VCPU TSC is not cleared by a warm reset (*), which leaves some types of Linux
guests (non-pvops guests and those with the kernel parameter no-kvmclock set)
vulnerable to the overflow in cyc2ns_offset fixed by upstream commit
9993bc635d01a6ee7f6b833b4ee65ce7c06350b1 ("sched/x86: Fix overflow in
cyc2ns_offset").
To put it in a nutshell, if such a Linux guest without the patch above applied
has been up more than 208 days and attempts a warm reset chances are that
the newly booted kernel will panic or hang.
(*) Intel Xeon E5 processors show the same broken behavior due to
the errata "TSC is Not Affected by Warm Reset" (Intel® Xeon®
Processor E5 Family Specification Update - August 2013): "The
TSC (Time Stamp Counter MSR 10H) should be cleared on
reset. Due to this erratum the TSC is not affected by warm
reset."
Cc: Will Auld <will.auld@intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Fernando Luis Vazquez Cao <fernando@oss.ntt.co.jp>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Fernando Luis Vázquez Cao <fernando_b1@lab.ntt.co.jp>
Add some MPX related definiation, and hardcode sizes and offsets
of xsave features 3 and 4. It also add corresponding part to
kvm_get/put_xsave, and vmstate.
Signed-off-by: Liu Jinsong <jinsong.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Fix cpuid leaf 0x0d which incorrectly parsed eax and ebx.
However, before this patch the CPUID worked fine -- the .offset
field contained the size _and_ was stored in the register that
is supposed to hold the size (eax), and likewise the .size field
contained the offset _and_ was stored in the register trhat is
supposed to hold the offset (ebx).
Signed-off-by: Liu Jinsong <jinsong.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Commit 787aaf5 (target-i386: forward CPUID cache leaves when -cpu host is
used, 2013-09-02) brings bits 31..26 of CPUID leaf 04h out of sync with
the APIC IDs that QEMU reserves for each package. This number must come
from "-smp" options rather than from the host CPUID.
It also turns out that this unsyncing makes Windows Server 2012R2 fail
to boot.
Tested-by: Peter Lieven <pl@kamp.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Message-id: 1384879786-6721-1-git-send-email-pbonzini@redhat.com
Signed-off-by: Anthony Liguori <aliguori@amazon.com>
# By Paolo Bonzini (2) and Jan Kiszka (1)
# Via Gleb Natapov
* qemu-kvm/uq/master:
kvmvapic: Prevent reading beyond the end of guest RAM
x86: cpuid: reconstruct leaf 0Dh data
x86: fix migration from pre-version 12
Message-id: 1382108641-4862-1-git-send-email-pbonzini@redhat.com
Signed-off-by: Anthony Liguori <aliguori@amazon.com>
Since this is only read in cpu_copy() and linux-user has a global
cpu_model, drop the field from generic code.
Signed-off-by: Andreas Färber <afaerber@suse.de>
The data in leaf 0Dh depends on information from other feature bits.
Instead of passing it blindly from the host, compute it based on
whether these feature bits are enabled.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
On KVM, the KVM_SET_XSAVE would be executed with a 0 xstate_bv,
and not restore anything.
Since FP and SSE data are always valid, set them in xstate_bv at reset
time. In fact, that value is the same that KVM_GET_XSAVE returns on
pre-XSAVE hosts.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
There's no Intel CPU with family=6,model=2, and Linux and Windows guests
disable SEP when seeing that combination due to Pentium Pro erratum #82.
In addition to just having SEP ignored by guests, Skype (and maybe other
applications) runs sysenter directly without passing through ntdll on
Windows, and crashes because Windows ignored the SEP CPUID bit.
So, having model > 2 is a better default on qemu64 and qemu32 for two
reasons: making SEP really available for guests, and avoiding crashing
applications that work on bare metal.
model=3 would fix the problem, but it causes CPU enumeration problems
for Windows guests[1]. So let's set model=6, that matches "Athlon
(PM core)" on AMD and "P2 with on-die L2 cache" on Intel and it allows
Windows to use all CPUs as well as fixing sysenter.
