Already used by nested SVM for tracing nested vmexit: kvm_nested_vmexit
marks exits from L2 to L0 while kvm_nested_vmexit_inject marks vmexits
that are reflected to L1.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Instead of fixing up the vmcs12 after the nested vmexit, pass key
parameters already when calling nested_vmx_vmexit. This will help
tracing those vmexits.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When userspace sets MSR_IA32_FEATURE_CONTROL to 0, make sure we leave
root and non-root mode, fully disabling VMX. The register state of the
VCPU is undefined after this step, so userspace has to set it to a
proper state afterward.
This enables to reboot a VM while it is running some hypervisor code.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
According to the SDM, only bits 0-3 of DR6 "may" be cleared by "certain"
debug exception. So do update them on #DB exception in KVM, but leave
the rest alone, only setting BD and BS in addition to already set bits
in DR6. This also aligns us with kvm_vcpu_check_singlestep.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
In contrast to VMX, SVM dose not automatically transfer DR6 into the
VCPU's arch.dr6. So if we face a DR6 read, we must consult a new vendor
hook to obtain the current value. And as SVM now picks the DR6 state
from its VMCB, we also need a set callback in order to write updates of
DR6 back.
Fixes a regression of 020df0794f.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Whenever we change arch.dr7, we also have to call kvm_update_dr7. In
case guest debugging is off, this will synchronize the new state into
hardware.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off: Peter Lieven <pl@kamp.de>
Signed-off: Gleb Natapov
Signed-off: Vadim Rozenfeld <vrozenfe@redhat.com>
After some consideration I decided to submit only Hyper-V reference
counters support this time. I will submit iTSC support as a separate
patch as soon as it is ready.
v1 -> v2
1. mark TSC page dirty as suggested by
Eric Northup <digitaleric@google.com> and Gleb
2. disable local irq when calling get_kernel_ns,
as it was done by Peter Lieven <pl@amp.de>
3. move check for TSC page enable from second patch
to this one.
v3 -> v4
Get rid of ref counter offset.
v4 -> v5
replace __copy_to_user with kvm_write_guest
when updateing iTSC page.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
After the previous patch from Marcelo, the comment before this write
became obsolete. In fact, the write is unnecessary. The calls to
kvm_write_tsc ultimately result in a master clock update as soon as
all TSCs agree and the master clock is re-enabled. This master
clock update will rewrite tsc_timestamp.
So, together with the comment, delete the dead write too.
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
To fix a problem related to different resolution of TSC and system clock,
the offset in TSC units is approximated by
delta = vcpu->hv_clock.tsc_timestamp - vcpu->last_guest_tsc
(Guest TSC value at (Guest TSC value at last VM-exit)
the last kvm_guest_time_update
call)
Delta is then later scaled using mult,shift pair found in hv_clock
structure (which is correct against tsc_timestamp in that
structure).
However, if a frequency change is performed between these two points,
this delta is measured using different TSC frequencies, but scaled using
mult,shift pair for one frequency only.
The end result is an incorrect delta.
The bug which this code works around is not the only cause for
clock backwards events. The global accumulator is still
necessary, so remove the max_kernel_ns fix and rely on the
global accumulator for no clock backwards events.
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Commit e66d2ae7c6 moved the assignment
vcpu->arch.apic_base = value above a condition with
(vcpu->arch.apic_base ^ value), causing that check
to always fail. Use old_value, vcpu->arch.apic_base's
old value, in the condition instead.
Signed-off-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Limit PIT timer frequency similarly to the limit applied by
LAPIC timer.
Cc: stable@kernel.org
Reviewed-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
After free_loaded_vmcs executes, the "loaded_vmcs" structure
is kfreed, and now vmx->loaded_vmcs points to a kfreed area.
Subsequent free_loaded_vmcs then attempts to manipulate
vmx->loaded_vmcs.
Switch the order to avoid the problem.
https://bugzilla.redhat.com/show_bug.cgi?id=1047892
Reviewed-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
fix the 'vcpi' typos when apic_debug is enabled.
Signed-off-by: Chen Fan <chen.fan.fnst@cn.fujitsu.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
According to Table C-1 of Intel SDM 3C, a VM exit happens on an I/O instruction when
"use I/O bitmaps" VM-execution control was 0 _and_ the "unconditional I/O exiting"
VM-execution control was 1. So we can't just check "unconditional I/O exiting" alone.
This patch was improved by suggestion from Jan Kiszka.
Reviewed-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Zhihui Zhang <zzhsuny@gmail.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Three reasons for doing this: 1. arch.walk_mmu points to arch.mmu anyway
in case nested EPT wasn't in use. 2. this aligns VMX with SVM. But 3. is
most important: nested_cpu_has_ept(vmcs12) queries the VMCS page, and if
one guest VCPU manipulates the page of another VCPU in L2, we may be
fooled to skip over the nested_ept_uninit_mmu_context, leaving mmu in
nested state. That can crash the host later on if nested_ept_get_cr3 is
invoked while L1 already left vmxon and nested.current_vmcs12 became
NULL therefore.
Cc: stable@kernel.org
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Update arch.apic_base before triggering recalculate_apic_map. Otherwise
the recalculation will work against the previous state of the APIC and
will fail to build the correct map when an APIC is hardware-enabled
again.
This fixes a regression of 1e08ec4a13.
Cc: stable@vger.kernel.org
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
It is possible for __direct_map to be called on invalid root_hpa
(-1), two examples:
1) try_async_pf -> can_do_async_pf
-> vmx_interrupt_allowed -> nested_vmx_vmexit
2) vmx_handle_exit -> vmx_interrupt_allowed -> nested_vmx_vmexit
Then to load_vmcs12_host_state and kvm_mmu_reset_context.
Check for this possibility, let fault exception be regenerated.
BZ: https://bugzilla.redhat.com/show_bug.cgi?id=924916
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
If kvm_get_dr or kvm_set_dr reports that it raised a fault, we must not
advance the instruction pointer. Otherwise the exception will hit the
wrong instruction.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
It's a pathological case, but still a valid one: If L1 disables APIC
virtualization and also allows L2 to directly write to the APIC page, we
have to forcibly enable APIC virtualization while in L2 if the in-kernel
APIC is in use.
This allows to run the direct interrupt test case in the vmx unit test
without x2APIC.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Giving proper names to the 0 and 1 was once suggested. But since 0 is
returned to the userspace, giving it another name can introduce extra
confusion. This patch just explains the meanings instead.
