Qemu-3.0 supports Hyper-V-style PV TLB flush, Windows guests can benefit
from this feature as KVM knows which vCPUs are not currently scheduled (and
thus don't require any immediate action).
Signed-off-by: NVitaly Kuznetsov <vkuznets@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Qemu-3.0 supports so-called 'Reenlightenment' notifications and this (in
conjunction with 'hv-frequencies') can be used make Hyper-V on KVM pass
stable TSC page clocksource to L2 guests.
Signed-off-by: NVitaly Kuznetsov <vkuznets@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Qemu-2.12 gained 'hv-frequencies' cpu flag to enable Hyper-V frequency
MSRs. These MSRs are required (but not sufficient) to make Hyper-V on
KVM pass stable TSC page clocksource to L2 guests.
Signed-off-by: NVitaly Kuznetsov <vkuznets@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

When computing a baseline CPU for a specific hypervisor we have to make
sure to include only CPU features supported by the hypervisor. Otherwise
the computed CPU could not be used for starting a new domain.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>
Reviewed-by: NCollin Walling <walling@linux.ibm.com>

Modern host CPU models from domain capabilities XMLs are reported as
guest CPU definitions with feature policies. This patch updates
virCPUx86Baseline to properly handle such CPU models.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

When specifying a new CPU model in cpu_map.xml as an extension to an
existing model, we forgot to copy the signature (family + model) from
the original CPU model.

We don't use this way of specifying CPU models, but it's still supported
and it becomes useful when someone wants to quickly hack up a CPU model
for testing or when creating additional variants of existing models to
help with fixing some spectral issues.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NPavel Hrdina <phrdina@redhat.com>

When translating CPUID data into CPU model + features, the code
sometimes uses an unexpected CPU model. There may be several reasons for
this, starting with wrong expectations and ending with an actual bug in
our code. These debug messages will help determining the reason.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NPavel Hrdina <phrdina@redhat.com>

A microcode update can cause the CPUID bits to change; an example
from the past was the update that disabled TSX on several Haswell and
Broadwell machines.

In order to track the x86 microcode version in the QEMU capabilities,
we have to fetch it and store it in the host CPU.  This also makes the
version visible in "virsh capabilities", which is a nice side effect.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

The function will be used to initialize internal data of the x86 CPU
driver (including the CPU map).
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

Right-aligning backslashes when defining macros or using complex
commands in Makefiles looks cute, but as soon as any changes is
required to the code you end up with either distractingly broken
alignment or unnecessarily big diffs where most of the changes
are just pushing all backslashes a few characters to one side.

Generated using

  $ git grep -El '[[:blank:]][[:blank:]]\\$' | \
    grep -E '*\.([chx]|am|mk)$$' | \
    while read f; do \
      sed -Ei 's/[[:blank:]]*[[:blank:]]\\$/ \\/g' "$f"; \
    done
Signed-off-by: NAndrea Bolognani <abologna@redhat.com>

Even though only family and model are used for matching CPUID data with
CPU models from cpu_map.xml, stepping is used by x86DataFilterTSX which
is supposed to disable TSX on CPU models with broken TSX support. Thus
we need to start parsing stepping from QEMU to make sure we don't
disable TSX on CPUs which provide working TSX implementation. See the
following patch for a real world example of such CPU.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

When decoding CPUID data to virCPUDef we need to be careful about using
a CPU model which cannot be directly used on the current host. Normally,
libvirt would notice the features which prevent the model from being
usable and it would disable them in the computed virCPUDef, but this
won't work in case the definition of the CPU model in QEMU contains more
features than what we have in cpu_map.xml. We need to count with the
usability blockers we got from QEMU and explicitly disable all of them
to make the computed virCPUDef usable.

https://bugzilla.redhat.com/show_bug.cgi?id=1464832Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

The "preferred" parameter is not used by any caller of cpuDecode
anymore. It's only used internally in cpu_x86 to implement cpuBaseline.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

All APIs which expect a list of CPU models supported by hypervisors were
switched from char **models and int models to just accept a pointer to
virDomainCapsCPUModels object stored in domain capabilities. This avoids
the need to transform virDomainCapsCPUModelsPtr into a NULL-terminated
list of model names and also allows the various cpu driver APIs to
access additional details (such as its usability) about each CPU model.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

The function checks whether all CPU features used in a CPU definition
are specified in cpu_map.xml.

https://bugzilla.redhat.com/show_bug.cgi?id=1460086Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

CPU features unknown to a hypervisor will not be present in dataDisabled
even though the features won't naturally be enabled because.
Thus any features we asked for which are not in dataEnabled should be
considered disabled.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

Because of the changes done in the previous commit, @host is already a
migratable CPU and there's no need to do any additional filtering.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

This reverts commit 5f96b3fe which was
pushed accidentally.

