LAM uses CR3[61] and CR3[62] to configure/enable LAM on user pointers.
LAM uses CR4[28] to configure/enable LAM on supervisor pointers.
For CR3 LAM bits, no additional handling needed:
- TCG
LAM is not supported for TCG of target-i386. helper_write_crN() and
helper_vmrun() check max physical address bits before calling
cpu_x86_update_cr3(), no change needed, i.e. CR3 LAM bits are not allowed
to be set in TCG.
- gdbstub
x86_cpu_gdb_write_register() will call cpu_x86_update_cr3() to update cr3.
Allow gdb to set the LAM bit(s) to CR3, if vcpu doesn't support LAM,
KVM_SET_SREGS will fail as other reserved bits.
For CR4 LAM bit, its reservation depends on vcpu supporting LAM feature or
not.
- TCG
LAM is not supported for TCG of target-i386. helper_write_crN() and
helper_vmrun() check CR4 reserved bit before calling cpu_x86_update_cr4(),
i.e. CR4 LAM bit is not allowed to be set in TCG.
- gdbstub
x86_cpu_gdb_write_register() will call cpu_x86_update_cr4() to update cr4.
Mask out LAM bit on CR4 if vcpu doesn't support LAM.
- x86_cpu_reset_hold() doesn't need special handling.
Signed-off-by: Binbin Wu <binbin.wu@linux.intel.com>
Tested-by: Xuelian Guo <xuelian.guo@intel.com>
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Message-ID: <20240112060042.19925-3-binbin.wu@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Linear Address Masking (LAM) is a new Intel CPU feature, which allows
software to use of the untranslated address bits for metadata.
The bit definition:
CPUID.(EAX=7,ECX=1):EAX[26]
Add CPUID definition for LAM.
Note LAM feature is not supported for TCG of target-i386, LAM CPIUD bit
will not be added to TCG_7_1_EAX_FEATURES.
More info can be found in Intel ISE Chapter "LINEAR ADDRESS MASKING(LAM)"
https://cdrdv2.intel.com/v1/dl/getContent/671368
Signed-off-by: Robert Hoo <robert.hu@linux.intel.com>
Co-developed-by: Binbin Wu <binbin.wu@linux.intel.com>
Signed-off-by: Binbin Wu <binbin.wu@linux.intel.com>
Tested-by: Xuelian Guo <xuelian.guo@intel.com>
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Message-ID: <20240112060042.19925-2-binbin.wu@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The PCOMMIT instruction was never included in any physical processor.
TCG implements it as a no-op instruction, but its utility is debatable
to say the least. Drop it from the decoder since it is only available
with "-cpu max", which does not guarantee migration compatibility
across versions, and deprecate the property just in case someone is
using it as "pcommit=off".
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Observed the following failure while booting the SEV-SNP guest and the
guest fails to boot with the smp parameters:
"-smp 192,sockets=1,dies=12,cores=8,threads=2".
qemu-system-x86_64: sev_snp_launch_update: SNP_LAUNCH_UPDATE ret=-5 fw_error=22 'Invalid parameter'
qemu-system-x86_64: SEV-SNP: CPUID validation failed for function 0x8000001e, index: 0x0.
provided: eax:0x00000000, ebx: 0x00000100, ecx: 0x00000b00, edx: 0x00000000
expected: eax:0x00000000, ebx: 0x00000100, ecx: 0x00000300, edx: 0x00000000
qemu-system-x86_64: SEV-SNP: failed update CPUID page
Reason for the failure is due to overflowing of bits used for "Node per
processor" in CPUID Fn8000001E_ECX. This field's width is 3 bits wide and
can hold maximum value 0x7. With dies=12 (0xB), it overflows and spills
over into the reserved bits. In the case of SEV-SNP, this causes CPUID
enforcement failure and guest fails to boot.
The PPR documentation for CPUID_Fn8000001E_ECX [Node Identifiers]
=================================================================
Bits Description
31:11 Reserved.
10:8 NodesPerProcessor: Node per processor. Read-only.
ValidValues:
Value Description
0h 1 node per processor.
7h-1h Reserved.
7:0 NodeId: Node ID. Read-only. Reset: Fixed,XXh.
=================================================================
As in the spec, the valid value for "node per processor" is 0 and rest
are reserved.
