Normally CPUState::cpu_index is used to pick the right CPU for various
operations. However default consecutive numbering does not always work
for POWERPC.
These indexes are reflected in /proc/device-tree/cpus/PowerPC,POWER7@XX
and used to call KVM VCPU's ioctls. In order to achieve this,
kvmppc_fixup_cpu() was introduced. Roughly speaking, it multiplies
cpu_index by the number of threads per core.
This approach has disadvantages such as:
1. NUMA configuration stays broken after the fixup;
2. CPU-targeted commands from the QEMU Monitor do not work properly as
CPU indexes have been fixed and there is no clear way for the user to
know what the new CPU indexes are.
This introduces a @cpu_dt_id field in the CPUPPCState struct which
is initialized from @cpu_index by default and can be fixed later
to meet the device tree requirements.
This adds an API to handle @cpu_dt_id.
This removes kvmppc_fixup_cpu() as it is not more needed, @cpu_dt_id
is calculated in ppc_cpu_realize().
Laurent Dufour [Fri, 21 Feb 2014 09:29:06 +0000 (10:29 +0100)]
target-ppc: Introduce hypervisor call H_GET_TCE
This patch introduces the hypervisor call H_GET_TCE which is basically the
reverse of H_PUT_TCE, as defined in the Power Architecture Platform
Requirements (PAPR).
The hcall H_GET_TCE is required by the kdump kernel which is calling it to
retrieve the TCE set up by the panicing kernel.
target-ppc: Update ppc_hash64_store_hpte to support updating in-kernel htab
This support updating htab managed by the hypervisor. Currently we don't have
any user for this feature. This actually bring the store_hpte interface
in-line with the load_hpte one. We may want to use this when we want to
emulate henter hcall in qemu for HV kvm.
target-ppc: Fix page table lookup with kvm enabled
With kvm enabled, we store the hash page table information in the hypervisor.
Use ioctl to read the htab contents. Without this we get the below error when
trying to read the guest address
Correctly update the htab_mask using the return value of
KVM_PPC_ALLOCATE_HTAB ioctl. Also we don't update sdr1
on GET_SREGS for HV. We check for external htab and if
found true, we don't need to update sdr1
Tom Musta [Mon, 24 Feb 2014 14:16:16 +0000 (08:16 -0600)]
target-ppc: Use Additional Temporary in stqcx Case
Per Alex Graf's suggestion, the recently added case to gen_conditional_store
for stqcx should use an additional temporary when accessing the second
doubleword. This avoids the mutation of the EA argument to the function,
which is counter intuitive.
spapr-vlan: flush queue whenever can_receive can go from false to true
When the guests adds buffers to receive queue, the network device
should flush its queue of pending packets. This is done with
qemu_flush_queued_packets.
This adds a call to qemu_flush_queued_packets() which wakes up the main
loop and let QEMU update the network device status which now is "can
receive". The patch basically does the same thing as e8b4c68 does.
Note that the implementation of vncipher deviates from the RTL in
ISA V2.07. However it does match the verbal description in the
third paragraph. The RTL will be fixed in ISA V2.07B. The
implementation here has been tested against actual P8 hardware.
Tom Musta [Wed, 12 Feb 2014 21:23:09 +0000 (15:23 -0600)]
target-ppc: Altivec 2.07: Change Bit Masks to Support 64-bit Rotates and Shifts
Existing code in the VROTATE, VSL and VSR macros for the Altivec rotate and shift
helpers uses a formula to compute a bit mask used to extract the rotate/shift
amount from the VRB register. What is desired is:
mask = (1 << (3 + log2(sizeof(element)))) - 1
but what is implemented is:
mask = (1 << (3 + (sizeof(element)/2))) - 1
This produces correct answers when "element" is uint8_t, uint16_t or uint_32t. But
it breaks down when element is uint64_t.
This patch corrects the situation. Since the mask is known at compile time, the
macros are changed to simply accept the mask as an argument.
Subsequent patches in this series will add double-word variants of rotates and
shifts and thus take advantage of this fix.
