The "acknowledge interrupt on exit" feature controls processor behavior
for external interrupt acknowledgement. When this control is set, the
processor acknowledges the interrupt controller to acquire the
interrupt vector on VM exit.

After enabling this feature, an interrupt which arrived when target cpu is
running in vmx non-root mode will be handled by vmx handler instead of handler
in idt. Currently, vmx handler only fakes an interrupt stack and jump to idt
table to let real handler to handle it. Further, we will recognize the interrupt
and only delivery the interrupt which not belong to current vcpu through idt table.
The interrupt which belonged to current vcpu will be handled inside vmx handler.
This will reduce the interrupt handle cost of KVM.

Also, interrupt enable logic is changed if this feature is turnning on:
Before this patch, hypervior call local_irq_enable() to enable it directly.
Now IF bit is set on interrupt stack frame, and will be enabled on a return from
interrupt handler if exterrupt interrupt exists. If no external interrupt, still
call local_irq_enable() to enable it.

Refer to Intel SDM volum 3, chapter 33.2.
Signed-off-by: NYang Zhang <yang.z.zhang@Intel.com>
Reviewed-by: NGleb Natapov <gleb@redhat.com>
Signed-off-by: NMarcelo Tosatti <mtosatti@redhat.com>

Userspace may deliver RTC interrupt without query the status. So we
want to track RTC EOI for this case.
Signed-off-by: NYang Zhang <yang.z.zhang@Intel.com>
Reviewed-by: NGleb Natapov <gleb@redhat.com>
Signed-off-by: NMarcelo Tosatti <mtosatti@redhat.com>

restore rtc_status from migration or save/restore
Signed-off-by: NYang Zhang <yang.z.zhang@Intel.com>
Reviewed-by: NGleb Natapov <gleb@redhat.com>
Signed-off-by: NMarcelo Tosatti <mtosatti@redhat.com>

Add a new parameter to know vcpus who received the interrupt.
Signed-off-by: NYang Zhang <yang.z.zhang@Intel.com>
Reviewed-by: NGleb Natapov <gleb@redhat.com>
Signed-off-by: NMarcelo Tosatti <mtosatti@redhat.com>

Add vcpu info to ioapic_update_eoi, so we can know which vcpu
issued this EOI.
Signed-off-by: NYang Zhang <yang.z.zhang@Intel.com>
Reviewed-by: NGleb Natapov <gleb@redhat.com>
Signed-off-by: NMarcelo Tosatti <mtosatti@redhat.com>

We only need to update vm_exit_intr_error_code if there is a valid exit
interruption information and it comes with a valid error code.
Signed-off-by: NJan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

If we are entering guest mode, we do not want L0 to interrupt this
vmentry with all its side effects on the vmcs. Therefore, injection
shall be disallowed during L1->L2 transitions, as in the previous
version. However, this check is conceptually independent of
nested_exit_on_intr, so decouple it.

If L1 traps external interrupts, we can kick the guest from L2 to L1,
also just like the previous code worked. But we no longer need to
consider L1's idt_vectoring_info_field. It will always be empty at this
point. Instead, if L2 has pending events, those are now found in the
architectural queues and will, thus, prevent vmx_interrupt_allowed from
being called at all.
Signed-off-by: NJan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

The basic idea is to always transfer the pending event injection on
vmexit into the architectural state of the VCPU and then drop it from
there if it turns out that we left L2 to enter L1, i.e. if we enter
prepare_vmcs12.

vmcs12_save_pending_events takes care to transfer pending L0 events into
the queue of L1. That is mandatory as L1 may decide to switch the guest
state completely, invalidating or preserving the pending events for
later injection (including on a different node, once we support
migration).

This concept is based on the rule that a pending vmlaunch/vmresume is
not canceled. Otherwise, we would risk to lose injected events or leak
them into the wrong queues. Encode this rule via a WARN_ON_ONCE at the
entry of nested_vmx_vmexit.
Signed-off-by: NJan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

Check if the interrupt or NMI window exit is for L1 by testing if it has
the corresponding controls enabled. This is required when we allow
direct injection from L0 to L2
Signed-off-by: NJan Kiszka <jan.kiszka@siemens.com>
Reviewed-by: NGleb Natapov <gleb@redhat.com>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

Signed-off-by: NGleb Natapov <gleb@redhat.com>

Emulation of undefined opcode should inject #UD instead of causing
emulation failure. Do that by moving Undefined flag check to emulation
stage and injection #UD there.
Signed-off-by: NGleb Natapov <gleb@redhat.com>

During invalid guest state emulation vcpu cannot enter guest mode to try
to reexecute instruction that emulator failed to emulate, so emulation
will happen again and again.  Prevent that by telling the emulator that
instruction reexecution should not be attempted.
Signed-off-by: NGleb Natapov <gleb@redhat.com>

Unimplemented instruction detection is broken for group instructions
since it relies on "flags" field of opcode to be zero, but all
instructions in a group inherit flags from a group encoding. Fix that by
having a separate flag for unimplemented instructions.
Signed-off-by: NGleb Natapov <gleb@redhat.com>

This fixes a regression introduced in commit 03ebebeb ("KVM: x86
emulator: Leave segment limit and attributs alone in real mode").