Looking back at the history of decoding of CPUID_Fn8000001E_ECX, noticed
that there were cases where "node per processor" can be more than 1. It
is valid only for pre-F17h (pre-EPYC) architectures. For EPYC or later
CPUs, the linux kernel does not use this information to build the L3
topology.
Also noted that the CPUID Function 0x8000001E_ECX is available only when
TOPOEXT feature is enabled. This feature is enabled only for EPYC(F17h)
or later processors. So, previous generation of processors do not not
enumerate 0x8000001E_ECX leaf.
There could be some corner cases where the older guests could enable the
TOPOEXT feature by running with -cpu host, in which case legacy guests
might notice the topology change. To address those cases introduced a
new CPU property "legacy-multi-node". It will be true for older machine
types to maintain compatibility. By default, it will be false, so new
decoding will be used going forward.
The documentation is taken from Preliminary Processor Programming
Reference (PPR) for AMD Family 19h Model 11h, Revision B1 Processors 55901
Rev 0.25 - Oct 6, 2022.
Cc: qemu-stable@nongnu.org
Fixes: 31ada106d8 ("Simplify CPUID_8000_001E for AMD")
Link: https://bugzilla.kernel.org/show_bug.cgi?id=206537
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Signed-off-by: Babu Moger <babu.moger@amd.com>
Message-ID: <0ee4b0a8293188a53970a2b0e4f4ef713425055e.1714757834.git.babu.moger@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
accel/tcg/ files requires the following definitions:
- TARGET_LONG_BITS
- TARGET_PAGE_BITS
- TARGET_PHYS_ADDR_SPACE_BITS
- TCG_GUEST_DEFAULT_MO
The first 3 are defined in "cpu-param.h". The last one
in "cpu.h", with a bunch of definitions irrelevant for
TCG. By moving the TCG_GUEST_DEFAULT_MO definition to
"cpu-param.h", we can simplify various accel/tcg includes.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Acked-by: Nicholas Piggin <npiggin@gmail.com>
Message-Id: <20231211212003.21686-4-philmd@linaro.org>
Allows to set guest-phys-bits (cpuid leaf 80000008, eax[23:16])
via -cpu $model,guest-phys-bits=$nr.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-ID: <20240318155336.156197-3-kraxel@redhat.com>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The low bit of MMU indices for x86 TCG indicates whether the processor is
in 32-bit mode and therefore linear addresses have to be masked to 32 bits.
However, the index was computed incorrectly, leading to possible conflicts
in the TLB for any address above 4G.
Analyzed-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Fixes: b1661801c1 ("target/i386: Fix physical address truncation", 2024-02-28)
Cc: qemu-stable@nongnu.org
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2206
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The address translation logic in get_physical_address() will currently
truncate physical addresses to 32 bits unless long mode is enabled.
This is incorrect when using physical address extensions (PAE) outside
of long mode, with the result that a 32-bit operating system using PAE
to access memory above 4G will experience undefined behaviour.
The truncation code was originally introduced in commit 33dfdb5 ("x86:
only allow real mode to access 32bit without LMA"), where it applied
only to translations performed while paging is disabled (and so cannot
affect guests using PAE).
Commit 9828198 ("target/i386: Add MMU_PHYS_IDX and MMU_NESTED_IDX")
rearranged the code such that the truncation also applied to the use
of MMU_PHYS_IDX and MMU_NESTED_IDX. Commit 4a1e9d4 ("target/i386: Use
atomic operations for pte updates") brought this truncation into scope
for page table entry accesses, and is the first commit for which a
Windows 10 32-bit guest will reliably fail to boot if memory above 4G
is present.
The truncation code however is not completely redundant. Even though the
maximum address size for any executed instruction is 32 bits, helpers for
operations such as BOUND, FSAVE or XSAVE may ask get_physical_address()
to translate an address outside of the 32-bit range, if invoked with an
argument that is close to the 4G boundary. Likewise for processor
accesses, for example TSS or IDT accesses, when EFER.LMA==0.
So, move the address truncation in get_physical_address() so that it
applies to 32-bit MMU indexes, but not to MMU_PHYS_IDX and MMU_NESTED_IDX.