Signed-off-by: Takuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Since the commit 15ad7146 ("KVM: Use the scheduler preemption notifiers
to make kvm preemptible"), the remaining stuff in this function is a
simple cond_resched() call with an extra need_resched() check which was
there to avoid dropping VCPUs unnecessarily. Now it is meaningless.
Signed-off-by: Takuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
A guest can cause a BUG_ON() leading to a host kernel crash.
When the guest writes to the ICR to request an IPI, while in x2apic
mode the following things happen, the destination is read from
ICR2, which is a register that the guest can control.
kvm_irq_delivery_to_apic_fast uses the high 16 bits of ICR2 as the
cluster id. A BUG_ON is triggered, which is a protection against
accessing map->logical_map with an out-of-bounds access and manages
to avoid that anything really unsafe occurs.
The logic in the code is correct from real HW point of view. The problem
is that KVM supports only one cluster with ID 0 in clustered mode, but
the code that has the bug does not take this into account.
Reported-by: Lars Bull <larsbull@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
In kvm_lapic_sync_from_vapic and kvm_lapic_sync_to_vapic there is the
potential to corrupt kernel memory if userspace provides an address that
is at the end of a page. This patches concerts those functions to use
kvm_write_guest_cached and kvm_read_guest_cached. It also checks the
vapic_address specified by userspace during ioctl processing and returns
an error to userspace if the address is not a valid GPA.
This is generally not guest triggerable, because the required write is
done by firmware that runs before the guest. Also, it only affects AMD
processors and oldish Intel that do not have the FlexPriority feature
(unless you disable FlexPriority, of course; then newer processors are
also affected).
Fixes: b93463aa59 ('KVM: Accelerated apic support')
Reported-by: Andrew Honig <ahonig@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: Andrew Honig <ahonig@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Under guest controllable circumstances apic_get_tmcct will execute a
divide by zero and cause a crash. If the guest cpuid support
tsc deadline timers and performs the following sequence of requests
the host will crash.
- Set the mode to periodic
- Set the TMICT to 0
- Set the mode bits to 11 (neither periodic, nor one shot, nor tsc deadline)
- Set the TMICT to non-zero.
Then the lapic_timer.period will be 0, but the TMICT will not be. If the
guest then reads from the TMCCT then the host will perform a divide by 0.
This patch ensures that if the lapic_timer.period is 0, then the division
does not occur.
Reported-by: Andrew Honig <ahonig@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: Andrew Honig <ahonig@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
We can easily emulate the HLT activity state for L1: If it decides that
L2 shall be halted on entry, just invoke the normal emulation of halt
after switching to L2. We do not depend on specific host features to
provide this, so we can expose the capability unconditionally.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
VM_(ENTRY|EXIT)_CONTROLS vmcs fields are read/written on each guest
entry but most times it can be avoided since values do not changes.
Keep fields copy in memory to avoid unnecessary reads from vmcs.
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
We should not be using jump labels before they were initialized. Push back
the callback to until after jump label initialization.
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
If a nested guest does a NM fault but its CR0 doesn't contain the TS
flag (because it was already cleared by the guest with L1 aid) then we
have to activate FPU ourselves in L0 and then continue to L2. If TS flag
is set then we fallback on the previous behavior, forward the fault to
L1 if it asked for.
Signed-off-by: Anthoine Bourgeois <bourgeois@bertin.fr>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
We need to copy padding to kernel space first before looking at it.
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
I noticed that srcu_read_lock/unlock both have a memory barrier,
so just by moving srcu_read_unlock earlier we can get rid of
one call to smp_mb() using smp_mb__after_srcu_read_unlock instead.
Unsurprisingly, the gain is small but measureable using the unit test
microbenchmark:
before
vmcall in the ballpark of 1410 cycles
after
vmcall in the ballpark of 1360 cycles
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Currently cpuid emulation is traced only when executed by intercept.
Move trace point so that emulator invocation is traced too.
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
All decode_register() callers check if instruction has rex prefix
to properly decode one byte operand. It make sense to move the check
inside.
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
When I was looking at RHEL5.9's failure to start with
unrestricted_guest=0/emulate_invalid_guest_state=1, I got it working with a
slightly older tree than kvm.git. I now debugged the remaining failure,
which was introduced by commit 660696d1 (KVM: X86 emulator: fix
source operand decoding for 8bit mov[zs]x instructions, 2013-04-24)
introduced a similar mis-emulation to the one in commit 8acb4207 (KVM:
fix sil/dil/bpl/spl in the mod/rm fields, 2013-05-30). The incorrect
decoding occurs in 8-bit movzx/movsx instructions whose 8-bit operand
is sil/dil/bpl/spl.
Needless to say, "movzbl %bpl, %eax" does occur in RHEL5.9's decompression
prolog, just a handful of instructions before finally giving control to
the decompressed vmlinux and getting out of the invalid guest state.
Because OpMem8 bypasses decode_modrm, the same handling of the REX prefix
must be applied to OpMem8.
Reported-by: Michele Baldessari <michele@redhat.com>
Cc: stable@vger.kernel.org
Cc: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Yet another instruction that we fail to emulate, this time found
in Windows 2008R2 32-bit.
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
mst can't be blamed for lack of switch entries: the
issue is with msrs actually.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The loop was always using 0 as the index. This means that
any rubbish after the first element of the array went undetected.
It seems reasonable to assume that no KVM userspace did that.
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The KVM_SET_XCRS ioctl must accept anything that KVM_GET_XCRS
could return. XCR0's bit 0 is always 1 in real processors with
XSAVE, and KVM_GET_XCRS will always leave bit 0 set even if the
emulated processor does not have XSAVE. So, KVM_SET_XCRS must
ignore that bit when checking for attempts to enable unsupported
save states.
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
We currently use some ad-hoc arch variables tied to legacy KVM device
assignment to manage emulation of instructions that depend on whether
non-coherent DMA is present. Create an interface for this, adapting
legacy KVM device assignment and adding VFIO via the KVM-VFIO device.
For now we assume that non-coherent DMA is possible any time we have a
VFIO group. Eventually an interface can be developed as part of the
VFIO external user interface to query the coherency of a group.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Default to operating in coherent mode. This simplifies the logic when
we switch to a model of registering and unregistering noncoherent I/O
with KVM.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
So far we've succeeded at making KVM and VFIO mostly unaware of each
other, but areas are cropping up where a connection beyond eventfds
and irqfds needs to be made. This patch introduces a KVM-VFIO device
that is meant to be a gateway for such interaction. The user creates
the device and can add and remove VFIO groups to it via file
descriptors. When a group is added, KVM verifies the group is valid
and gets a reference to it via the VFIO external user interface.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This basically came from the need to be able to boot 32-bit Atom SMP
guests on an AMD host, i.e. a host which doesn't support MOVBE. As a
matter of fact, qemu has since recently received MOVBE support but we
cannot share that with kvm emulation and thus we have to do this in the
host. We're waay faster in kvm anyway. :-)
So, we piggyback on the #UD path and emulate the MOVBE functionality.
