do not already exist in
the current device tree, this is going to be used from
the client-architecture-support handler.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr: Add ibm, client-architecture-support call
Alexey Kardashevskiy [Fri, 23 May 2014 02:26:54 +0000 (12:26 +1000)]
spapr: Add ibm, client-architecture-support call

The PAPR+ specification defines a ibm,client-architecture-support (CAS)
RTAS call which purpose is to provide a negotiation mechanism for
the guest and the hypervisor to work out the best compatibility parameters.
During the negotiation process, the guest provides an array of various
options and capabilities which it supports, the hypervisor adjusts
the device tree and (optionally) reboots the guest.

At the moment the Linux guest calls CAS method at early boot so SLOF
gets called. SLOF allocates a memory buffer for the device tree changes
and calls a custom KVMPPC_H_CAS hypercall. QEMU parses the options,
composes a diff for the device tree, copies it to the buffer provided
by SLOF and returns to SLOF. SLOF updates the device tree and returns
control to the guest kernel. Only then the Linux guest parses the device
tree so it is possible to avoid unnecessary reboot in most cases.

The device tree diff is a header with an update format version
(defined as 1 in this patch) followed by a device tree with the properties
which require update.

If QEMU detects that it has to reboot the guest, it silently does so
as the guest expects reboot to happen because this is usual pHyp firmware
behavior.

This defines custom KVMPPC_H_CAS hypercall. The current SLOF already
has support for it.

This implements stub which returns very basic tree (root node,
no properties) to the guest.

As the return buffer does not contain any change, no change in behavior is
expected.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Define Processor Compatibility Masks
Alexey Kardashevskiy [Fri, 23 May 2014 02:26:53 +0000 (12:26 +1000)]
target-ppc: Define Processor Compatibility Masks

This introduces PCR mask for supported compatibility modes.
This will be used later by the ibm,client-architecture-support call.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Implement "compat" CPU option
Alexey Kardashevskiy [Fri, 23 May 2014 02:26:52 +0000 (12:26 +1000)]
target-ppc: Implement "compat" CPU option

This adds basic support for the "compat" CPU option. By specifying
the compat property, the user can manually switch guest CPU mode from
"raw" to "architected".

This defines feature disable bits which are not used yet as, for example,
PowerISA 2.07 says if 2.06 mode is selected, the TM bit does not matter -
transactional memory (TM) will be disabled because 2.06 does not define
it at all. The same is true for VSX and 2.05 mode. So just setting a mode
must be ok.

This does not change the existing behavior as the actual compatibility
mode support is coming in next patches.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
[agraf: fix compilation on 32bit hosts]
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr: Move SMT-related properties out of skeleton fdt
Alexey Kardashevskiy [Fri, 23 May 2014 02:26:51 +0000 (12:26 +1000)]
spapr: Move SMT-related properties out of skeleton fdt

The upcoming support of the "ibm,client-architecture-support"
reconfiguration call will be able to change dynamically the number
of threads per core (SMT mode). From the device tree prospective
this does not change the number of CPU nodes (as it is one node per
a CPU core) but affects content and size of the ibm,ppc-interrupt-server#s
and ibm,ppc-interrupt-gserver#s properties.

This moves ibm,ppc-interrupt-server#s and ibm,ppc-interrupt-gserver#s
out of the device tree skeleton.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Add "compat" CPU option
Alexey Kardashevskiy [Fri, 23 May 2014 02:26:50 +0000 (12:26 +1000)]
target-ppc: Add "compat" CPU option

PowerISA defines a compatibility mode for server POWERPC CPUs which
is supported by the PCR special register which is hypervisor privileged.
To support this mode for guests, SPAPR defines a set of virtual PVRs,
one per PowerISA spec version. When a hypervisor needs a guest to work in
a compatibility mode, it puts a virtual PVR value into @cpu-version
property of a CPU node.

This introduces a "compat" CPU option which defines maximal compatibility
mode enabled. The supported modes are power6/power7/power8.