Because of the changes done in the previous commit, @host is already a
migratable CPU and there's no need to do any additional filtering.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

This new internal API makes a copy of virCPUDef while removing all
features which would block migration. It uses cpu_map.xml as a database
of such features, which should only be used as a fallback when we cannot
get the data from a hypervisor. The main goal of this API is to decouple
this filtering from virCPUUpdate so that the hypervisor driver can
filter the features according to the hypervisor.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

The public API flags are handled by the cpuBaselineXML wrapper. The
internal cpuBaseline API only needs to know whether it is supposed to
drop non-migratable features.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

cpuBaseline is responsible for computing a baseline CPU while feature
expansion is done by virCPUExpandFeatures. The cpuBaselineXML wrapper
(used by hypervisor drivers to implement virConnectBaselineCPU API)
calls cpuBaseline followed by virCPUExpandFeatures if requested by
VIR_CONNECT_BASELINE_CPU_EXPAND_FEATURES flag.

The features in the three changed test files had to be sorted using
"sort -k 3" because virCPUExpandFeatures returns a sorted list of
features.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

Having to use cpuBaseline with VIR_CONNECT_BASELINE_CPU_EXPAND_FEATURES
flag to expand CPU features is strange. Not to mention that cpuBaseline
can only expand host CPU definitions (i.e., it completely ignores
feature policies). The new virCPUExpandFeatures API is designed to work
with both host and guest CPU definitions.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

When guest CPU definition uses VIR_CPU_CHECK_FULL checks, we need to
make sure QEMU does not add or remove any features.

https://bugzilla.redhat.com/show_bug.cgi?id=822148
https://bugzilla.redhat.com/show_bug.cgi?id=824989Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

When starting a domain with custom guest CPU specification QEMU may add
or remove some CPU features. There are several reasons for this, e.g.,
QEMU/KVM does not support some requested features or the definition of
the requested CPU model in libvirt's cpu_map.xml differs from the one
QEMU is using. We can't really avoid this because CPU models are allowed
to change with machine types and libvirt doesn't know (and probably
doesn't even want to know) about such changes.

Thus when we want to make sure guest ABI doesn't change when a domain
gets migrated to another host, we need to update our live CPU definition
according to the CPU QEMU created. Once updated, we will change CPU
checking to VIR_CPU_CHECK_FULL to make sure the virtual CPU created
after migration exactly matches the one on the source.

https://bugzilla.redhat.com/show_bug.cgi?id=822148
https://bugzilla.redhat.com/show_bug.cgi?id=824989Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

When creating host CPU definition usable with a given emulator, the CPU
should not be defined using an unsupported CPU model. The new @models
and @nmodels parameters can be used to limit CPU models which can be
used in the result.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

cpuNodeData has always been followed by cpuDecode as no hypervisor
driver is really interested in raw CPUID data for a host CPU. Let's
create a new CPU driver API which returns virCPUDefPtr directly.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

All Intel Haswell processors (except Xeon E7 v3 with stepping >= 4) have
TSX disabled by microcode update. As not all CPUs are guaranteed to be
patched with microcode updates we need to explicitly disable TSX on
affected CPUs to avoid its accidental usage.

https://bugzilla.redhat.com/show_bug.cgi?id=1406791Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

The API is useful for creating virCPUData in a hypervisor driver from
data we got by querying the hypervisor.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

The API is useful for creating virCPUData in a hypervisor driver from
data we got by querying the hypervisor.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

The API is useful for creating virCPUData in a hypervisor driver from
data we got by querying the hypervisor.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

The CPU driver provides APIs to create and free virCPUDataPtr. Thus all
APIs exported from the driver should work with that rather than
requiring the caller to pass a pointer to an internal part of the
structure.

In other words

    virCPUx86DataAddCPUID(cpudata, &cpuid)

is much better than the original

    virCPUx86DataAddCPUID(&cpudata->data.x86, &cpuid)
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

The new API is called virCPUDataFree. Individual CPU drivers are no
longer required to implement their own freeing function unless they need
to free architecture specific data from virCPUData.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

Strings associated with virDomainHyperv values in domain_conf.c are used to
construct HyperV CPU features names to be compared with names defined in
cpu_x86_data.h and the names for HyperV "spinlocks" feature don't match.
This leads to a misleading warning:
"host doesn't support hyperv 'spinlocks' feature" even when it's supported.
Let's fix it and rename along with it VIR_CPU_x86_KVM_HV_SPINLOCK to
VIR_CPU_x86_KVM_HV_SPINLOCKS.
Signed-off-by: NMaxim Nestratov <mnestratov@virtuozzo.com>

virCPUDefStealModel is called with keepVendor == true which means the
cpu structure will keep its original vendor/vendor_id values. Thus it
makes no sense to copy them to the translated definition as they won't
be used there anyway. Except that the translated->vendor pointer might
get lost in x86Decode.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>