[1] https://bugzilla.redhat.com/show_bug.cgi?id=508623
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
I don't know yet if want this feature on by default, so for now I'm
just adding support for "-cpu ...,+kvm_pv_unhalt".
Signed-off-by: Andrew Jones <drjones@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Some users running cpu intensive tasks checking the cache CPUID leaves at
startup and making decisions based on the result reported that the guest was
not reflecting the host CPUID leaves when -cpu host is used.
This patch fix this.
Signed-off-by: Benoît Canet <benoit@irqsave.net>
[Rename new field to cache_info_passthrough - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This is an attempt to make the CPUID cache topology code clearer, by
replacing the magic numbers in the code with #defines, and moving all
the cache information to the same place in the file.
I took care of comparing the assembly output of compiling
target-i386/cpu.c before and after applying this change, to make sure
not a single bit was changed on cpu_x86_cpuid() before and after
applying this patch (unfortunately I had to manually check existing
differences, because of __LINE__ expansions on
object_class_dynamic_cast_assert() calls).
This even keeps the code bug-compatible with the previous version: today
the cache information returned on AMD cache information leaves (CPUID
0x80000005 & 0x80000006) do not match the information returned on CPUID
leaves 2 and 4. The L2 cache information on CPUID leaf 2 also doesn't
match the information on CPUID leaf 2. The new constants should make it
easier to eventually fix those inconsistencies. All inconsistencies I
have found are documented in code comments.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: liguang <lig.fnst@cn.fujitsu.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Convert stderr messages calling error_get_pretty()
to error_report().
Timestamp is prepended by -msg timstamp option with it.
Per Markus's comment below, A conversion from fprintf() to
error_report() is always an improvement, regardless of
error_get_pretty().
http://marc.info/?l=qemu-devel&m=137513283408601&w=2
But, it is not reasonable to convert them at one time
because fprintf() is used everwhere in qemu.
So, it should be done step by step with avoiding regression.
Signed-off-by: Seiji Aguchi <seiji.aguchi@hds.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
- since hyperv_* helper functions are used only in target-i386/kvm.c
move them there as static helpers
Requested-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Error **errp argument is not for emitting warnings, it means an error
has occurred and the caller should not make any assumptions about the
state of other return values (unless otherwise documented).
Therefore cpu_x86_create() must unref the new X86CPU itself, and
pc_new_cpu() must check for an Error rather than NULL return value.
While at it, clean up a superfluous NULL check.
Reported-by: Jan Kiszka <jan.kiszka@siemens.com>
Cc: qemu-stable@nongnu.org
Cc: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Bug description: QEMU currently gets all bits from GET_SUPPORTED_CPUID
for CPUID leaf 0xA and passes them directly to the guest. This makes
the guest ABI depend on host kernel and host CPU capabilities, and
breaks live migration if we migrate between hosts with different
capabilities (e.g., different number of PMU counters).
Add a "pmu" property to X86CPU, and set it to true only on "-cpu host",
or on pc-*-1.5 and older machine-types.
For now, setting pmu=on will enable the current passthrough mode that
doesn't have any ABI stability guarantees, but in the future we may
implement a mode where the PMU CPUID bits are stable and configurable.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
This will help us change the initialization code to not require carrying
some intermediate values in a x86_def_t struct (and eventually kill the
x86_def_t struct entirely).
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Commit c643bed99 moved qemu_init_vcpu() calls to common CPUState code.
This causes x86 cpu-add to fail with "KVM: setting VAPIC address failed".
The reason for the failure is that CPUClass::kvm_fd is not yet
initialized in the following call graph:
->x86_cpu_realizefn
->x86_cpu_apic_realize
->qdev_init
->device_set_realized
->device_reset (hotplugged == 1)
->apic_reset_common
->vapic_base_update
->kvm_apic_vapic_base_update
This causes attempted KVM vCPU ioctls to fail.
By contrast, in the non-hotplug case the APIC is reset much later, when
the vCPU is already initialized.
As a quick and safe solution, move the qemu_init_vcpu() call back into
the targets' realize functions.