With it, an 8-core SMP guest boots in under 6 seconds.
Also, requesting MOVBE emulation needs to happen explicitly to work,
i.e. qemu -cpu n270,+movbe...
Just FYI, a fairly straight-forward boot of a MOVBE-enabled 3.9-rc6+
kernel in kvm executes MOVBE ~60K times.
Signed-off-by: Andre Przywara <andre@andrep.de>
Signed-off-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Add initial support for handling three-byte instructions in the
emulator.
Signed-off-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Call it EmulateOnUD which is exactly what we're trying to do with
vendor-specific instructions.
Rename ->only_vendor_specific_insn to something shorter, while at it.
Signed-off-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Add a field to the current emulation context which contains the
instruction opcode length. This will streamline handling of opcodes of
different length.
Signed-off-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Add a kvm ioctl which states which system functionality kvm emulates.
The format used is that of CPUID and we return the corresponding CPUID
bits set for which we do emulate functionality.
Make sure ->padding is being passed on clean from userspace so that we
can use it for something in the future, after the ioctl gets cast in
stone.
s/kvm_dev_ioctl_get_supported_cpuid/kvm_dev_ioctl_get_cpuid/ while at
it.
Signed-off-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
If the host supports it, we can and should expose it to the guest as
well, just like we already do with PIN_BASED_VIRTUAL_NMIS.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
__vmx_complete_interrupts stored uninjected NMIs in arch.nmi_injected,
not arch.nmi_pending. So we actually need to check the former field in
vmcs12_save_pending_event. This fixes the eventinj unit test when run
in nested KVM.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
As long as the hardware provides us 2MB EPT pages, we can also expose
them to the guest because our shadow EPT code already supports this
feature.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
We will use that in the later patch to find the kvm ops handler
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
Page pinning is not mandatory in kvm async page fault processing since
after async page fault event is delivered to a guest it accesses page once
again and does its own GUP. Drop the FOLL_GET flag in GUP in async_pf
code, and do some simplifying in check/clear processing.
Suggested-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Gu zheng <guz.fnst@cn.fujitsu.com>
Signed-off-by: chai wen <chaiw.fnst@cn.fujitsu.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
This patch contains the following two changes:
1. Fix the bug in nested preemption timer support. If vmexit L2->L0
with some reasons not emulated by L1, preemption timer value should
be save in such exits.
2. Add support of "Save VMX-preemption timer value" VM-Exit controls
to nVMX.
With this patch, nested VMX preemption timer features are fully
supported.
Signed-off-by: Arthur Chunqi Li <yzt356@gmail.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
72f857950f broke shadow on EPT. This patch reverts it and fixes PAE
on nEPT (which reverted commit fixed) in other way.
Shadow on EPT is now broken because while L1 builds shadow page table
for L2 (which is PAE while L2 is in real mode) it never loads L2's
GUEST_PDPTR[0-3]. They do not need to be loaded because without nested
virtualization HW does this during guest entry if EPT is disabled,
but in our case L0 emulates L2's vmentry while EPT is enables, so we
cannot rely on vmcs12->guest_pdptr[0-3] to contain up-to-date values
and need to re-read PDPTEs from L2 memory. This is what kvm_set_cr3()
is doing, but by clearing cache bits during L2 vmentry we drop values
that kvm_set_cr3() read from memory.
So why the same code does not work for PAE on nEPT? kvm_set_cr3()
reads pdptes into vcpu->arch.walk_mmu->pdptrs[]. walk_mmu points to
vcpu->arch.nested_mmu while nested guest is running, but ept_load_pdptrs()
uses vcpu->arch.mmu which contain incorrect values. Fix that by using
walk_mmu in ept_(load|save)_pdptrs.
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Tested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
kvm_mmu initialization is mostly filling in function pointers, there is
no way for it to fail. Clean up unused return values.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
They do the same thing, and destroy_kvm_mmu can be confused with
kvm_mmu_destroy.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
The new_cr3 MMU callback has been a wrapper for mmu_free_roots since commit
e676505 (KVM: MMU: Force cr3 reload with two dimensional paging on mov
cr3 emulation, 2012-07-08).
The commit message mentioned that "mmu_free_roots() is somewhat of an overkill,
but fixing that is more complicated and will be done after this minimal fix".
One year has passed, and no one really felt the need to do a different fix.
Wrap the call with a kvm_mmu_new_cr3 function for clarity, but remove the
callback.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
The free MMU callback has been a wrapper for mmu_free_roots since mmu_free_roots
itself was introduced (commit 17ac10a, [PATCH] KVM: MU: Special treatment
for shadow pae root pages, 2007-01-05), and has always been the same for all
MMU cases. Remove the indirection as it is useless.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
This makes the interface more deterministic for userspace, which can expect
(after configuring only the features it supports) to get exactly the same
state from the kernel, independent of the host CPU and kernel version.
Suggested-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
A guest can still attempt to save and restore XSAVE states even if they
have been masked in CPUID leaf 0Dh. This usually is not visible to
the guest, but is still wrong: "Any attempt to set a reserved bit (as
determined by the contents of EAX and EDX after executing CPUID with
EAX=0DH, ECX= 0H) in XCR0 for a given processor will result in a #GP
exception".
The patch also performs the same checks as __kvm_set_xcr in KVM_SET_XSAVE.
This catches migration from newer to older kernel/processor before the
guest starts running.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
XSAVE entries that KVM does not support are reported by
KVM_GET_SUPPORTED_CPUID for leaf 0Dh index 0 if the host supports them;
they should be left out unless there is also hypervisor support for them.