This does not change the existing behaviour, new property will be used
by next patches.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoPPC: openpic_kvm: Implement reset
Alexander Graf [Thu, 22 May 2014 15:12:23 +0000 (17:12 +0200)]
PPC: openpic_kvm: Implement reset

When we trigger a system reset, the in-kernel openpic controller should also
get reset. This happens through a write to the GCR.RESET register which is
the same mechanism a guest would use to manually reset the device.

Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoopenpic: Reset IRQ source private members
Paul Janzen [Thu, 22 May 2014 06:09:45 +0000 (23:09 -0700)]
openpic: Reset IRQ source private members

The openpic emulation code maintains an allowable-CPU's bitmap
("destmask") for each IRQ source which is calculated from the IDR
register value whenever the guest OS writes to it.  However, if the
guest OS relies on the system to set the IDR register to a default
value at reset, and does not write IDR, then destmask does not get
updated, and interrupts do not get propagated to the guest.
Additionally, if an IRQ source is marked as critical, the source's
internal "output" and "nomask" fields are not correctly reset when the
PIC is reset.

Fix both these issues by calling write_IRQreg_idr from within
openpic_reset, instead of simply setting the IDR register to the
specified idr_reset value.

Signed-off-by: Paul Janzen <pcj@pauljanzen.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoopenpic: Move definition of openpic_reset
Paul Janzen [Thu, 22 May 2014 04:46:52 +0000 (21:46 -0700)]
openpic: Move definition of openpic_reset

This patch moves the definition of openpic_reset after the various
register read/write functions. No functional change.  It is in
preparation for using the register read/write functions in
openpic_reset.

Signed-off-by: Paul Janzen <pcj@pauljanzen.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Set the correct endianness in ELF dump header
Bharata B Rao [Mon, 19 May 2014 17:59:22 +0000 (19:59 +0200)]
target-ppc: Set the correct endianness in ELF dump header

Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com>
Reviewed-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce callback for interrupt endianness
Greg Kurz [Mon, 19 May 2014 17:59:05 +0000 (19:59 +0200)]
target-ppc: Introduce callback for interrupt endianness

POWER7, POWER7+ and POWER8 families use the ILE bit of the LPCR
special purpose register to decide the endianness to use when
entering interrupt handlers. When running a Linux guest, this
provides a hint on the endianness used by the kernel. And when
it comes to dumping a guest, the information is needed to write
ELF headers using the kernel endianness.

Suggested-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Reviewed-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
[agraf: change subject line]
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Support dump for little endian ppc64
Bharata B Rao [Mon, 19 May 2014 17:58:35 +0000 (19:58 +0200)]
target-ppc: Support dump for little endian ppc64

Fix ppc64 arch specific dump code to support all combinations of little/big
endian hosts/guests. FWIW the current code is broken for altivec registers
when guest and host have a different endianness: these 128-bit registers
are written to guest memory as a two 64-bit entities and we should also swap
them.

Unit testing was done with the following program provided by Tom Musta:

#include <stdio.h>
#include <stdint.h>
#include <stdlib.h>

int main(int argc, char** argv)
{

__uint128_t v = ((__uint128_t)0x0001020304050607ull << 64) |
0x08090a0b0c0d0e0full;

register void * vptr asm ("r11");
vptr = &v;

for(;;)
asm volatile ("lvx 30,0,11" );
}

When sending SIGABRT to this program and examining the core file, we get:

- ppc64  : 00 01 02 03 04 05 06 07 08 09 0a 0b 0c 0d 0e 0f
- ppc64le: 0f 0e 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00

We expect to find the very same layout in the QEMU dump since they are
real core files. This is what we get:

- ppc64 host, ppc64 guest   : 00 01 02 03 04 05 06 07 08 09 0a 0b 0c 0d 0e 0f
- ppc64 host, ppc64le guest : 0f 0e 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00
- x86_64 host, ppc64 guest  : 00 01 02 03 04 05 06 07 08 09 0a 0b 0c 0d 0e 0f
- x86_64 host, ppc64le guest: 0f 0e 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00

We introduce a NoteFuncArg type to avoid adding extra arguments to all note
functions.

Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com>
[ rebased on top of current master branch,
  introduced NoteFuncArg,
  use new cpu_to_dump{16,32,64} endian helpers,
  fix altivec support,
  Greg Kurz <gkurz@linux.vnet.ibm.com> ]
Reviewed-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agodump: Make DumpState and endian conversion routines available for arch-specific dump...
Bharata B Rao [Mon, 19 May 2014 17:57:44 +0000 (19:57 +0200)]
dump: Make DumpState and endian conversion routines available for arch-specific dump code

Make DumpState and endian conversion routines available for arch-specific dump
code by moving into dump.h. DumpState will be needed by arch-specific dump
code to access target endian information from DumpState->ArchDumpInfo. Also
break the dependency of dump.h from stubs/dump.c by creating a separate
dump-arch.h.

This patch doesn't change any functionality.

Signed-off-by: Bharata B Rao <bharata@linux.ibm.com>
[ rebased on top of current master branch,
  renamed endian helpers to cpu_to_dump{16,32,64},
  pass a DumpState * argument to endian helpers,
  Greg Kurz <gkurz@linux.vnet.ibm.com> ]
Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
[agraf: fix to apply]
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agomacio: handle non-block ATAPI DMA transfers
Mark Cave-Ayland [Sun, 18 May 2014 12:20:55 +0000 (13:20 +0100)]
macio: handle non-block ATAPI DMA transfers

Currently the macio DMA routines assume that all DMA requests are for read/write
block transfers. This is not always the case for ATAPI, for example when
requesting a TOC where the response is generated directly in the IDE buffer.

Detect these non-block ATAPI DMA transfers (where no lba is specified in the
command) and copy the results directly into RAM as indicated by the DBDMA
descriptor. This fixes CDROM access under MorphOS.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr: Add ibm, chip-id property in device tree
Alexey Kardashevskiy [Thu, 15 May 2014 11:23:14 +0000 (21:23 +1000)]
spapr: Add ibm, chip-id property in device tree

This adds a "ibm,chip-id" property for CPU nodes which should be the same
for all cores in the same CPU socket. The recent guest kernels use this
information to associate threads with sockets.

Refer to the kernel commit 256f2d4b463d3030ebc8d2b54f427543814a2bdc
for more details.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr: Add support for time base offset migration
Alexey Kardashevskiy [Thu, 1 May 2014 10:37:09 +0000 (20:37 +1000)]
spapr: Add support for time base offset migration

This allows guests to have a different timebase origin from the host.

This is needed for migration, where a guest can migrate from one host
to another and the two hosts might have a different timebase origin.
However, the timebase seen by the guest must not go backwards, and
should go forwards only by a small amount corresponding to the time
taken for the migration.

This is only supported for recent POWER hardware which has the TBU40
(timebase upper 40 bits) register. That includes POWER6, 7, 8 but not
970.

This adds kvm_access_one_reg() to access a special register which is not
in env->spr. This requires kvm_set_one_reg/kvm_get_one_reg patch.

The feature must be present in the host kernel.

This bumps vmstate_spapr::version_id and enables new vmstate_ppc_timebase
only for it. Since the vmstate_spapr::minimum_version_id remains
unchanged, migration from older QEMU is supported but without
vmstate_ppc_timebase.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoPPC: e500: Move to u-boot as firmware
Alexander Graf [Sun, 19 Jan 2014 23:27:25 +0000 (00:27 +0100)]
PPC: e500: Move to u-boot as firmware

Almost all platforms QEMU emulates have some sort of firmware they can load
to expose a guest environment that closely resembles the way it would look
like on real hardware.

This patch introduces such a firmware on our e500 platforms. U-boot is the
default firmware for most of these systems and as such our preferred choice.

For backwards compatibility reasons (and speed and simplicity) we skip u-boot
when you use -kernel and don't pass in -bios. For all other combinations like
-kernel and -bios or no -kernel you get u-boot as firmware.

This allows you to modify the boot environment, execute a networked boot through
the e1000 emulation and execute u-boot payloads.

Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoPPC: Add u-boot firmware for e500
Alexander Graf [Sun, 19 Jan 2014 23:25:40 +0000 (00:25 +0100)]
PPC: Add u-boot firmware for e500

This adds a special build of u-boot tailored for the e500 platforms we
emulate. It is based on the current version of upstream u-boot which
contains all the code necessary to drive our QEMU provided machines.

Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoPPC: e500: Expose kernel load address in dt
Alexander Graf [Sun, 19 Jan 2014 23:21:04 +0000 (00:21 +0100)]
PPC: e500: Expose kernel load address in dt

We want to move to a model where firmware loads our kernel. To achieve
this we need to be able to tell firmware where the kernel lies.

Let's copy the mechanism we already use for -M pseries and expose the
kernel load address and size through the device tree.

Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoPPC: Add dcbtls emulation
Alexander Graf [Sun, 19 Jan 2014 16:50:09 +0000 (17:50 +0100)]
PPC: Add dcbtls emulation

The dcbtls instruction is able to lock data inside the L1 cache.

Unfortunately we don't emulate any caches, so we have to tell the guest
that its locking attempt failed.

However, by implementing the instruction we at least don't give the
guest a program exception which it definitely does not expect.

Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoPPC: Properly emulate L1CSR0 and L1CSR1
Alexander Graf [Sun, 19 Jan 2014 16:49:11 +0000 (17:49 +0100)]
PPC: Properly emulate L1CSR0 and L1CSR1

There are 2 L1 cache control registers - one for data (L1CSR0) and
one for instructions (L1CSR1).

Emulate both of them well enough to give the guest the illusion that
it could actually do anything about its caches.

Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoPPC: Add L1CFG1 SPR emulation
Alexander Graf [Sun, 19 Jan 2014 16:47:43 +0000 (17:47 +0100)]
PPC: Add L1CFG1 SPR emulation

In addition to the L1 data cache configuration register L1CFG0 there is
also another one for the L1 instruction cache called L1CFG1.

Emulate that one with the same values as the data one.

Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoPPC: Fix SPR access control of L1CFG0
Alexander Graf [Thu, 23 Jan 2014 10:43:49 +0000 (11:43 +0100)]
PPC: Fix SPR access control of L1CFG0

The L1CFG0 register on e200 and e500 is "User RO" according to the
specifications. So let's make it user readable and world unwritable.

Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoPPC: Add definitions for GIVORs
Alexander Graf [Sun, 19 Jan 2014 16:41:14 +0000 (17:41 +0100)]
PPC: Add definitions for GIVORs

We're missing SPR definitions for GIVORs. Add them to the list of SPRs.

Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoPPC: Make all e500 CPUs SVR aware
Alexander Graf [Sun, 19 Jan 2014 16:39:54 +0000 (17:39 +0100)]
PPC: Make all e500 CPUs SVR aware

Our pre-e500mc e500 CPU types didn't get instanciated with SVR information,
even though those systems do support the SVR register.

Spawn them with the SVR tag so that they don't get confused when someone tries
to read SPR_SVR.

Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoPPC: Fail on leaking temporaries
Alexander Graf [Sun, 19 Jan 2014 16:28:33 +0000 (17:28 +0100)]
PPC: Fail on leaking temporaries

When QEMU gets compiled with --enable-debug-tcg we can check for temporary
leakage. Implement the necessary target code for this and fail emulation
when we hit a leakage.

This hopefully ensures that we don't get new leaks.

Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoPPC: Fix TCG chunks that don't free their temps
Alexander Graf [Sun, 19 Jan 2014 16:26:33 +0000 (17:26 +0100)]
PPC: Fix TCG chunks that don't free their temps

We want to make sure that every instruction cleans up after itself and
clears every temporary it allocated.

While checking whether this is already the case, I came across a few
cases where it isn't. This patch fixes every translation I found that
doesn't free their allocated temporaries.

Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoPPC: e500: implement PCI INTx routing
Bharat Bhushan [Mon, 12 May 2014 09:45:40 +0000 (15:15 +0530)]
PPC: e500: implement PCI INTx routing

This patch adds pci pin to irq_num routing callback.
This callback is called from pci_device_route_intx_to_irq to
find which pci device maps to which irq.
This fix is required for pci-device passthrough using vfio.

Also without this patch we gets below prints

"
  PCI: Bug - unimplemented PCI INTx routing (e500-pcihost)
  qemu-system-ppc64: PCI: Bug - unimplemented PCI INTx routing (e500-pcihost) "

and Legacy interrupt does not work with pci device passthrough.