Reported-by: Chen Fan <chen.fan.fnst@cn.fujitsu.com>
Acked-by: Igor Mammedov <imammedo@redhat.com> (for i386)
Tested-by: Jia Liu <proljc@gmail.com> (for openrisc)
Signed-off-by: Andreas Färber <afaerber@suse.de>
Completes migration of target-specific code to new target-*/gdbstub.c.
Acked-by: Michael Walle <michael@walle.cc> (for lm32)
Acked-by: Max Filippov <jcmvbkbc@gmail.com> (for xtensa)
Signed-off-by: Andreas Färber <afaerber@suse.de>
CPUState::gdb_num_regs replaces num_g_regs.
CPUClass::gdb_num_core_regs replaces NUM_CORE_REGS.
Allows building gdb_register_coprocessor() for xtensa, too.
As a side effect this should fix coprocessor register numbering for SMP.
Acked-by: Michael Walle <michael@walle.cc> (for lm32)
Acked-by: Max Filippov <jcmvbkbc@gmail.com> (for xtensa)
Signed-off-by: Andreas Färber <afaerber@suse.de>
Change breakpoint_invalidate() argument to CPUState alongside.
Since all targets now assign a softmmu-only field, we can drop helpers
cpu_class_set_{do_unassigned_access,vmsd}() and device_class_set_vmsd().
Prepares for changing cpu_memory_rw_debug() argument to CPUState.
Acked-by: Max Filippov <jcmvbkbc@gmail.com> (for xtensa)
Signed-off-by: Andreas Färber <afaerber@suse.de>
Where no extra implementation is needed, fall back to CPUClass::set_pc().
Acked-by: Michael Walle <michael@walle.cc> (for lm32)
Signed-off-by: Andreas Färber <afaerber@suse.de>
This moves setting the Program Counter from gdbstub into target code.
Use vaddr type as upper-bound replacement for target_ulong.
Signed-off-by: Andreas Färber <afaerber@suse.de>
x86 was using additional CPU_DUMP_* flags, so make that configurable in
CPUClass::reset_dump_flags.
This adds reset logging for alpha, unicore32 and xtensa.
Acked-by: Michael Walle <michael@walle.cc> (for lm32)
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Since commit 878096eeb2 (cpu: Turn
cpu_dump_{state,statistics}() into CPUState hooks) CPUArchState is no
longer needed.
Add documentation and make the functions available through qemu/log.h
outside NEED_CPU_H to allow use in qom/cpu.c. Moving them to qom/cpu.h
was not yet possible due to convoluted include paths, so that some
devices grow an implicit and unneeded dependency on qom/cpu.h for now.
Acked-by: Michael Walle <michael@walle.cc> (for lm32)
Reviewed-by: Richard Henderson <rth@twiddle.net>
[AF: Simplified mb_cpu_do_interrupt() and do_interrupt_all() changes]
Signed-off-by: Andreas Färber <afaerber@suse.de>
This allows to move the call into CPUState's realizefn.
Therefore move the stub into libqemustub.a.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Make cpustats monitor command available unconditionally.
Prepares for changing kvm_handle_internal_error() and kvm_cpu_exec()
arguments to CPUState.
Signed-off-by: Andreas Färber <afaerber@suse.de>
Relocate assignment of x86 get_arch_id to have all hooks in one place.
Reviewed-by: Jens Freimann <jfrei@linux.vnet.ibm.com>
Reviewed-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Spotted by Coverity,
x86_reg_info_32[] is CPU_NB_REGS32 elements long, so accessing
x86_reg_info_32[CPU_NB_REGS32] will be one element off array.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: liguang <lig.fnst@cn.fujitsu.com>
Reviewed by: Jesse Larrew <jlarrew@linux.vnet.ibm.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
The CPUID level value on Conroe, Penryn, and Nehalem are too low. This
causes at least one known problem: the -smp "threads" option doesn't
work as expect if level is < 4, because thread count information is
provided to the guest on CPUID[EAX=4,ECX=2].EAX
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
The CPUID model values on Conroe, Penryn, and Nehalem are too
conservative and don't reflect the values found on real Conroe, Penryn,
and Nehalem CPUs.