Sub-leafs are correctly handled in supported_xcr0_bit, fix index 0
to match.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
In commit e935b8372c ("KVM: Convert kvm_lock to raw_spinlock"),
the kvm_lock was made a raw lock. However, the kvm mmu_shrink()
function tries to grab the (non-raw) mmu_lock within the scope of
the raw locked kvm_lock being held. This leads to the following:
BUG: sleeping function called from invalid context at kernel/rtmutex.c:659
in_atomic(): 1, irqs_disabled(): 0, pid: 55, name: kswapd0
Preemption disabled at:[<ffffffffa0376eac>] mmu_shrink+0x5c/0x1b0 [kvm]
Pid: 55, comm: kswapd0 Not tainted 3.4.34_preempt-rt
Call Trace:
[<ffffffff8106f2ad>] __might_sleep+0xfd/0x160
[<ffffffff817d8d64>] rt_spin_lock+0x24/0x50
[<ffffffffa0376f3c>] mmu_shrink+0xec/0x1b0 [kvm]
[<ffffffff8111455d>] shrink_slab+0x17d/0x3a0
[<ffffffff81151f00>] ? mem_cgroup_iter+0x130/0x260
[<ffffffff8111824a>] balance_pgdat+0x54a/0x730
[<ffffffff8111fe47>] ? set_pgdat_percpu_threshold+0xa7/0xd0
[<ffffffff811185bf>] kswapd+0x18f/0x490
[<ffffffff81070961>] ? get_parent_ip+0x11/0x50
[<ffffffff81061970>] ? __init_waitqueue_head+0x50/0x50
[<ffffffff81118430>] ? balance_pgdat+0x730/0x730
[<ffffffff81060d2b>] kthread+0xdb/0xe0
[<ffffffff8106e122>] ? finish_task_switch+0x52/0x100
[<ffffffff817e1e94>] kernel_thread_helper+0x4/0x10
[<ffffffff81060c50>] ? __init_kthread_worker+0x
After the previous patch, kvm_lock need not be a raw spinlock anymore,
so change it back.
Reported-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Cc: kvm@vger.kernel.org
Cc: gleb@redhat.com
Cc: jan.kiszka@siemens.com
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
If #PF happens during delivery of an exception into L2 and L1 also do
not have the page mapped in its shadow page table then L0 needs to
generate vmexit to L2 with original event in IDT_VECTORING_INFO, but
current code combines both exception and generates #DF instead. Fix that
by providing nVMX specific function to handle page faults during page
table walk that handles this case correctly.
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
All exceptions should be checked for intercept during delivery to L2,
but we check only #PF currently. Drop nested_run_pending while we are
at it since exception cannot be injected during vmentry anyway.
Signed-off-by: Gleb Natapov <gleb@redhat.com>
[Renamed the nested_vmx_check_exception function. - Paolo]
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
If an exception causes vmexit directly it should not be reported in
IDT_VECTORING_INFO during the exit. For that we need to be able to
distinguish between exception that is injected into nested VM and one that
is reinjected because its delivery failed. Fortunately we already have
mechanism to do so for nested SVM, so here we just use correct function
to requeue exceptions and make sure that reinjected exception is not
moved to IDT_VECTORING_INFO during vmexit emulation and not re-checked
for interception during delivery.
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
EXIT_REASON_VMLAUNCH/EXIT_REASON_VMRESUME exit does not mean that nested
VM will actually run during next entry. Move setting nested_run_pending
closer to vmentry emulation code and move its clearing close to vmexit to
minimize amount of code that will erroneously run with nested_run_pending
set.
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Bit 12 is undefined in any of the following cases:
- If the "NMI exiting" VM-execution control is 1 and the "virtual NMIs"
VM-execution control is 0.
- If the VM exit sets the valid bit in the IDT-vectoring information field
Signed-off-by: Gleb Natapov <gleb@redhat.com>
[Add parentheses around & within && - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Now that we provide EPT support, there is no reason to torture our
guests by hiding the relieving unrestricted guest mode feature. We just
need to relax CR0 checks for always-on bits as PE and PG can now be
switched off.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Implement and advertise VM_EXIT_SAVE_IA32_EFER. L0 traps EFER writes
unconditionally, so we always find the current L2 value in the
architectural state.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Fiddling with CR3 for L2 is L1's job. It may set its own, different
identity map or simple leave it alone if unrestricted guest mode is
enabled. This also fixes reading back the current CR3 on L2 exits for
reporting it to L1.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
kvm_set_cr0 performs checks on the state transition that may prevent
loading L1's cr0. For now we rely on the hardware to catch invalid
states loaded by L1 into its VMCS. Still, consistency checks on the host
state part of the VMCS on guest entry will have to be improved later on.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Set "blocked by NMI" flag if EPT violation happens during IRET from NMI
otherwise NMI can be called recursively causing stack corruption.
Signed-off-by: Gleb Natapov <gleb@redhat.com>
After nested vmentry stale cache can be used to reload L2 PDPTR pointers
which will cause L2 guest to fail. Fix it by invalidating cache on nested
vmentry emulation.
https://bugzilla.kernel.org/show_bug.cgi?id=60830
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Page tables in a read-only memory slot will currently cause a triple
fault because the page walker uses gfn_to_hva and it fails on such a slot.
OVMF uses such a page table; however, real hardware seems to be fine with
that as long as the accessed/dirty bits are set. Save whether the slot
is readonly, and later check it when updating the accessed and dirty bits.
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Opcode CA
This gets used by a DOS based NetWare guest.
Signed-off-by: Bruce Rogers <brogers@suse.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Convert the remaining couple of random shrinkers in the tree to the new
API.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Glauber Costa <glommer@openvz.org>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Gleb Natapov <gleb@redhat.com>
Cc: Chuck Lever <chuck.lever@oracle.com>
Cc: J. Bruce Fields <bfields@redhat.com>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Cc: Arve Hjønnevåg <arve@android.com>
Cc: Carlos Maiolino <cmaiolino@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Chuck Lever <chuck.lever@oracle.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: David Rientjes <rientjes@google.com>
Cc: Gleb Natapov <gleb@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: J. Bruce Fields <bfields@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: John Stultz <john.stultz@linaro.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Kent Overstreet <koverstreet@google.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Thomas Hellstrom <thellstrom@vmware.com>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
vcpu in page_fault_can_be_fast() is not used so remove it
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
The offset to add to the hosts monotonic time, kvmclock_offset, is
calculated against the monotonic time at KVM_SET_CLOCK ioctl time.
Request a master clock update at this time, to reduce a potentially
unbounded difference between the values of the masterclock and
the clock value used to calculate kvmclock_offset.
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Support for single-step in the emulator (new in 3.12) does not work for
MMIO or PIO writes, because they are completed without returning to
the emulator. This is not worse than what we had in 3.11; still, add
comments so that the issue is not forgotten.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
These will happen due to MMIO.
Suggested-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Note that we are using APIC_DM_REMRD which has reserved usage.