Signed-off-by: Bharat Bhushan <Bharat.Bhushan@freescale.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
[agraf: remove double semicolon]
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoPPC: e500: some pci related cleanup
Bharat Bhushan [Mon, 12 May 2014 09:45:39 +0000 (15:15 +0530)]
PPC: e500: some pci related cleanup

- Use PCI_NUM_PINS rather than hardcoding
 - use "pin" wherever possible

Signed-off-by: Bharat Bhushan <Bharat.Bhushan@freescale.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoKVM: PPC: Don't secretly add 1T segment feature to CPU
Alexander Graf [Sun, 11 May 2014 16:37:00 +0000 (18:37 +0200)]
KVM: PPC: Don't secretly add 1T segment feature to CPU

When we select a CPU type that does not support 1TB segments, we should
not expose 1TB just because KVM supports 1TB segments. User configuration
always wins over feature availability.

Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Refactor AES Instructions
Tom Musta [Thu, 13 Mar 2014 14:13:30 +0000 (09:13 -0500)]
target-ppc: Refactor AES Instructions

This patch refactors the PowerPC Advanced Encryption Standard (AES) instructions
to use the common AES tables (include/qemu/aes.h).

Specifically:
    - vsbox is recoded to use the AES_sbox table.
    - vcipher, vcipherlast and vncipherlast are all recoded to use the optimized
      AES_t[ed][0-4] tables.
    - vncipher is recoded to use a combination of InvS-Box, InvShiftRows and
      InvMixColumns tables.  It was not possible to use AES_Td[0-4] due to a
      slight difference in how PowerPC implements vncipher.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-arm: Use Common Tables in AES Instructions
Tom Musta [Thu, 13 Mar 2014 14:13:29 +0000 (09:13 -0500)]
target-arm: Use Common Tables in AES Instructions

This patch refactors the ARM cryptographic instructions to use the
(newly) added common tables from include/qemu/aes.h.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-i386: Use Common ShiftRows and InvShiftRows Tables
Tom Musta [Thu, 13 Mar 2014 14:13:28 +0000 (09:13 -0500)]
target-i386: Use Common ShiftRows and InvShiftRows Tables

This patch eliminates the (now) redundant copy of the Advanced Encryption Standard (AES)
ShiftRows and InvShiftRows tables; the code is updated to use the common tables declared in
include/qemu/aes.h.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoutil: Add InvMixColumns
Tom Musta [Thu, 13 Mar 2014 14:13:27 +0000 (09:13 -0500)]
util: Add InvMixColumns

This patch adds the table implementation of the Advanced Encryption Standard (AES)
InvMixColumns transformation.

The patch is intentionally asymmetrical -- the MixColumns table is not added because
there is no known use for it at this time.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoutil: Add AES ShiftRows and InvShiftRows Tables
Tom Musta [Thu, 13 Mar 2014 14:13:26 +0000 (09:13 -0500)]
util: Add AES ShiftRows and InvShiftRows Tables

This patch adds tables that implement the Advanced Encryption Standard (AES) ShiftRows
and InvShiftRows transformations.  These are commonly used in instruction models.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agoutil: Add S-Box and InvS-Box Arrays to Common AES Utils
Tom Musta [Thu, 13 Mar 2014 14:13:25 +0000 (09:13 -0500)]
util: Add S-Box and InvS-Box Arrays to Common AES Utils

This patch adds tables for the S-Box and InvS-Box transformations commonly used by various
Advanced Encription Standard (AES) instruction models.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agospapr_pci: fix MSI limit
Alexey Kardashevskiy [Sun, 4 May 2014 14:09:48 +0000 (00:09 +1000)]
spapr_pci: fix MSI limit

At the moment XICS does not support interrupts reuse so sPAPR PHB
implements this. sPAPRPHBState holds array of 32 spapr_pci_msi to
describe PCI config address, first MSI and number of MSIs. Once
allocated for a device, QEMU tries reusing this config until the number
of MSIs changes.

Existing SPAPR guests call ibm,change-msi in a loop until the handler
returns the requested number of vectors.

Recently introduced check for the maximum number of MSI/MSIX vectors
supported by a device only works for a device which is new for PHB's
MSI cache. If it is already there, the check is not performed which
leads to new IRQ block allocation. This happens during PCI hotplug
even when the user hot plug the same device which he just hot unplugged.