This causes at least one known problems: Windows XP disables sysenter
when (family == 6 && model <= 2), but Skype tries to use the sysenter
instruction anyway because it is reported as available on CPUID, making
it crash.
This patch sets appropriate model values that correspond to real Conroe,
Penryn, and Nehalem CPUs.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
The Atom core (cpu name "n270" in QEMU speak) supports MOVBE. This is
needed when booting 3.8 and later linux kernels built with the MATOM
target because we require MOVBE in order to boot properly now.
Signed-off-by: Borislav Petkov <bp@suse.de>
[ehabkost: added compat code to disable MOVBE on pc-*-1.4 and older]
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Introduce x86_cpu_compat_set_features(), that can be used to set/unset
feature bits on specific CPU models for machine-type compatibility.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
This changes the model number of 486 to 8 (DX4) which matches the
feature set presented, and actually has the CPUID instruction.
This adds a compatibility property, to keep model=0 on pc-*-1.4 and older.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
[AF: Add compat_props entry]
Tested-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
After initializing the object from its x86_def_t and before setting any
additional -cpu arguments, set any global properties for the designated
subclass <name>-{i386,x86_64}-cpu.
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
This field will contain the feature bits that were filtered out because
of missing host support.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
This property will be useful for libvirt, as libvirt already has logic
based on low-level feature bits (not feature names), so it will be
really easy to convert the current libvirt logic to something using the
"feature-words" property.
The property will have two main use cases:
- Checking host capabilities, by checking the features of the "host"
CPU model
- Checking which features are enabled on each CPU model
Example output:
$ ./QMP/qmp --path=/tmp/m \
qom-get --path=/machine/icc-bridge/icc/child[0] \
--property=feature-words
item[0].cpuid-register: EDX
item[0].cpuid-input-eax: 2147483658
item[0].features: 0
item[1].cpuid-register: EAX
item[1].cpuid-input-eax: 1073741825
item[1].features: 0
item[2].cpuid-register: EDX
item[2].cpuid-input-eax: 3221225473
item[2].features: 0
item[3].cpuid-register: ECX
item[3].cpuid-input-eax: 2147483649
item[3].features: 101
item[4].cpuid-register: EDX
item[4].cpuid-input-eax: 2147483649
item[4].features: 563346425
item[5].cpuid-register: EBX
item[5].cpuid-input-eax: 7
item[5].features: 0
item[5].cpuid-input-ecx: 0
item[6].cpuid-register: ECX
item[6].cpuid-input-eax: 1
item[6].features: 2155880449
item[7].cpuid-register: EDX
item[7].cpuid-input-eax: 1
item[7].features: 126614521
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Instead of open-coding the filtering code for each feature word, change
the existing code to use the feature_word_info array, that has exactly
the same CPUID eax/ecx/register values for each feature word.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
FEAT_7_0_EBX uses ECX as input, so we have to take that into account
when reporting feature word values.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
This replaces the feature-bit fields on both X86CPU and x86_def_t
structs with an array.
With this, we will be able to simplify code that simply does the same
operation on all feature words (e.g. kvm_check_features_against_host(),
filter_features_for_kvm(), add_flagname_to_bitmaps(), CPU feature-bit
property lookup/registration, and the proposed "feature-words" property)
The following field replacements were made on X86CPU and x86_def_t:
(cpuid_)features -> features[FEAT_1_EDX]
(cpuid_)ext_features -> features[FEAT_1_ECX]
(cpuid_)ext2_features -> features[FEAT_8000_0001_EDX]
(cpuid_)ext3_features -> features[FEAT_8000_0001_ECX]
(cpuid_)ext4_features -> features[FEAT_C000_0001_EDX]
(cpuid_)kvm_features -> features[FEAT_KVM]
(cpuid_)svm_features -> features[FEAT_SVM]
(cpuid_)7_0_ebx_features -> features[FEAT_7_0_EBX]
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Break lines on kvm_check_features_against_host(), kvm_cpu_fill_host(),
and builtin_x86_defs, so they don't get too long once the *_features
fields are replaced by an array.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
It allows APIC to be hotplugged.