In future if APIC_DM_REMRD usage is standardized, then we should
find some other way or go back to old method.
Suggested-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Raghavendra K T <raghavendra.kt@linux.vnet.ibm.com>
Acked-by: Gleb Natapov <gleb@redhat.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
kvm_hc_kick_cpu allows the calling vcpu to kick another vcpu out of halt state.
the presence of these hypercalls is indicated to guest via
kvm_feature_pv_unhalt.
Fold pv_unhalt flag into GET_MP_STATE ioctl to aid migration
During migration, any vcpu that got kicked but did not become runnable
(still in halted state) should be runnable after migration.
Signed-off-by: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Signed-off-by: Suzuki Poulose <suzuki@in.ibm.com>
[Raghu: Apic related changes, folding pvunhalted into vcpu_runnable
Added flags for future use (suggested by Gleb)]
[ Raghu: fold pv_unhalt flag as suggested by Eric Northup]
Signed-off-by: Raghavendra K T <raghavendra.kt@linux.vnet.ibm.com>
Acked-by: Gleb Natapov <gleb@redhat.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Advertise VM_EXIT_SAVE_IA32_PAT and VM_EXIT_LOAD_IA32_PAT.
Signed-off-by: Arthur Chunqi Li <yzt356@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Do not report that we can enter the guest in 64-bit mode if the host is
32-bit only. This is not supported by KVM.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
At least WB must be possible.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When asking vmx to load the PAT MSR for us while switching from L1 to L2
or vice versa, we have to update arch.pat as well as it may later be
used again to load or read out the MSR content.
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Tested-by: Arthur Chunqi Li <yzt356@gmail.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Some additional comments to preexisting code:
Explain who (L0 or L1) handles EPT violation and misconfiguration exits.
Don't mention "shadow on either EPT or shadow" as the only two options.
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This is the last patch of the basic Nested EPT feature, so as to allow
bisection through this patch series: The guest will not see EPT support until
this last patch, and will not attempt to use the half-applied feature.
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
If we let L1 use EPT, we should probably also support the INVEPT instruction.
In our current nested EPT implementation, when L1 changes its EPT table
for L2 (i.e., EPT12), L0 modifies the shadow EPT table (EPT02), and in
the course of this modification already calls INVEPT. But if last level
of shadow page is unsync not all L1's changes to EPT12 are intercepted,
which means roots need to be synced when L1 calls INVEPT. Global INVEPT
should not be different since roots are synced by kvm_mmu_load() each
time EPTP02 changes.
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
KVM's existing shadow MMU code already supports nested TDP. To use it, we
need to set up a new "MMU context" for nested EPT, and create a few callbacks
for it (nested_ept_*()). This context should also use the EPT versions of
the page table access functions (defined in the previous patch).
Then, we need to switch back and forth between this nested context and the
regular MMU context when switching between L1 and L2 (when L1 runs this L2
with EPT).
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Inject nEPT fault to L1 guest. This patch is original from Xinhao.
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
need_remote_flush() assumes that shadow page is in PT64 format, but
with addition of nested EPT this is no longer always true. Fix it by
bits definitions that depend on host shadow page type.
Reported-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Since nEPT doesn't support A/D bit, so we should not set those bit
when build shadow page table.
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This is the first patch in a series which adds nested EPT support to KVM's
nested VMX. Nested EPT means emulating EPT for an L1 guest so that L1 can use
EPT when running a nested guest L2. When L1 uses EPT, it allows the L2 guest
to set its own cr3 and take its own page faults without either of L0 or L1
getting involved. This often significanlty improves L2's performance over the
previous two alternatives (shadow page tables over EPT, and shadow page
tables over shadow page tables).
This patch adds EPT support to paging_tmpl.h.
paging_tmpl.h contains the code for reading and writing page tables. The code
for 32-bit and 64-bit tables is very similar, but not identical, so
paging_tmpl.h is #include'd twice in mmu.c, once with PTTTYPE=32 and once
with PTTYPE=64, and this generates the two sets of similar functions.
There are subtle but important differences between the format of EPT tables
and that of ordinary x86 64-bit page tables, so for nested EPT we need a
third set of functions to read the guest EPT table and to write the shadow
EPT table.
So this patch adds third PTTYPE, PTTYPE_EPT, which creates functions (prefixed
with "EPT") which correctly read and write EPT tables.
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Some guest paging modes do not support A/D bits. Add support for such
modes in shadow page code. For such modes PT_GUEST_DIRTY_MASK,
PT_GUEST_ACCESSED_MASK, PT_GUEST_DIRTY_SHIFT and PT_GUEST_ACCESSED_SHIFT
should be set to zero.
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This patch makes guest A/D bits definition to be dependable on paging
mode, so when EPT support will be added it will be able to define them
differently.
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
For preparation, we just move gpte_access(), prefetch_invalid_gpte(),
s_rsvd_bits_set(), protect_clean_gpte() and is_dirty_gpte() from mmu.c
to paging_tmpl.h.
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
kvm_set_cr3() attempts to check if the new cr3 is a valid guest physical
address. The problem is that with nested EPT, cr3 is an *L2* physical
address, not an L1 physical address as this test expects.
As the comment above this test explains, it isn't necessary, and doesn't
correspond to anything a real processor would do. So this patch removes it.
Note that this wrong test could have also theoretically caused problems
in nested NPT, not just in nested EPT. However, in practice, the problem
was avoided: nested_svm_vmexit()/vmrun() do not call kvm_set_cr3 in the
nested NPT case, and instead set the vmcb (and arch.cr3) directly, thus
circumventing the problem. Additional potential calls to the buggy function
are avoided in that we don't trap cr3 modifications when nested NPT is
enabled. However, because in nested VMX we did want to use kvm_set_cr3()
(as requested in Avi Kivity's review of the original nested VMX patches),
we can't avoid this problem and need to fix it.
Reviewed-by: Orit Wasserman <owasserm@redhat.com>
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The existing code for handling cr3 and related VMCS fields during nested
exit and entry wasn't correct in all cases:
If L2 is allowed to control cr3 (and this is indeed the case in nested EPT),
during nested exit we must copy the modified cr3 from vmcs02 to vmcs12, and
we forgot to do so. This patch adds this copy.
If L0 isn't controlling cr3 when running L2 (i.e., L0 is using EPT), and
whoever does control cr3 (L1 or L2) is using PAE, the processor might have
saved PDPTEs and we should also save them in vmcs12 (and restore later).