This moves the check earlier.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Shift Significand
Tom Musta [Mon, 21 Apr 2014 20:55:21 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Shift Significand

Add emulation of the PowerPC Decimal Floating Point Shift Significand
Left Immediate (dscli[q][.]) and DFP Shift Significant Right Immediate
(dscri[q][.]) instructions.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Insert Biased Exponent
Tom Musta [Mon, 21 Apr 2014 20:55:20 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Insert Biased Exponent

Add emulation of the PowerPC Decimal Floating Point Insert Biased
Exponent instructions diex[q][.].

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Extract Biased Exponent
Tom Musta [Mon, 21 Apr 2014 20:55:19 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Extract Biased Exponent

Add emulation of the PowerPC Decimal Floating Point Extract
Biased Exponent instructions dxex[q][.].

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Encode BCD to DPD
Tom Musta [Mon, 21 Apr 2014 20:55:18 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Encode BCD to DPD

Add emulation of the PowerPC Decimal Floating Point Encode Binary
Coded Decimal to Densely Packed Decimal instructions denbcd[q][.].

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Decode DPD to BCD
Tom Musta [Mon, 21 Apr 2014 20:55:17 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Decode DPD to BCD

Add emulation of the Power PC Decimal Floating Point Decode
Densely Packed Decimal to Binary Coded Decimal instructions
ddedpd[q][.].

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Convert to Fixed
Tom Musta [Mon, 21 Apr 2014 20:55:16 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Convert to Fixed

Add emulation of the PowerPC Decimal Floating Point Convert to Fixed
instructions dctfix[q][.].

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Convert to Fixed
Tom Musta [Mon, 21 Apr 2014 20:55:15 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Convert to Fixed

Add emulation of the PowerPC Decimal Floating Point Convert to
Fixed instructions dctfix[q][.].

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce Round to DFP Short/Long
Tom Musta [Mon, 21 Apr 2014 20:55:14 +0000 (15:55 -0500)]
target-ppc: Introduce Round to DFP Short/Long

Add emulation of the PowerPC Round to DFP Short (drsp[.]) and Round to
DFP Long (drdpq[.]) instructions.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Convert to Long/Extended
Tom Musta [Mon, 21 Apr 2014 20:55:13 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Convert to Long/Extended

Add emulation of the PowerPC Convert to DFP Long (dctdp[.]) and
Convert to DFP Extended (dctqpq[.]) instructions.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Round to Integer
Tom Musta [Mon, 21 Apr 2014 20:55:12 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Round to Integer

Add emulation of the PowerPC Decimal Floating Point (DFP) Round
to FP Integer With Inexact (drintx[q][.]) and DFP Round to FP
Integer Without Inexact (drintn[q][.]) instructions.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Reround
Tom Musta [Mon, 21 Apr 2014 20:55:11 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Reround

Add emulation of the PowerPC Decimal Floating Point Reround instructions
drrnd[q][.].

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Quantize
Tom Musta [Mon, 21 Apr 2014 20:55:10 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Quantize

Add emulation of the PowerPC Decimal Floating Point Quantize instructions
dquai[q][.] and dqua[q][.].

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Test Significance
Tom Musta [Mon, 21 Apr 2014 20:55:09 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Test Significance

Add emulation of the PowerPC Decimal Floating Point Test Significance
instructions dtstsf[q][.].

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Test Exponent
Tom Musta [Mon, 21 Apr 2014 20:55:08 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Test Exponent

Add emulation of the PowerPC Decimal Floating Point Test Exponent
instructions dtstex[q][.].

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Test Data Group
Tom Musta [Mon, 21 Apr 2014 20:55:07 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Test Data Group

Add emulation of the PowerPC Decimal Floating Point Test Data
Group instructions dtstdg[q][.].

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Test Data Class
Tom Musta [Mon, 21 Apr 2014 20:55:06 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Test Data Class

Add emulation of the PowerPC Decimal Floating Point Test Data Class
instructions dtstdc[q][.].