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2040
Fixes: 4a1e9d4d11 ("target/i386: Use atomic operations for pte updates", 2022-10-18)
Cc: qemu-stable@nongnu.org
Co-developed-by: Michael Brown <mcb30@ipxe.org>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Accesses from a 32-bit environment (32-bit code segment for instruction
accesses, EFER.LMA==0 for processor accesses) have to mask away the
upper 32 bits of the address. While a bit wasteful, the easiest way
to do so is to use separate MMU indexes. These days, QEMU anyway is
compiled with a fixed value for NB_MMU_MODES. Split MMU_USER_IDX,
MMU_KSMAP_IDX and MMU_KNOSMAP_IDX in two.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Remove knowledge of specific MMU indexes (other than MMU_NESTED_IDX and
MMU_PHYS_IDX) from mmu_translate(). This will make it possible to split
32-bit and 64-bit MMU indexes.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This commit adds support for x2APIC transitions when writing to
MSR_IA32_APICBASE register and finally adds CPUID_EXT_X2APIC to
TCG_EXT_FEATURES.
The set_base in APICCommonClass now returns an integer to indicate error in
execution. apic_set_base return -1 on invalid APIC state transition,
accelerator can use this to raise appropriate exception.
Signed-off-by: Bui Quang Minh <minhquangbui99@gmail.com>
Message-Id: <20240111154404.5333-4-minhquangbui99@gmail.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This commit extends the APIC ID to 32-bit long and remove the 255 max APIC
ID limit in userspace APIC. The array that manages local APICs is now
dynamically allocated based on the max APIC ID of created x86 machine.
Also, new x2APIC IPI destination determination scheme, self IPI and x2APIC
mode register access are supported.
Signed-off-by: Bui Quang Minh <minhquangbui99@gmail.com>
Message-Id: <20240111154404.5333-3-minhquangbui99@gmail.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This commit creates apic_register_read/write which are used by both
apic_mem_read/write for MMIO access and apic_msr_read/write for MSR access.
The apic_msr_read/write returns -1 on error, accelerator can use this to
raise the appropriate exception.
Signed-off-by: Bui Quang Minh <minhquangbui99@gmail.com>
Message-Id: <20240111154404.5333-2-minhquangbui99@gmail.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
For user-only mode, use MMU_USER_IDX.
For system mode, use CPUClass.mmu_index.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
ALU instructions can write to both memory and flags. If the CC_SRC*
and CC_DST locations have been written already when a memory access
causes a fault, the value in CC_SRC* and CC_DST might be interpreted
with the wrong CC_OP (the one that is in effect before the instruction.
Besides just using the wrong result for the flags, something like
subtracting -1 can have disastrous effects if the current CC_OP is
CC_OP_EFLAGS: this is because QEMU does not expect bits outside the ALU
flags to be set in CC_SRC, and env->eflags can end up set to all-ones.
In the case of the attached testcase, this sets IOPL to 3 and would
cause an assertion failure if SUB is moved to the new decoder.
This mechanism is not really needed for BMI instructions, which can
only write to a register, but put it to use anyway for cleanliness.
In the case of BZHI, the code has to be modified slightly to ensure
that decode->cc_src is written, otherwise the new assertions trigger.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
cpu_cc_compute_all() has an argument that is always equal to CC_OP for historical
reasons (dating back to commit a7812ae412, "TCG variable type checking.", 2008-11-17,
which added the argument to helper_cc_compute_all). It does not make sense for the
argument to have any other value, so remove it and clean up some lines that are not
too long anymore.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
In 32-bit mode, pc = eip + cs_base is also 32-bit, and must wrap.
Failure to do so results in incorrect memory exceptions to the guest.
Before 732d548732, this was implicitly done via truncation to
target_ulong but only in qemu-system-i386, not qemu-system-x86_64.
To fix this, we must add conditional zero-extensions.
Since we have to test for 32 vs 64-bit anyway, note that cs_base
is always zero in 64-bit mode.
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2022
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20231212172510.103305-1-richard.henderson@linaro.org>
In the nr_threads' comment, specify it represents the
number of threads in the "core" to avoid confusion.
Also add comment for nr_dies in CPUX86State.