* map APIC's mmio at board level if it is present
* do not register mmio region for each APIC, since
only one is used/mapped
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
X86CPU should have parent bus so it could provide bus for child APIC.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
The property is used from board level to set APIC ID for CPUs it
creates. Do so in a new pc_new_cpu() helper, to be reused for hot-plug.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
This helper replaces '_' with '-' in a uniform way.
As a side effect, even custom mappings must use '-' now.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
[AF: Split off; operate on NUL-terminated string rather than '=' delimiter]
Signed-off-by: Andreas Färber <afaerber@suse.de>
get_arch_id() adds possibility for generic code to get a guest-visible
CPU ID without accessing CPUArchState.
If derived classes don't override it, it will return cpu_index.
Override it on target-i386 in X86CPU to return the APIC ID.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: liguang <lig.fnst@cn.fujitsu.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Move CPU creation and features parsing into a separate cpu_x86_create()
function, so that board would be able to set board-specific CPU
properties before CPU is realized.
Keep cpu_x86_init() for compatibility with the code that uses cpu_init()
and doesn't need to modify CPU properties.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
* Add braces to 'if' statements;
* Remove last TAB character from the source.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
[AF: Changed whitespace]
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
When APIC is hotplugged during CPU hotplug, device_set_realized()
calls device_reset() on it. And if QEMU runs in KVM mode, following
call chain will fail:
apic_reset_common()
-> kvm_apic_vapic_base_update()
-> kvm_vcpu_ioctl(cpu->kvm_fd,...)
due to cpu->kvm_fd not being initialized yet.
cpu->kvm_fd is initialized during qemu_init_vcpu() but x86_cpu_apic_init()
can't be moved after it because kvm_init_vcpu() -> kvm_arch_reset_vcpu()
relies on APIC to determine if CPU is BSP for setting initial env->mp_state.
So split APIC device creation from its initialization and realize APIC
after CPU is created, when it's safe to call APIC's reset method.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: liguang <lig.fnst@cn.fujitsu.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
We were missing a bunch of feature lists. Fix this by simply dumping
the meta list feature_word_info.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
kvm_enabled() cannot be true at this point because accelerators are
initialized much later during init. Also, hiding this makes it very hard
to discover for users. Simply dump unconditionally if CONFIG_KVM is set.
Add explanation for "host" CPU type.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
The PCLMULQDQ instruction has been introduced on the Westmere CPU.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Many of these should be cleaned up with proper qdev-/QOM-ification.
Right now there are many catch-all headers in include/hw/ARCH depending
on cpu.h, and this makes it necessary to compile these files per-target.
However, fixing this does not belong in these patches.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
A common dependency of the constant's current users:
- hw/apic_common.c
- hw/i386/kvmvapic.c
- target-i386/cpu.c
is "target-i386/cpu.h".
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
Message-id: 1363821803-3380-9-git-send-email-lersek@redhat.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
commit 5ec01c2e96 broke "-cpu ..,enforce",
as it has moved kvm_check_features_against_host() after the
filter_features_for_kvm() call. filter_features_for_kvm() removes all
features not supported by the host, so this effectively made
kvm_check_features_against_host() impossible to fail.
This patch changes the call so we check for host feature support before
filtering the feature bits.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Message-id: 1364935692-24004-1-git-send-email-ehabkost@redhat.com
Cc: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This removes a global per-target function and thus takes us one step
closer to compiling multiple targets into one executable.
It will also allow to override the interrupt handling for certain CPU
families.
Signed-off-by: Andreas Färber <afaerber@suse.de>
Both fields are used in VMState, thus need to be moved together.
Explicitly zero them on reset since they were located before
breakpoints.
Pass PowerPCCPU to kvmppc_handle_halt().
Signed-off-by: Andreas Färber <afaerber@suse.de>
Expose vmstate_cpu as vmstate_x86_cpu and hook it up to CPUClass::vmsd.
Adapt opaques and VMState fields to X86CPU. Drop cpu_{save,load}().
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
This patch addresses the issue fully described here:
http://lists.nongnu.org/archive/html/qemu-devel/2013-02/msg01804.html
Linux kernels prior to 2.6.36 do not disable the PCI device during
enumeration process. Since lower and higher parts of a 64bit BAR
are programmed separately this leads to qemu receiving a request to occupy
a completely wrong address region for a short period of time.