Tom Musta [Wed, 12 Feb 2014 21:23:01 +0000 (15:23 -0600)]
target-ppc: Altivec 2.07: Multiply Even/Odd Word Instructions
This patch adds the Multilpy Even/Odd Word instructions that are introduced
in Power ISA Version 2.07:
- Vector Multiply Even Unsigned Word (vmuleuw)
- Vector Multiply Even Signed Word (vmulesw)
- Vector Multiply Odd Unsigned Word (vmulouw)
- Vector Multiply Odd Signed Word (vmulosw)
Tom Musta [Wed, 12 Feb 2014 21:23:00 +0000 (15:23 -0600)]
target-ppc: Altivec 2.07: Change VMUL_DO to Support 64-bit Integers
This VMUL_DO macro provides support for the various vmule* and vmulo*
instructions. These instructions multiply vector elements, producing
products that are one size larger; e.g. vmuleub multiplies unsigned 8-bit
elements and produces a 16 bit unsigned element.
The existing macro works correctly for the existing instructions (8-bit,
and 16-bit source elements) but does not work correctly for 32-bit
source elements.
This patch adds an explicit cast to the multiplicands, forcing them to be
of the target element type. This is required for the forthcoming patches
that add the vmul[eo][us]w instructions.
Tom Musta [Wed, 12 Feb 2014 21:22:57 +0000 (15:22 -0600)]
target-ppc: Altivec 2.07: Add Support for R-Form Dual Instructions
Some Alitvec instructions introduced in Power ISA Version 2.07 use bit 31
(aka the "Rc" bit) as an opcode but also use bit 21 as an actual Rc
bit. QEMU for PowerPC typically uses bits 0-5 and 21-30 for opcodes.
This patch introduces a generator macro that injects an auxiliary handler
which decodes both bits 21 and 31 and invokes one of four standard
handlers. Since the instructions are not, in general, from the same version
of the ISA, two sets of PPC_*/PPC2_* flags are supported.
This patch also introduces a macro to insert two entries into the opcode
table -- one for bit 21 equal to 0 and one for bit 21 equal to 1.
Tom Musta [Wed, 12 Feb 2014 21:22:56 +0000 (15:22 -0600)]
target-ppc: Altivec 2.07: Add Opcode Macro for VX Form Instructions
This patch adds a macro to insert an entry into the opcode table for Altivec
Power ISA Version 2.07 instructions. The macro is similar to the GEN_VXFORM macro
except that it tags the entry with the PPC2_ALTIVEC_207 flag rather than
PPC_ALTIVEC.
Tom Musta [Wed, 12 Feb 2014 21:22:55 +0000 (15:22 -0600)]
target-ppc: Altivec 2.07: Add Support for Dual Altivec Instructions
Some Alitvec instructions introduced in Power ISA Version 2.07 use bit 31
(aka the "Rc" bit) as an opcode bit. However, QEMU for PowerPC uses
bits 0-5 and 21-30 for opcodes and not bit 31.
This patch introduces macros that will handle this situation by injecting
an auxiliary handler which decodes bit 31 in invokes one of two standard
handlers. Since the instructions are not, in general, from the same version
of the ISA, two sets of PPC_*/PPC2_* instruction tags are supported.
Tom Musta [Wed, 12 Feb 2014 21:22:53 +0000 (15:22 -0600)]
target-ppc: Altivec 2.07: Update AVR Structure
This patch updates the ppc_avr_t data structure to include elements for
signed 64-bit integers and (conditionally) unsigned 128 bit integers.
These elements will be in instructions models later on in this patch series.
Tom Musta [Mon, 10 Feb 2014 17:26:59 +0000 (11:26 -0600)]
target-ppc: Store Quadword
This patch adds support for the Store Quadword instruction in user mode. Prior
to Power ISA 2.07, stq was legal only in privileged mode. Support for Little
Endian mode is also new in ISA 2.07.
Tom Musta [Mon, 10 Feb 2014 17:26:58 +0000 (11:26 -0600)]
target-ppc: Load Quadword
This patch adds the Book I (user space) Load Quadword (lq) instruction.
This instruction was introduced into Book I in Power ISA V2.07. Previous
versions of the architecture supported this as a privileged instruction.
Previous versions of the architecture also did not support Little Endian
mode.