Signed-off-by: Zhao Liu <zhao1.liu@intel.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Babu Moger <babu.moger@amd.com>
Tested-by: Yongwei Ma <yongwei.ma@intel.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Message-ID: <20231024090323.1859210-5-zhao1.liu@linux.intel.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
The OBJECT_DECLARE_CPU_TYPE() macro forward-declares each
ArchCPUClass type. These forward declarations are sufficient
for code in hw/ to use the QOM definitions. No need to expose
these structure definitions. Keep each local to their target/
by moving them to the corresponding "cpu.h" header.
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20231013140116.255-13-philmd@linaro.org>
Hegerogeneous code needs access to the FOO_CPU_TYPE_NAME()
macro to resolve target CPU types. Move the declaration
(along with the required FOO_CPU_TYPE_SUFFIX) to "cpu-qom.h".
"target/foo/cpu-qom.h" is supposed to be target agnostic
(include-able by any target). Add such mention in the
header.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Acked-by: LIU Zhiwei <zhiwei_liu@linux.alibaba.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20231013140116.255-7-philmd@linaro.org>
Enforce the style described by commit 067109a11c ("docs/devel:
mention the spacing requirement for QOM"):
The first declaration of a storage or class structure should
always be the parent and leave a visual space between that
declaration and the new code. It is also useful to separate
backing for properties (options driven by the user) and internal
state to make navigation easier.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Message-Id: <20231013140116.255-2-philmd@linaro.org>
Retain the separate structure to emphasize its importance.
Enforce CPUArchState always follows CPUState without padding.
Reviewed-by: Anton Johansson <anjo@rev.ng>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
On parts that enumerate IA32_VMX_BASIC MSR bit as 1, any exception vector
can be delivered with or without an error code if the other consistency
checks are satisfied.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Latest Intel platform GraniteRapids-D introduces AMX-COMPLEX, which adds
two instructions to perform matrix multiplication of two tiles containing
complex elements and accumulate the results into a packed single precision
tile.
AMX-COMPLEX is enumerated via CPUID.(EAX=7,ECX=1):EDX[bit 8]. Add the CPUID
definition for AMX-COMPLEX, AMX-COMPLEX will be enabled automatically when
using '-cpu host' and KVM advertises AMX-COMPLEX to userspace.
Signed-off-by: Tao Su <tao1.su@linux.intel.com>
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-ID: <20230830074324.84059-1-tao1.su@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Current QEMU can expose waitpkg to guests when it is available. However,
VMX_SECONDARY_EXEC_ENABLE_USER_WAIT_PAUSE is still not recognized and
masked by QEMU. This can lead to an unexpected situation when a L1
hypervisor wants to expose waitpkg to a L2 guest. The L1 hypervisor can
assume that VMX_SECONDARY_EXEC_ENABLE_USER_WAIT_PAUSE exists as waitpkg is
available. The L1 hypervisor then can accidentally expose waitpkg to the
L2 guest. This will cause invalid opcode exception in the L2 guest when
it executes waitpkg related instructions.
This patch adds VMX_SECONDARY_EXEC_ENABLE_USER_WAIT_PAUSE support, and
sets up dependency between the bit and CPUID_7_0_ECX_WAITPKG. QEMU should
not expose waitpkg feature if VMX_SECONDARY_EXEC_ENABLE_USER_WAIT_PAUSE is
not available to avoid unexpected invalid opcode exception in L2 guests.
Signed-off-by: Ake Koomsin <ake@igel.co.jp>
Message-ID: <20230807093339.32091-2-ake@igel.co.jp>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Currently, bit 13, 14, 15 and 24 of MSR_IA32_ARCH_CAPABILITIES are
disclosed for fixing security issues, so add those bit definitions.
Signed-off-by: Tao Su <tao1.su@linux.intel.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Message-ID: <20230706054949.66556-5-tao1.su@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
CPUID.(EAX=7,ECX=2):EDX[bit 5] enumerates MCDT_NO. Processors enumerate
this bit as 1 do not exhibit MXCSR Configuration Dependent Timing (MCDT)
behavior and do not need to be mitigated to avoid data-dependent behavior
for certain instructions.
Since MCDT_NO is in a new sub-leaf, add a new CPUID feature word
FEAT_7_2_EDX. Also update cpuid_level_func7 by FEAT_7_2_EDX.