We have found that the boot process screws up completely if kvm-apic range
is overlapped even for a short period of time (it is fine for other
regions though).
This patch raises the priority of the kvm-apic memory region, so it is
never pushed out by PCI devices. The patch is quite safe as it does not
touch memory manager.
Signed-off-by: Alexey Korolev <akorolex@gmail.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
As this is the first of the BMI insns to be implemented,
this carries quite a bit more baggage than normal.
Signed-off-by: Richard Henderson <rth@twiddle.net>
In order to instantiate a CPU subtype we will need to know which type,
so move the cpu_model splitting into cpu_x86_init().
Parameters need to be set on the X86CPU instance, so move
cpu_x86_parse_featurestr() into cpu_x86_init() as well.
This leaves cpu_x86_register() operating on the model name only.
Signed-off-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Consolidate CPU functions in cpu.c.
Allows to make cpu_x86_register() static.
No functional changes.
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
The target-specific ENV_GET_CPU() macros have allowed us to navigate
from CPUArchState to CPUState. The reverse direction was not supported.
Avoid introducing CPU_GET_ENV() macros by initializing an untyped
pointer that is initialized in derived instance_init functions.
The field may not be called "env" due to it being poisoned.
Acked-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Adapt the signature of x86_cpu_realize(), hook up to
DeviceClass::realize and set realized = true in cpu_x86_init().
The QOM realizefn cannot depend on errp being non-NULL as in
cpu_x86_init(), so use a local Error to preserve error handling behavior
on APIC initialization errors.
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
[AF: Invoke parent's realizefn]
Signed-off-by: Andreas Färber <afaerber@suse.de>
Setting tsc-frequency from x86_def_t is NOP because default tsc_khz
in x86_def_t is 0 and CPUX86State.tsc_khz is also initialized to 0
by default. So there is no need to overwrite tsc_khz with default 0
because field was already initialized to 0.
Custom tsc-frequency setting is not affected due to it being set
without using x86_def_t.
Field tsc_khz in x86_def_t becomes unused with this patch, so drop it
as well.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Move custom features parsing after built-in cpu_model defaults are set
and set custom features directly on CPU instance. That allows to make a
clear distinction between built-in cpu model defaults that eventually
should go into class_init() and extra property setting which is done
after defaults are set on CPU instance.
Impl. details:
* use object_property_parse() property setter so it would be a mechanical
change to switch to global properties later.
* And after all current features/properties are converted into static
properties, it will take a trivial patch to switch to global properties.
Which will allow to:
* get CPU instance initialized with all parameters passed on -cpu ...
cmd. line from object_new() call.
* call cpu_model/featurestr parsing only once before CPUs are created
* open a road for removing CPUxxxState.cpu_model_str field, when other
CPUs are similarly converted to subclasses and static properties.
- re-factor error handling, to use Error instead of fprintf()s, since
it is anyway passed in for property setter.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Commit 8935499831 makes cpuid return to guest host's vendor value
instead of built-in one by default if kvm_enabled() == true and allows
to override this behavior if 'vendor' is specified on -cpu command line.
But every time guest calls cpuid to get 'vendor' value, host's value is
read again and again in default case.
It complicates semantics of vendor property and makes it harder to use.
Instead of reading 'vendor' value from host every time cpuid[vendor] is
called, override 'vendor' value only once in cpu_x86_find_by_name(), when
built-in CPU model is found and if(kvm_enabled() == true).
It provides the same default semantics
if (kvm_enabled() == true) vendor = host's vendor
else vendor = built-in vendor
and then later:
if (custom vendor) vendor = custom vendor
'vendor' value is overridden when user provides it on -cpu command line,
and there is no need for vendor_override field anymore, remove it.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Vendor property setter takes string as vendor value but cpudefs
use uint32_t vendor[123] fields to define vendor value. It makes it
difficult to unify and use property setter for values from cpudefs.
Simplify code by using vendor property setter, vendor[123] fields
are converted into vendor[13] array to keep its value. And vendor
property setter is used to access/set value on CPU.