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Compares
Tom Musta [Mon, 21 Apr 2014 20:55:05 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Compares

Add emulation of the PowerPC Decimal Floating Point Compare instructions
dcmpu[q] and dcmpo[q].

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Divide
Tom Musta [Mon, 21 Apr 2014 20:55:04 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Divide

Add emulation of the PowerPC Decimal Floating Point Divide instructions
ddiv[q][.]

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Multiply
Tom Musta [Mon, 21 Apr 2014 20:55:03 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Multiply

Add emulation of the PowerPC Decimal Floating Point Multiply instructions
dmul[q][.]

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Subtract
Tom Musta [Mon, 21 Apr 2014 20:55:02 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Subtract

Add emulation of the PowerPC Decimal Floating Point Subtract instructions
dsub[q][.]

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Add
Tom Musta [Mon, 21 Apr 2014 20:55:01 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Add

Add emulation of the PowerPC Decimal Floating Point Add instructions dadd[q][.]

Various GCC unused annotations are removed since it is now safe to remove them.

Signed-off-by: Tom Musta <tommusta@gmail.com>
[agraf: move brace in function definition]
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Post Processor Utilities
Tom Musta [Mon, 21 Apr 2014 20:55:00 +0000 (15:55 -0500)]
target-ppc: Introduce DFP Post Processor Utilities

Add post-processing utilities to the PowerPC Decimal Floating Point
(DFP) helper code.  Post-processors are small routines that execute
after a preliminary DFP result is computed.  They are used, among other
things, to compute status bits.

This change defines a function type for post processors as well as a
generic routine to run a list (array) of post-processors.

Actual post-processor implementations will be added as needed by specific
DFP helpers in subsequent changes.

Some routines are annotated with the GCC unused attribute in order to
preserve build bisection.  The annotation will be removed in subsequent
patches.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce DFP Helper Utilities
Tom Musta [Mon, 21 Apr 2014 20:54:59 +0000 (15:54 -0500)]
target-ppc: Introduce DFP Helper Utilities

Add a new file (dfp_helper.c) to the PowerPC implementation for Decimal Floating
Point (DFP) emulation.  This first version of the file declares a structure that
will be used by DFP helpers.  It also implements utilities that will initialize
such a structure for either a long (64 bit) DFP instruction or an extended (128
bit, aka "quad") instruction.

Some utility functions are annotated with the unused attribute in order to preserve
build bisection.

Signed-off-by: Tom Musta <tommusta@gmail.com>
[agraf: Add never reached assert on dfp_prepare_rounding_mode()]
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce Decoder Macros for DFP
Tom Musta [Mon, 21 Apr 2014 20:54:58 +0000 (15:54 -0500)]
target-ppc: Introduce Decoder Macros for DFP

Add decoder macros for the various Decimal Floating Point
instruction forms.  Illegal instruction masks are used to not only
guard against reserved instruction field use, but also to catch
illegal quad word forms that use odd-numbered floating point registers.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Introduce Generator Macros for DFP Arithmetic Forms
Tom Musta [Mon, 21 Apr 2014 20:54:57 +0000 (15:54 -0500)]
target-ppc: Introduce Generator Macros for DFP Arithmetic Forms

Add general support for generators of PowerPC Decimal Floating Point helpers.

Some utilities are annotated with GCC attribute unused in order to preserve
build bisection.  These annotations will be removed in later patches.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Define FPR Pointer Type for Helpers
Tom Musta [Mon, 21 Apr 2014 20:54:56 +0000 (15:54 -0500)]
target-ppc: Define FPR Pointer Type for Helpers

Define a floating pointer register pointer type in the PowerPC
helper header.  The type will be used to pass FPR register operands
to Decimal Floating Point (DFP) helpers.  A pointer is used because
the quad word forms of PowerPC DFP instructions operate on adjacent
pairs of floating point registers and thus can be thought of as
arrays of length 2.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agolibdecnumber: Fix decNumberSetBCD
Tom Musta [Mon, 21 Apr 2014 20:54:55 +0000 (15:54 -0500)]
libdecnumber: Fix decNumberSetBCD

Fix a simple bug in the decNumberSetBCD() function.  This function
encodes a decNumber with "n" BCD digits.  The original code erroneously
computed the number of declets from the dn argument, which is the output
decNumber value, and hence may contain garbage.  Instead, the input "n"
value is used.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agolibdecnumber: Introduce decNumberIntegralToInt64
Tom Musta [Mon, 21 Apr 2014 20:54:54 +0000 (15:54 -0500)]
libdecnumber: Introduce decNumberIntegralToInt64

Introduce a new conversion function to the libdecnumber library.
This function converts a decNumber to a signed 64-bit integer.
In order to support 64-bit integers (which may have up to 19
decimal digits), the existing "powers of 10" array is expanded
from 10 to 19 entries.