Signed-off-by: Tao Su <tao1.su@linux.intel.com>
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-ID: <20230706054949.66556-3-tao1.su@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anton Johansson <anjo@rev.ng>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230621135633.1649-4-anjo@rev.ng>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Since commit 604664726f ("target/i386: Restrict cpu_exec_interrupt()
handler to sysemu"), do_cpu_sipi() isn't called anymore on user
emulation. Remove the now pointless stub.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230602224628.59546-2-philmd@linaro.org>
As reported by the Intel's doc:
"FB_CLEAR: The processor will overwrite fill buffer values as part of
MD_CLEAR operations with the VERW instruction.
On these processors, L1D_FLUSH does not overwrite fill buffer values."
If this cpu feature is present in host, allow QEMU to choose whether to
show it to the guest too.
One disadvantage of not exposing it is that the guest will report
a non existing vulnerability in
/sys/devices/system/cpu/vulnerabilities/mmio_stale_data
because the mitigation is present only when the cpu has
(FLUSH_L1D and MD_CLEAR) or FB_CLEAR
features enabled.
Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Message-Id: <20230201135759.555607-3-eesposit@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
As reported by Intel's doc:
"L1D_FLUSH: Writeback and invalidate the L1 data cache"
If this cpu feature is present in host, allow QEMU to choose whether to
show it to the guest too.
One disadvantage of not exposing it is that the guest will report
a non existing vulnerability in
/sys/devices/system/cpu/vulnerabilities/mmio_stale_data
because the mitigation is present only when the cpu has
(FLUSH_L1D and MD_CLEAR) or FB_CLEAR
features enabled.
Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Message-Id: <20230201135759.555607-2-eesposit@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Add the following featute bits.
vnmi: Virtual NMI (VNMI) allows the hypervisor to inject the NMI into the
guest without using Event Injection mechanism meaning not required to
track the guest NMI and intercepting the IRET.
The presence of this feature is indicated via the CPUID function
0x8000000A_EDX[25].
automatic-ibrs :
The AMD Zen4 core supports a new feature called Automatic IBRS.
It is a "set-and-forget" feature that means that, unlike e.g.,
s/w-toggled SPEC_CTRL.IBRS, h/w manages its IBRS mitigation
resources automatically across CPL transitions.
The presence of this feature is indicated via the CPUID function
0x80000021_EAX[8].
The documention for the features are available in the links below.
a. Processor Programming Reference (PPR) for AMD Family 19h Model 01h,
Revision B1 Processors
b. AMD64 Architecture Programmer’s Manual Volumes 1–5 Publication No. Revision
40332 4.05 Date October 2022
Signed-off-by: Santosh Shukla <santosh.shukla@amd.com>
Signed-off-by: Kim Phillips <kim.phillips@amd.com>
Signed-off-by: Babu Moger <babu.moger@amd.com>
Link: https://www.amd.com/system/files/TechDocs/55898_B1_pub_0.50.zip
Link: https://www.amd.com/system/files/TechDocs/40332_4.05.pdf
Message-Id: <20230504205313.225073-7-babu.moger@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Add the following feature bits.
no-nested-data-bp : Processor ignores nested data breakpoints.
lfence-always-serializing : LFENCE instruction is always serializing.
null-sel-cls-base : Null Selector Clears Base. When this bit is
set, a null segment load clears the segment base.
The documentation for the features are available in the links below.
a. Processor Programming Reference (PPR) for AMD Family 19h Model 01h,
Revision B1 Processors
b. AMD64 Architecture Programmer’s Manual Volumes 1–5 Publication No. Revision
40332 4.05 Date October 2022
Signed-off-by: Babu Moger <babu.moger@amd.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Link: https://www.amd.com/system/files/TechDocs/55898_B1_pub_0.50.zip
Link: https://www.amd.com/system/files/TechDocs/40332_4.05.pdf
Message-Id: <20230504205313.225073-5-babu.moger@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Add the following feature bits.
amd-psfd : Predictive Store Forwarding Disable:
PSF is a hardware-based micro-architectural optimization
designed to improve the performance of code execution by
predicting address dependencies between loads and stores.
While SSBD (Speculative Store Bypass Disable) disables both
PSF and speculative store bypass, PSFD only disables PSF.