- Make for() cycle reusable for the next patch by adding
x86_cpu_vendor_words2str()
Intel's CPUID spec[1] says:
"
5.1.1 ...
These registers contain the ASCII string: GenuineIntel
...
"
List[2] of known vendor values shows that they all are 12 ASCII
characters long, padded where necessary with space.
Current supported values are all ASCII characters packed in
ebx, edx, ecx. So lets state that QEMU supports 12 printable ASCII
characters packed in ebx, edx, ecx registers for cpuid(0) instruction.
*1 - http://www.intel.com/Assets/PDF/appnote/241618.pdf
*2 - http://en.wikipedia.org/wiki/CPUID#EAX.3D0:_Get_vendor_ID
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
It is no longer needed since dropping cpudef config file support.
Cleaning this up removes knowledge about other models from x86_def_t,
in preparation for reusing x86_def_t as intermediate step towards pure
QOM X86CPU subclasses.
Signed-off-by: Andreas Färber <afaerber@suse.de>
Catch NULL name argument early to avoid repeated checks.
Similarly, check for -cpu host early and untangle from iterating through
model definitions. This prepares for introducing X86CPU subclasses.
Signed-off-by: Andreas Färber <afaerber@suse.de>
This keeps compatibility on machine-types pc-1.2 and older, and prints a
warning in case the requested configuration won't get the correct
topology.
I couldn't think of a better way to warn about broken topology when in
compat mode other than using error_report(). The warning message will
probably be buried in a log file somewhere, but it's better than
nothing.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
This function will be used by both the CPU initialization code and the
fw_cfg table initialization code.
Later this function will be updated to generate APIC IDs according to
the CPU topology.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Currently, the pc-1.4 machine init function enables PV EOI and then
calls the pc-1.2 machine init function. The problem with this approach
is that now we can't enable any additional compatibility code inside the
pc-1.2 init function because it would end up enabling the compatibility
behavior on pc-1.3 and pc-1.4 as well.
This reverses the logic so that the pc-1.2 machine init function will
disable PV EOI, and then call the pc-1.4 machine init function.
This way we can change older machine-types to enable compatibility
behavior, and the newer machine-types (pc-1.3, pc-q35-1.4 and
pc-i440fx-1.4) would just use the default behavior.
(This means that one nice side-effect of this change is that pc-q35-1.4
will get PV EOI enabled by default, too)
It would be interesting to eventually change pc_init_pci_no_kvmclock()
and pc_init_isa() to reuse pc_init_pci_1_2() as well (so we don't need
to duplicate compatibility code on those two functions). But this will
be probably much easier to do after we create a PCInitArgs struct for
the PC initialization arguments, and/or after we use global-properties
to implement the compatibility modes present in pc_init_pci_1_2().
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
This is a cleanup that tries to solve two small issues:
- We don't need a separate kvm_pv_eoi_features variable just to keep a
constant calculated at compile-time, and this style would require
adding a separate variable (that's declared twice because of the
CONFIG_KVM ifdef) for each feature that's going to be
enabled/disabled by machine-type compat code.
- The pc-1.3 code is setting the kvm_pv_eoi flag on cpuid_kvm_features
even when KVM is disabled at runtime. This small inconsistency in
the cpuid_kvm_features field isn't a problem today because
cpuid_kvm_features is ignored by the TCG code, but it may cause
unexpected problems later when refactoring the CPUID handling code.
This patch eliminates the kvm_pv_eoi_features variable and simply uses
kvm_enabled() inside the enable_kvm_pv_eoi() compat function, so it
enables kvm_pv_eoi only if KVM is enabled. I believe this makes the
behavior of enable_kvm_pv_eoi() clearer and easier to understand.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Acked-by: Gleb Natapov <gleb@redhat.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Replace by SYS_BUS_DEVICE() QOM cast macro using a scripted conversion.
Avoids the old macro creeping into new code.
Resolve a Coding Style warning in openpic code.
Signed-off-by: Andreas Färber <afaerber@suse.de>
Cc: Anthony Liguori <anthony@codemonkey.ws>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
kvm_check_features_against_host() should be called when features can't
be changed, and when features are converted to properties it would be
possible to change them until realize time, so correct way is to call
kvm_check_features_against_host() in x86_cpu_realize().