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Orit Wasserman <owasserm@redhat.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Recent KVM, since http://kerneltrap.org/mailarchive/linux-kvm/2010/5/2/6261577
switch the EFER MSR when EPT is used and the host and guest have different
NX bits. So if we add support for nested EPT (L1 guest using EPT to run L2)
and want to be able to run recent KVM as L1, we need to allow L1 to use this
EFER switching feature.
To do this EFER switching, KVM uses VM_ENTRY/EXIT_LOAD_IA32_EFER if available,
and if it isn't, it uses the generic VM_ENTRY/EXIT_MSR_LOAD. This patch adds
support for the former (the latter is still unsupported).
Nested entry and exit emulation (prepare_vmcs_02 and load_vmcs12_host_state,
respectively) already handled VM_ENTRY/EXIT_LOAD_IA32_EFER correctly. So all
that's left to do in this patch is to properly advertise this feature to L1.
Note that vmcs12's VM_ENTRY/EXIT_LOAD_IA32_EFER are emulated by L0, by using
vmx_set_efer (which itself sets one of several vmcs02 fields), so we always
support this feature, regardless of whether the host supports it.
Reviewed-by: Orit Wasserman <owasserm@redhat.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Current code always uses arch.mmu to check the reserved bits on guest gpte
which is valid only for L1 guest, we should use arch.nested_mmu instead when
we translate gva to gpa for the L2 guest
Fix it by using @mmu instead since it is adapted to the current mmu mode
automatically
The bug can be triggered when nested npt is used and L1 guest and L2 guest
use different mmu mode
Reported-by: Jan Kiszka <jan.kiszka@siemens.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
After commit 21feb4eb64 tr base is zeroed
during vmexit. Set it to L1's HOST_TR_BASE. This should fix
https://bugzilla.kernel.org/show_bug.cgi?id=60679
Reported-by: Yongjie Ren <yongjie.ren@intel.com>
Reviewed-by: Arthur Chunqi Li <yzt356@gmail.com>
Tested-by: Yongjie Ren <yongjie.ren@intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
During nested vmentry into vm86 mode a vcpu state is found to be incorrect
because rflags does not have VM flag set since it is read from the cache
and has L1's value instead of L2's. If emulate_invalid_guest_state=1 L0
KVM tries to emulate it, but emulation does not work for nVMX and it
never should happen anyway. Fix that by using vmx_set_rflags() to set
rflags during nested vmentry which takes care of updating register cache.
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This lets debugging work better during emulation of invalid
guest state.
This time the check is done after emulation, but before writeback
of the flags; we need to check the flags *before* execution of the
instruction, we cannot check singlestep_rip because the CS base may
have already been modified.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Conflicts:
arch/x86/kvm/x86.c
This lets debugging work better during emulation of invalid
guest state.
The check is done before emulating the instruction, and (in the case
of guest debugging) reuses EMULATE_DO_MMIO to exit with KVM_EXIT_DEBUG.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
If posted interrupts are enabled, we can no longer track if an IRQ was
coalesced based on IRR. So drop this logic also from the classic
software path and simplify apic_test_and_set_irr to apic_set_irr.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
[KVM maintainers:
The underlying support for this is in perf/core now. So please merge
this patch into the KVM tree.]
This is not arch perfmon, but older CPUs will just ignore it. This makes
it possible to do at least some TSX measurements from a KVM guest
v2: Various fixes to address review feedback
v3: Ignore the bits when no CPUID. No #GP. Force raw events with TSX bits.
v4: Use reserved bits for #GP
v5: Remove obsolete argument
Acked-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When L2 exits to L1, segment infomations of L1 are not set correctly.
According to Intel SDM 27.5.2(Loading Host Segment and Descriptor
Table Registers), segment base/limit/access right of L1 should be
set to some designed value when L2 exits to L1. This patch fixes
this.
Signed-off-by: Arthur Chunqi Li <yzt356@gmail.com>
Reviewed-by: Gleb Natapov <gnatapov@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Fix read/write to IA32_FEATURE_CONTROL MSR in nested environment.
This patch simulate this MSR in nested_vmx and the default value is
0x0. BIOS should set it to 0x5 before VMXON. After setting the lock
bit, write to it will cause #GP(0).
Another QEMU patch is also needed to handle emulation of reset
and migration. Reset to vCPU should clear this MSR and migration
should reserve value of it.
This patch is based on Nadav's previous commit.
http://permalink.gmane.org/gmane.comp.emulators.kvm.devel/88478
Signed-off-by: Nadav Har'El <nyh@math.technion.ac.il>
Signed-off-by: Arthur Chunqi Li <yzt356@gmail.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Void pointers don't need no casting, drop it.
Signed-off-by: Mathias Krause <minipli@googlemail.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Use a const pointer type instead of casting away the const qualifier
from const arrays. Keep the pointer array on the stack, nonetheless.
Making it static just increases the object size.
Signed-off-by: Mathias Krause <minipli@googlemail.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Set rflags after successfully emulateing VMXON/VMXOFF in VMX.
Signed-off-by: Arthur Chunqi Li <yzt356@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Move nested_vmx_succeed/nested_vmx_failInvalid/nested_vmx_failValid
ahead of handle_vmon to eliminate double declaration in the same
file
Signed-off-by: Arthur Chunqi Li <yzt356@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Now that kvm_arch_memslots_updated() catches every increment of the
memslots->generation, checking if the mmio generation has reached its
maximum value is enough.
Signed-off-by: Takuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This is called right after the memslots is updated, i.e. when the result
of update_memslots() gets installed in install_new_memslots(). Since
the memslots needs to be updated twice when we delete or move a memslot,
kvm_arch_commit_memory_region() does not correspond to this exactly.
In the following patch, x86 will use this new API to check if the mmio
generation has reached its maximum value, in which case mmio sptes need
to be flushed out.
Signed-off-by: Takuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Acked-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Currently, fast page fault incorrectly tries to fix mmio page fault when
the generation number is invalid (spte.gen != kvm.gen). It then returns
to guest to retry the fault since it sees the last spte is nonpresent.
This causes an infinite loop.
Since fast page fault only works for direct mmu, the issue exists when
1) tdp is enabled. It is only triggered only on AMD host since on Intel host
the mmio page fault is recognized as ept-misconfig whose handler call
fault-page path with error_code = 0
2) guest paging is disabled. Under this case, the issue is hardly discovered
since paging disable is short-lived and the sptes will be invalid after
memslot changed for 150 times
Fix it by filtering out MMIO page faults in page_fault_can_be_fast.