PSFD may be desirable for the software which is concerned
with the speculative behavior of PSF but desires a smaller
performance impact than setting SSBD.
Depends on the following kernel commit:
b73a54321ad8 ("KVM: x86: Expose Predictive Store Forwarding Disable")
stibp-always-on :
Single Thread Indirect Branch Prediction mode has enhanced
performance and may be left always on.
The documentation for the features are available in the links below.
a. Processor Programming Reference (PPR) for AMD Family 19h Model 01h,
Revision B1 Processors
b. SECURITY ANALYSIS OF AMD PREDICTIVE STORE FORWARDING
Signed-off-by: Babu Moger <babu.moger@amd.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Link: https://www.amd.com/system/files/documents/security-analysis-predictive-store-forwarding.pdf
Link: https://www.amd.com/system/files/TechDocs/55898_B1_pub_0.50.zip
Message-Id: <20230504205313.225073-4-babu.moger@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Latest Intel platform Granite Rapids has introduced a new instruction -
PREFETCHIT0/1, which moves code to memory (cache) closer to the
processor depending on specific hints.
The bit definition:
CPUID.(EAX=7,ECX=1):EDX[bit 14]
Add CPUID definition for PREFETCHIT0/1.
Signed-off-by: Jiaxi Chen <jiaxi.chen@linux.intel.com>
Signed-off-by: Tao Su <tao1.su@linux.intel.com>
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-Id: <20230303065913.1246327-7-tao1.su@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
AVX-NE-CONVERT is a new set of instructions which can convert low
precision floating point like BF16/FP16 to high precision floating point
FP32, as well as convert FP32 elements to BF16. This instruction allows
the platform to have improved AI capabilities and better compatibility.
The bit definition:
CPUID.(EAX=7,ECX=1):EDX[bit 5]
Add CPUID definition for AVX-NE-CONVERT.
Signed-off-by: Jiaxi Chen <jiaxi.chen@linux.intel.com>
Signed-off-by: Tao Su <tao1.su@linux.intel.com>
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-Id: <20230303065913.1246327-6-tao1.su@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
AVX-VNNI-INT8 is a new set of instructions in the latest Intel platform
Sierra Forest, aims for the platform to have superior AI capabilities.
This instruction multiplies the individual bytes of two unsigned or
unsigned source operands, then adds and accumulates the results into the
destination dword element size operand.
The bit definition:
CPUID.(EAX=7,ECX=1):EDX[bit 4]
AVX-VNNI-INT8 is on a new feature bits leaf. Add a CPUID feature word
FEAT_7_1_EDX for this leaf.
Add CPUID definition for AVX-VNNI-INT8.
Signed-off-by: Jiaxi Chen <jiaxi.chen@linux.intel.com>
Signed-off-by: Tao Su <tao1.su@linux.intel.com>
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-Id: <20230303065913.1246327-5-tao1.su@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
AVX-IFMA is a new instruction in the latest Intel platform Sierra
Forest. This instruction packed multiplies unsigned 52-bit integers and
adds the low/high 52-bit products to Qword Accumulators.
The bit definition:
CPUID.(EAX=7,ECX=1):EAX[bit 23]
Add CPUID definition for AVX-IFMA.
Signed-off-by: Jiaxi Chen <jiaxi.chen@linux.intel.com>
Signed-off-by: Tao Su <tao1.su@linux.intel.com>
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-Id: <20230303065913.1246327-4-tao1.su@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Latest Intel platform Granite Rapids has introduced a new instruction -
AMX-FP16, which performs dot-products of two FP16 tiles and accumulates
the results into a packed single precision tile. AMX-FP16 adds FP16
capability and allows a FP16 GPU trained model to run faster without
loss of accuracy or added SW overhead.
The bit definition:
CPUID.(EAX=7,ECX=1):EAX[bit 21]
Add CPUID definition for AMX-FP16.
Signed-off-by: Jiaxi Chen <jiaxi.chen@linux.intel.com>
Signed-off-by: Tao Su <tao1.su@linux.intel.com>
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-Id: <20230303065913.1246327-3-tao1.su@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
CMPccXADD is a new set of instructions in the latest Intel platform
Sierra Forest. This new instruction set includes a semaphore operation
that can compare and add the operands if condition is met, which can
improve database performance.