Signed-off-by: Tom Musta <tommusta@gmail.com>
[agraf: fix 32bit host compile]
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agolibdecnumber: Introduce decNumberFrom[U]Int64
Tom Musta [Mon, 21 Apr 2014 20:54:53 +0000 (15:54 -0500)]
libdecnumber: Introduce decNumberFrom[U]Int64

Introduce two conversion functions to the libdecnumber library.
These conversions transform 64 bit integers to the internal decNumber
representation.  Both a signed and unsigned version is added.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agotarget-ppc: Enable Building of libdecnumber
Tom Musta [Mon, 21 Apr 2014 20:54:52 +0000 (15:54 -0500)]
target-ppc: Enable Building of libdecnumber

Enable compilation of the newly added libdecnumber library code.
Object file targets are added to Makefile.target using a newly
introduced flag CONFIG_LIBDECNUMBER.  The flag is added
to the PowerPC targets (ppc[64]-linux-user, ppc[64]-softmmu).

Signed-off-by: Tom Musta <tommusta@gmail.com>
[agraf: add ppcemb and ppc64abi32 config]
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agolibdecnumber: Eliminate Unused Variable in decSetSubnormal
Tom Musta [Mon, 21 Apr 2014 20:54:51 +0000 (15:54 -0500)]
libdecnumber: Eliminate Unused Variable in decSetSubnormal

Eliminate an unused variable in the decSetSubnormal routine.  The
variable dnexp is declared and eventually set but never used, and
thus may trigger an unused-but-set-variable warning.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agolibdecnumber: Eliminate redundant declarations
Tom Musta [Mon, 21 Apr 2014 20:54:50 +0000 (15:54 -0500)]
libdecnumber: Eliminate redundant declarations

Eliminate redundant declarations of symbols DPD2BIN and BIN2DPD in
various .c source files.  These symbols are already declared in decDPD.h and
thus will trigger 'redundant redeclaration of ?XXX?' warnings, which, of
course, may fail QEMU compilation.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agolibdecnumber: Change gstdint.h to stdint.h
Tom Musta [Mon, 21 Apr 2014 20:54:49 +0000 (15:54 -0500)]
libdecnumber: Change gstdint.h to stdint.h

Replace the inclusion of gstdint.h with the standard stdint.h
header file.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agolibdecnumber: Modify dconfig.h to Integrate with QEMU
Tom Musta [Mon, 21 Apr 2014 20:54:48 +0000 (15:54 -0500)]
libdecnumber: Modify dconfig.h to Integrate with QEMU

Modify the dconfig.h header file so that libdecnumber code integrates QEMU
configuration.   Specifically:

  - the WORDS_BIGENDIAN preprocessor macro is used in libdecnumber code to
    determines endianness.  It is derived from the existing QEMU macro
    HOST_WORDS_BIGENDIAN which is defined in config-host.h.

  - the DECPUN macro determines the number of decimal digits (aka declets) per
    unit (byte).  This is 3 for PowerPC DFP.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
10 years agolibdecnumber: Prepare libdecnumber for QEMU include structure
Tom Musta [Mon, 21 Apr 2014 20:54:47 +0000 (15:54 -0500)]
libdecnumber: Prepare libdecnumber for QEMU include structure

Consistent with other libraries in QEMU, the libdecnumber header files were
placed in include/libdecnumber, separate from the C code.  This is different
from the original libdecnumber source, where they were co-located.

Change the libdecnumber source code so that it reflects this split.  Specifically,
modify directives of the form:

    #include "xxx.h"

to look like:

    #include "libdecnumber/xxx.h"

Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>