Note that this patch also adds the PPC_64BX flag to the Power8 model,
which enables the lq instruction.
Tom Musta [Mon, 10 Feb 2014 17:26:54 +0000 (11:26 -0600)]
target-ppc: Add Target Address SPR (TAR) to Power8
This patch adds support for the Target Address Register (TAR) to the Power8
model.
Because supported SPRs are typically identified in an init_proc_*()
function and because the Power8 model is currently just using the
init_proc_POWER7() function, a new init_proc_POWER8() function
is added and plugged into the P8 model.
Tom Musta [Mon, 10 Feb 2014 17:25:09 +0000 (11:25 -0600)]
target-ppc: Fix xxpermdi When T==A or T==B
The existing implementation of xxpermdi is defective if the target
VSR is also a source VSR. This patch fixes the defect in this case
but also preserves the simpler, two TCG operation implementation
when the target is not once of the two sources.
Alexander Graf [Thu, 6 Feb 2014 15:08:15 +0000 (16:08 +0100)]
qdev: Keep global allocation counter per bus
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.
-> 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.
The existing load_elf() just returns -1 if it fails to load ELF. However
it could be smarter than this and tell more about the failure such as
wrong endianness or incompatible platform.
This adds additional return codes for wrong architecture, wrong
endianness and if the image is not ELF at all.
This adds a load_elf_strerror() helper to convert return codes into
string messages.
This fixes handling of what load_elf() returns for s390x, other
callers just check the return value for <0 and this remains unchanged.
Currently everybody uses ELF kernel images with "-kernel" option on
pseries machine but QEMU still tries to boot from an image even it
fails to recognize it is ELF. This produces undefined behaviour if
the user tries a kernel image compiled for another architecture.
virtex_ml507: Add support for loading initrd images
Signed-off-by: Edgar E. Iglesias <[email protected]>
[agraf: fix up stray quotes and newlines in strings] Signed-off-by: Alexander Graf <[email protected]>
When ppc_store_slb() is called from kvm_arch_get_registers(), it stores
a SLB in CPUPPCState::slb[slot]. However it drops the slot number from
ESID so when kvm_arch_put_registers() puts SLBs back to KVM, they do not
have correct "index" field anymore. This broke migration with LPCR_AIR
enabled as now the guest is handling interrupts in virtual mode and unable
to reconstruct correct SLBs anymore.
This adds "index" field for valid SLBs when putting them to KVM.
Tom Musta [Tue, 7 Jan 2014 16:06:09 +0000 (10:06 -0600)]
target-ppc: Enable frsqrtes on Power7 and Power8
The frsqrtes instruction was introduced prior to ISA 2.06 and is
support on both the Power7 and Power8 processors. However, this
instruction is handled as illegal in the current QEMU emulation
machines. This patch enables the existing implemention of frsqrtes
in the P7 and P8 machines.
Tom Musta [Tue, 7 Jan 2014 16:06:05 +0000 (10:06 -0600)]
target-ppc: Fix and enable fri[mnpz]
The fri* series of instructions was introduced prior to ISA 2.06 and
is supported on Power7 and Power8 hardware. However, the instruction
is still considered illegal in the P7 and P8 QEMU emulation models.
This patch enables these instructions for the P7 and P8 machines.
Also, the existing helper is modified to correctly handle some of
the boundary cases (NaNs and the inexact flag).
Tom Musta [Tue, 7 Jan 2014 16:06:00 +0000 (10:06 -0600)]
target-ppc: Add ISA 2.06 fcfid[u][s] Instructions
This patch adds the fcfids, fcfidu and fcfidus instructions which
were introduced in Power ISA 2.06B. A common macro is provided to
eliminate repetitious code, and the existing fcfid instruction is
refactored to use this macro.
Tom Musta [Tue, 7 Jan 2014 16:05:59 +0000 (10:05 -0600)]
target-ppc: Add ISA2.06 Float to Integer Instructions
This patch adds the four floating point to integer conversion instructions
introduced by Power ISA V2.06:
- Floating Convert to Integer Word Unsigned (fctiwu)
- Floating Convert to Integer Word Unsigned with Round Toward
Zero (fctiwuz)
- Floating Convert to Integer Doubleword Unsigned (fctidu)
- Floating Convert to Integer Doubleword Unsigned with Round
Toward Zero (fctiduz)
A common macro is developed to eliminate repetitious code. Existing instructions
are also refactoried to use this macro (fctiw, fctiwz, fctid, fctidz).