Reported-by: Markus Trippelsdorf <markus@trippelsdorf.de>
Tested-by: Markus Trippelsdorf <markus@trippelsdorf.de>
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Some userspaces do not preserve unusable property. Since usable
segment has to be present according to VMX spec we can use present
property to amend userspace bug by making unusable segment always
nonpresent. vmx_segment_access_rights() already marks nonpresent segment
as unusable.
Cc: stable@vger.kernel.org # 3.9+
Reported-by: Stefan Pietsch <stefan.pietsch@lsexperts.de>
Tested-by: Stefan Pietsch <stefan.pietsch@lsexperts.de>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
On the x86 side, there are some optimizations and documentation updates.
The big ARM/KVM change for 3.11, support for AArch64, will come through
Catalin Marinas's tree. s390 and PPC have misc cleanups and bugfixes.
There is a conflict due to "s390/pgtable: fix ipte notify bit" having
entered 3.10 through Martin Schwidefsky's s390 tree. This pull request
has additional changes on top, so this tree's version is the correct one.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.13 (GNU/Linux)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=1zpG
-----END PGP SIGNATURE-----
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Pull KVM fixes from Paolo Bonzini:
"On the x86 side, there are some optimizations and documentation
updates. The big ARM/KVM change for 3.11, support for AArch64, will
come through Catalin Marinas's tree. s390 and PPC have misc cleanups
and bugfixes"
* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm: (87 commits)
KVM: PPC: Ignore PIR writes
KVM: PPC: Book3S PR: Invalidate SLB entries properly
KVM: PPC: Book3S PR: Allow guest to use 1TB segments
KVM: PPC: Book3S PR: Don't keep scanning HPTEG after we find a match
KVM: PPC: Book3S PR: Fix invalidation of SLB entry 0 on guest entry
KVM: PPC: Book3S PR: Fix proto-VSID calculations
KVM: PPC: Guard doorbell exception with CONFIG_PPC_DOORBELL
KVM: Fix RTC interrupt coalescing tracking
kvm: Add a tracepoint write_tsc_offset
KVM: MMU: Inform users of mmio generation wraparound
KVM: MMU: document fast invalidate all mmio sptes
KVM: MMU: document fast invalidate all pages
KVM: MMU: document fast page fault
KVM: MMU: document mmio page fault
KVM: MMU: document write_flooding_count
KVM: MMU: document clear_spte_count
KVM: MMU: drop kvm_mmu_zap_mmio_sptes
KVM: MMU: init kvm generation close to mmio wrap-around value
KVM: MMU: add tracepoint for check_mmio_spte
KVM: MMU: fast invalidate all mmio sptes
...
Pull asm/x86 changes from Ingo Molnar:
"Misc changes, with a bigger processor-flags cleanup/reorganization by
Peter Anvin"
* 'x86-asm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86, asm, cleanup: Replace open-coded control register values with symbolic
x86, processor-flags: Fix the datatypes and add bit number defines
x86: Rename X86_CR4_RDWRGSFS to X86_CR4_FSGSBASE
x86, flags: Rename X86_EFLAGS_BIT1 to X86_EFLAGS_FIXED
linux/const.h: Add _BITUL() and _BITULL()
x86/vdso: Convert use of typedef ctl_table to struct ctl_table
x86: __force_order doesn't need to be an actual variable
This reverts most of the f1ed0450a5. After
the commit kvm_apic_set_irq() no longer returns accurate information
about interrupt injection status if injection is done into disabled
APIC. RTC interrupt coalescing tracking relies on the information to be
accurate and cannot recover if it is not.
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Add a tracepoint write_tsc_offset for tracing TSC offset change.
We want to merge ftrace's trace data of guest OSs and the host OS using
TSC for timestamp in chronological order. We need "TSC offset" values for
each guest when merge those because the TSC value on a guest is always the
host TSC plus guest's TSC offset. If we get the TSC offset values, we can
calculate the host TSC value for each guest events from the TSC offset and
the event TSC value. The host TSC values of the guest events are used when we
want to merge trace data of guests and the host in chronological order.
(Note: the trace_clock of both the host and the guest must be set x86-tsc in
this case)
This tracepoint also records vcpu_id which can be used to merge trace data for
SMP guests. A merge tool will read TSC offset for each vcpu, then the tool
converts guest TSC values to host TSC values for each vcpu.
TSC offset is stored in the VMCS by vmx_write_tsc_offset() or
vmx_adjust_tsc_offset(). KVM executes the former function when a guest boots.
The latter function is executed when kvm clock is updated. Only host can read
TSC offset value from VMCS, so a host needs to output TSC offset value
when TSC offset is changed.
Since the TSC offset is not often changed, it could be overwritten by other
frequent events while tracing. To avoid that, I recommend to use a special
instance for getting this event:
1. set a instance before booting a guest
# cd /sys/kernel/debug/tracing/instances
# mkdir tsc_offset
# cd tsc_offset
# echo x86-tsc > trace_clock
# echo 1 > events/kvm/kvm_write_tsc_offset/enable
2. boot a guest
Signed-off-by: Yoshihiro YUNOMAE <yoshihiro.yunomae.ez@hitachi.com>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Gleb Natapov <gleb@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Acked-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Without this information, users will just see unexpected performance
problems and there is little chance we will get good reports from them:
note that mmio generation is increased even when we just start, or stop,
dirty logging for some memory slot, in which case users cannot expect
all shadow pages to be zapped.
printk_ratelimited() is used for this taking into account the problems
that we can see the information many times when we start multiple VMs
and guests can trigger this by reading ROM in a loop for example.
Signed-off-by: Takuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Document it to Documentation/virtual/kvm/mmu.txt
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Drop kvm_mmu_zap_mmio_sptes and use kvm_mmu_invalidate_zap_all_pages
instead to handle mmio generation number overflow
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Then it has the chance to trigger mmio generation number wrap-around
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
[Change from MMIO_MAX_GEN - 13 to MMIO_MAX_GEN - 150, because 13 is
very close to the number of calls to KVM_SET_USER_MEMORY_REGION
before the guest is started and there is any chance to create any
spte. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This patch tries to introduce a very simple and scale way to invalidate
all mmio sptes - it need not walk any shadow pages and hold mmu-lock
KVM maintains a global mmio valid generation-number which is stored in
kvm->memslots.generation and every mmio spte stores the current global
generation-number into his available bits when it is created
When KVM need zap all mmio sptes, it just simply increase the global
generation-number. When guests do mmio access, KVM intercepts a MMIO #PF
then it walks the shadow page table and get the mmio spte. If the
generation-number on the spte does not equal the global generation-number,
it will go to the normal #PF handler to update the mmio spte
Since 19 bits are used to store generation-number on mmio spte, we zap all
mmio sptes when the number is round
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Store the generation-number into bit3 ~ bit11 and bit52 ~ bit61, totally
19 bits can be used, it should be enough for nearly all most common cases
In this patch, the generation-number is always 0, it will be changed in
the later patch
[Gleb: masking generation bits from spte in get_mmio_spte_gfn() and
get_mmio_spte_access()]
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Bit 1 in the x86 EFLAGS is always set. Name the macro something that
actually tries to explain what it is all about, rather than being a
tautology.