The bit definition:
CPUID.(EAX=7,ECX=1):EAX[bit 7]
Add CPUID definition for CMPCCXADD.
Signed-off-by: Jiaxi Chen <jiaxi.chen@linux.intel.com>
Signed-off-by: Tao Su <tao1.su@linux.intel.com>
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-Id: <20230303065913.1246327-2-tao1.su@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Introduce support for one shot and periodic mode of Xen PV timers,
whereby timer interrupts come through a special virq event channel
with deadlines being set through:
1) set_timer_op hypercall (only oneshot)
2) vcpu_op hypercall for {set,stop}_{singleshot,periodic}_timer
hypercalls
Signed-off-by: Joao Martins <joao.m.martins@oracle.com>
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Paul Durrant <paul@xen.org>
The GSI callback (and later PCI_INTX) is a level triggered interrupt. It
is asserted when an event channel is delivered to vCPU0, and is supposed
to be cleared when the vcpu_info->evtchn_upcall_pending field for vCPU0
is cleared again.
Thankfully, Xen does *not* assert the GSI if the guest sets its own
evtchn_upcall_pending field; we only need to assert the GSI when we
have delivered an event for ourselves. So that's the easy part, kind of.
There's a slight complexity in that we need to hold the BQL before we
can call qemu_set_irq(), and we definitely can't do that while holding
our own port_lock (because we'll need to take that from the qemu-side
functions that the PV backend drivers will call). So if we end up
wanting to set the IRQ in a context where we *don't* already hold the
BQL, defer to a BH.
However, we *do* need to poll for the evtchn_upcall_pending flag being
cleared. In an ideal world we would poll that when the EOI happens on
the PIC/IOAPIC. That's how it works in the kernel with the VFIO eventfd
pairs — one is used to trigger the interrupt, and the other works in the
other direction to 'resample' on EOI, and trigger the first eventfd
again if the line is still active.
However, QEMU doesn't seem to do that. Even VFIO level interrupts seem
to be supported by temporarily unmapping the device's BARs from the
guest when an interrupt happens, then trapping *all* MMIO to the device
and sending the 'resample' event on *every* MMIO access until the IRQ
is cleared! Maybe in future we'll plumb the 'resample' concept through
QEMU's irq framework but for now we'll do what Xen itself does: just
check the flag on every vmexit if the upcall GSI is known to be
asserted.
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Paul Durrant <paul@xen.org>
Add the array of virq ports to each vCPU so that we can deliver timers,
debug ports, etc. Global virqs are allocated against vCPU 0 initially,
but can be migrated to other vCPUs (when we implement that).
The kernel needs to know about VIRQ_TIMER in order to accelerate timers,
so tell it via KVM_XEN_VCPU_ATTR_TYPE_TIMER. Also save/restore the value
of the singleshot timer across migration, as the kernel will handle the
hypercalls automatically now.
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Paul Durrant <paul@xen.org>
The kvm_xen_inject_vcpu_callback_vector() function will either deliver
the per-vCPU local APIC vector (as an MSI), or just kick the vCPU out
of the kernel to trigger KVM's automatic delivery of the global vector.
Support for asserting the GSI/PCI_INTX callbacks will come later.
Also add kvm_xen_get_vcpu_info_hva() which returns the vcpu_info of
a given vCPU.
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Paul Durrant <paul@xen.org>
The HVMOP_set_evtchn_upcall_vector hypercall sets the per-vCPU upcall
vector, to be delivered to the local APIC just like an MSI (with an EOI).
This takes precedence over the system-wide delivery method set by the
HVMOP_set_param hypercall with HVM_PARAM_CALLBACK_IRQ. It's used by
Windows and Xen (PV shim) guests but normally not by Linux.
Signed-off-by: Ankur Arora <ankur.a.arora@oracle.com>
Signed-off-by: Joao Martins <joao.m.martins@oracle.com>
[dwmw2: Rework for upstream kernel changes and split from HVMOP_set_param]
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Paul Durrant <paul@xen.org>
Allow guest to setup the vcpu runstates which is used as
steal clock.
Signed-off-by: Joao Martins <joao.m.martins@oracle.com>
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Paul Durrant <paul@xen.org>