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Freeing resources in one place would require setting 'error'
to not NULL, so add some more error reporting before jumping to
exit branch.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
No functional change, needed for simplifying conversion to properties.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
This adds the following feature words to the list of flags to be checked
by kvm_check_features_against_host():
- cpuid_7_0_ebx_features
- ext4_features
- kvm_features
- svm_features
This will ensure the "enforce" flag works as it should: it won't allow
QEMU to be started unless every flag that was requested by the user or
defined in the CPU model is supported by the host.
This patch may cause existing configurations where "enforce" wasn't
preventing QEMU from being started to abort QEMU. But that's exactly the
point of this patch: if a flag was not supported by the host and QEMU
wasn't aborting, it was a bug in the "enforce" code.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Feature names were taken from the X86_FEATURE_* constants in the Linux
kernel code.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Instead of carrying the CPUID leaf/register and feature name array on
the model_features_t struct, move that information into
feature_word_info so it can be reused by other functions.
The goal is to eventually kill model_features_t entirely, but to do that
we have to either convert x86_def_t.features to an array or use
offsetof() inside FeatureWordInfo (to replace the pointers inside
model_features_t). So by now just move most of the model_features_t
fields to FeatureWordInfo except for the two pointers to local
arguments.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
This introduces a FeatureWord enum, FeatureWordInfo struct (with
generation information about a feature word), and a FeatureWordArray
typedef, and changes add_flagname_to_bitmaps() code and
cpu_x86_parse_featurestr() to use the new typedefs instead of separate
variables for each feature word.
This will help us keep the code at kvm_check_features_against_host(),
cpu_x86_parse_featurestr() and add_flagname_to_bitmaps() sane while
adding new feature name arrays.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
KVM_CAP_PV_MMU capability reporting was removed from the kernel since
v2.6.33 (see commit a68a6a7282373), and was completely removed from the
kernel since v3.3 (see commit fb92045843). It doesn't make sense to keep
it enabled by default, as it would cause unnecessary hassle when using
the "enforce" flag.
This disables kvm_mmu on all machine-types. With this fix, the possible
scenarios when migrating from QEMU <= 1.3 to QEMU 1.4 are:
------------+----------+----------------------------------------------------
src kernel | dst kern.| Result
------------+----------+----------------------------------------------------
>= 2.6.33 | any | kvm_mmu was already disabled and will stay disabled
<= 2.6.32 | >= 3.3 | correct live migration is impossible
<= 2.6.32 | <= 3.2 | kvm_mmu will be disabled on next guest reboot *
------------+----------+----------------------------------------------------
* If they are running kernel <= 2.6.32 and want kvm_mmu to be kept
enabled on guest reboot, they can explicitly add +kvm_mmu to the QEMU
command-line. Using 2.6.33 and higher, it is not possible to enable
kvm_mmu explicitly anymore.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Note that target-alpha accesses this field from TCG, now using a
negative offset. Therefore the field is placed last in CPUState.
Pass PowerPCCPU to [kvm]ppc_fixup_cpu() to facilitate this change.
Move common parts of mips cpu_state_reset() to mips_cpu_reset().
Acked-by: Richard Henderson <rth@twiddle.net> (for alpha)
[AF: Rebased onto ppc CPU subclasses and openpic changes]
Signed-off-by: Andreas Färber <afaerber@suse.de>
To facilitate the field movements, pass MIPSCPU to malta_mips_config();
avoid that for mips_cpu_map_tc() since callers only access MIPS Thread
Contexts, inside TCG helpers.
Signed-off-by: Andreas Färber <afaerber@suse.de>
Since cpudef config is not supported anymore and all remaining sources
now always set x86_def_t.vendor[123] fields, remove setting default
vendor to simplify future re-factoring.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
When CPU properties are implemented, ext2_features may change
between object_new(CPU) and cpu_realize_fn(). Sanitizing
ext2_features for AMD based CPU at realize() time will keep
current behavior after CPU features are converted to properties.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