Tom Musta [Tue, 7 Jan 2014 16:05:57 +0000 (10:05 -0600)]
target-ppc: Add ISA 2.06 stbcx. and sthcx. Instructions
This patch adds the byte and halfword variants of the Store Conditional
instructions. A common macro is introduced and the existing implementations
of stwcx. and stdcx. are refactored to use this macro.
Tom Musta [Tue, 7 Jan 2014 16:05:56 +0000 (10:05 -0600)]
target-ppc: Add ISA2.06 lbarx, lharx Instructions
This patch adds the byte and halfword variants of the Load and
Reserve instructions. Since there is much commonality among
all forms of Load and Reserve, a macro is provided and the existing
implementations of lwarx and ldarx are refactoried to use this
macro.
Tom Musta [Tue, 7 Jan 2014 16:05:51 +0000 (10:05 -0600)]
target-ppc: Add ISA2.06 divdeu[o] Instructions
This patch adds the Divide Doubleword Extended Unsigned
instructions. This instruction requires dividing a 128-bit
value by a 64 bit value. Since 128 bit integer division is
not supported in TCG, a helper is used. An architecture
independent 128-bit division routine is added to host-utils.
Tom Musta [Tue, 7 Jan 2014 16:05:50 +0000 (10:05 -0600)]
target-ppc: Add Flag for ISA2.06 Divide Extended Instructions
This patch adds a flag for the Divide Extended instructions that
were introduced in Power ISA V2.06B. The flag is added to the
Power7 and Power8 models.
Tom Musta [Wed, 15 Jan 2014 14:10:42 +0000 (08:10 -0600)]
target-ppc: Move To/From VSR Instructions
This patch adds the Move To VSR instructions (mfvsrd, mfvsrwz)
and Move From VSR instructions (mtvsrd, mtvsrwa, mtvsrwz). These
instructions are unusual in that they are considered a floating
point instruction if the indexed VSR is in the first half of the
array (0-31) but they are considered vector instructions if the
indexed VSR is in the second half of the array (32-63).
Tom Musta [Wed, 15 Jan 2014 14:10:40 +0000 (08:10 -0600)]
target-ppc: VSX Stage 4: Add xscvsxdsp and xscvuxdsp
This patch adds the VSX Scalar Convert Unsigned Integer Doubleword
to Floating Point Format and Round to Single Precision (xscvuxdsp)
and VSX Scalar Convert Signed Integer Douglbeword to Floating Point
Format and Round to Single Precision (xscvsxdsp) instructions.
The existing integer to floating point conversion macro (VSX_CVT_INT_TO_FP)
is modified to support the rounding of the intermediate floating point
result to single precision.
Tom Musta [Wed, 15 Jan 2014 14:10:30 +0000 (08:10 -0600)]
target-ppc: VSX Stage 4: Add lxsiwax, lxsiwzx and lxsspx
This patch adds the scalar load instructions introduced in ISA
V2.07:
- Load VSX Scalar as Integer Word Algebraic Indexd (lxsiwax)
- Load VSX Scalar as Integer Word and Zero Indexed (lxsiwzx)
- Load VSX Scalar Single-Precision Indexed (lxsspx)
Tom Musta [Wed, 15 Jan 2014 14:10:29 +0000 (08:10 -0600)]
target-ppc: VSX Stage 4: Refactor lxsdx
This patch refactors the lxsdx generator. Resuable code is isolated
into a macro. The macro will be used in subsequent patches in this
series to implement other scalar load instructions.
Tom Musta [Wed, 15 Jan 2014 14:10:28 +0000 (08:10 -0600)]
target-ppc: VSX Stage 4: Add VSX 2.07 Flag
This patch adds a flag to identify those VSX instructions that are
new to Power ISA V2.07. The flag is added to the Power 8 processor
initialization so that the P8 models understand how to decode and
emulate instructions in this category.
projects / qemu.git / log