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Gleb Natapov <gleb@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Link: http://lkml.kernel.org/n/tip-f10rx5vjjm6tfnt8o1wseb3v@git.kernel.org
Let mmio spte only use bit62 and bit63 on upper 32 bits, then bit 52 ~ bit 61
can be used for other purposes
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
__kvm_set_xcr function does the CPL check when set xcr. __kvm_set_xcr is
called in two flows, one is invoked by guest, call stack shown as below,
handle_xsetbv(or xsetbv_interception)
kvm_set_xcr
__kvm_set_xcr
the other one is invoked by host, for example during system reset:
kvm_arch_vcpu_ioctl
kvm_vcpu_ioctl_x86_set_xcrs
__kvm_set_xcr
The former does need the CPL check, but the latter does not.
Cc: stable@vger.kernel.org
Signed-off-by: Zhang Haoyu <haoyu.zhang@huawei.com>
[Tweaks to commit message. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Its possible that idivl overflows (due to large delta stored in usdiff,
valid scenario).
Create an exception handler to catch the overflow exception (division by zero
is protected by vcpu->arch.virtual_tsc_khz check), and interpret it accordingly
(delta is larger than USEC_PER_SEC).
Fixes https://bugzilla.redhat.com/show_bug.cgi?id=969644
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Quote Gleb's mail:
| why don't we check for sp->role.invalid in
| kvm_mmu_prepare_zap_page before calling kvm_reload_remote_mmus()?
and
| Actually we can add check for is_obsolete_sp() there too since
| kvm_mmu_invalidate_all_pages() already calls kvm_reload_remote_mmus()
| after incrementing mmu_valid_gen.
[ Xiao: add some comments and the check of is_obsolete_sp() ]
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
As Marcelo pointed out that
| "(retention of large number of pages while zapping)
| can be fatal, it can lead to OOM and host crash"
We introduce a list, kvm->arch.zapped_obsolete_pages, to link all
the pages which are deleted from the mmu cache but not actually
freed. When page reclaiming is needed, we always zap this kind of
pages first.
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
kvm_zap_obsolete_pages uses lock-break technique to zap pages,
it will flush tlb every time when it does lock-break
We can reload mmu on all vcpus after updating the generation
number so that the obsolete pages are not used on any vcpus,
after that we do not need to flush tlb when obsolete pages
are zapped
It will do kvm_mmu_prepare_zap_page many times and use one
kvm_mmu_commit_zap_page to collapse tlb flush, the side-effects
is that causes obsolete pages unlinked from active_list but leave
on hash-list, so we add the comment around the hash list walker
Note: kvm_mmu_commit_zap_page is still needed before free
the pages since other vcpus may be doing locklessly shadow
page walking
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Zap at lease 10 pages before releasing mmu-lock to reduce the overload
caused by requiring lock
After the patch, kvm_zap_obsolete_pages can forward progress anyway,
so update the comments
[ It improves the case 0.6% ~ 1% that do kernel building meanwhile read
PCI ROM. ]
Note: i am not sure that "10" is the best speculative value, i just
guessed that '10' can make vcpu do not spend long time on
kvm_zap_obsolete_pages and do not cause mmu-lock too hungry.
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
The obsolete page will be zapped soon, do not reuse it to
reduce future page fault
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
It is good for debug and development
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
The current kvm_mmu_zap_all is really slow - it is holding mmu-lock to
walk and zap all shadow pages one by one, also it need to zap all guest
page's rmap and all shadow page's parent spte list. Particularly, things
become worse if guest uses more memory or vcpus. It is not good for
scalability
In this patch, we introduce a faster way to invalidate all shadow pages.
KVM maintains a global mmu invalid generation-number which is stored in
kvm->arch.mmu_valid_gen and every shadow page stores the current global
generation-number into sp->mmu_valid_gen when it is created
When KVM need zap all shadow pages sptes, it just simply increase the
global generation-number then reload root shadow pages on all vcpus.
Vcpu will create a new shadow page table according to current kvm's
generation-number. It ensures the old pages are not used any more.
Then the obsolete pages (sp->mmu_valid_gen != kvm->arch.mmu_valid_gen)
are zapped by using lock-break technique
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
It is the responsibility of kvm_mmu_zap_all that keeps the
consistent of mmu and tlbs. And it is also unnecessary after
zap all mmio sptes since no mmio spte exists on root shadow
page and it can not be cached into tlb
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Quote Gleb's mail:
| Back then kvm->lock protected memslot access so code like:
|
| mutex_lock(&vcpu->kvm->lock);
| kvm_mmu_zap_all(vcpu->kvm);
| mutex_unlock(&vcpu->kvm->lock);
|
| which is what 7aa81cc0 does was enough to guaranty that no vcpu will
| run while code is patched. This is no longer the case and
| mutex_lock(&vcpu->kvm->lock); is gone from that code path long time ago,
| so now kvm_mmu_zap_all() there is useless and the code is incorrect.
So we drop it and it will be fixed later
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Jan Kiszka
Vadim Rozenfeld
Paolo Bonzini
Marcelo Tosatti
Andrew Jones
Chen Fan
Zhihui Zhang
Jan Kiszka
Takuya Yoshikawa
Gleb Natapov
Andy Honig
Sasha Levin
Anthoine Bourgeois
Borislav Petkov
Michael S. Tsirkin
Alex Williamson
Aneesh Kumar K.V
chai wen
Arthur Chunqi Li
Bruce Rogers
Dave Chinner
Xiao Guangrong
Raghavendra K T
Srivatsa Vaddagiri
Nadav Har'El
Yang Zhang
Andi Kleen
Nadav Har'El
Mathias Krause
Linus Torvalds
Yoshihiro YUNOMAE
H. Peter Anvin
Zhanghaoyu (A)