The mentioned commit changed the segment descriptors for both real mode
and VM86 to only update the segment base instead of creating a
completely new descriptor with limit 0xffff so that unreal mode keeps
working across a segment register reload.

This leads to an invalid segment descriptor in the eyes of VMX, which
seems to be okay for real mode because KVM will fix it up before the
next VM entry or emulate the state, but it doesn't do this if the guest
is in VM86, so we end up with:

  KVM: entry failed, hardware error 0x80000021

Fix this by effectively reverting commit 03ebebeb for VM86 and leaving
it only in place for real mode, which is where it's really needed.
Signed-off-by: NKevin Wolf <kwolf@redhat.com>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

The variable kvm_rebooting is a common kvm variable, so move its
declaration from arch/x86/include/asm/kvm_host.h to
include/asm/kvm_host.h.

Fixes this sparse warning when building on arm64:

  virt/kvm/kvm_main.c:warning: symbol 'kvm_rebooting' was not declared. Should it be static?
Signed-off-by: NGeoff Levand <geoff@infradead.org>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

The routine kvm_spurious_fault() is an x86 specific routine, so
move it from virt/kvm/kvm_main.c to arch/x86/kvm/x86.c.

Fixes this sparse warning when building on arm64:

  virt/kvm/kvm_main.c:warning: symbol 'kvm_spurious_fault' was not declared. Should it be static?
Signed-off-by: NGeoff Levand <geoff@infradead.org>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

The variables vm_list and kvm_lock are common to all architectures, so
move the declarations from arch/x86/include/asm/kvm_host.h to
include/linux/kvm_host.h.

Fixes sparse warnings like these when building for arm64:

  virt/kvm/kvm_main.c: warning: symbol 'kvm_lock' was not declared. Should it be static?
  virt/kvm/kvm_main.c: warning: symbol 'vm_list' was not declared. Should it be static?
Signed-off-by: NGeoff Levand <geoff@infradead.org>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

The code was already properly aligned, now also add the braces to avoid
that err is checked even if alloc_apic_access_page didn't run and change
it. Found via Coccinelle by Fengguang Wu.
Signed-off-by: NJan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

Free vmx_msr_bitmap_longmode_x2apic and vmx_msr_bitmap_longmode if
kvm_init() fails.
Signed-off-by: NYang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

Obviously a copy&paste mistake: prepare_vmcs12 has to check L1's exit
controls for VM_EXIT_SAVE_IA32_PAT.
Signed-off-by: NJan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

For a given vcpu, kvm_apic_match_dest() will tell you whether
the vcpu in the destination list quickly. Drop kvm_calculate_eoi_exitmap()
and use kvm_apic_match_dest() instead.
Signed-off-by: NYang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

With the following commit, shadow pages can be zapped at random during
a shadow page talbe walk:
  KVM: MMU: Move kvm_mmu_free_some_pages() into kvm_mmu_alloc_page()
  7ddca7e4

This patch reverts it and fixes __direct_map() and FNAME(fetch)().
Signed-off-by: NTakuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

In order to migrate the PMU state correctly, we need to restore the
values of MSR_CORE_PERF_GLOBAL_STATUS (a read-only register) and
MSR_CORE_PERF_GLOBAL_OVF_CTRL (which has side effects when written).
We also need to write the full 40-bit value of the performance counter,
which would only be possible with a v3 architectural PMU's full-width
counter MSRs.

To distinguish host-initiated writes from the guest's, pass the
full struct msr_data to kvm_pmu_set_msr.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

The current name "kvm_mmu_free_some_pages" should be used for something
that actually frees some shadow pages, as we expect from the name, but
what the function is doing is to make some, KVM_MIN_FREE_MMU_PAGES,
shadow pages available: it does nothing when there are enough.

This patch changes the name to reflect this meaning better; while doing
this renaming, the code in the wrapper function is inlined into the main
body since the whole function will be inlined into the only caller now.
Signed-off-by: NTakuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Signed-off-by: NMarcelo Tosatti <mtosatti@redhat.com>

What this function is doing is to ensure that the number of shadow pages
does not exceed the maximum limit stored in n_max_mmu_pages: so this is
placed at every code path that can reach kvm_mmu_alloc_page().

Although it might have some sense to spread this function in each such
code path when it could be called before taking mmu_lock, the rule was
changed not to do so.

Taking this background into account, this patch moves it into
kvm_mmu_alloc_page() and simplifies the code.

Note: the unlikely hint in kvm_mmu_free_some_pages() guarantees that the
overhead of this function is almost zero except when we actually need to
allocate some shadow pages, so we do not need to care about calling it
multiple times in one path by doing kvm_mmu_get_page() a few times.
Signed-off-by: NTakuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Signed-off-by: NMarcelo Tosatti <mtosatti@redhat.com>

The CS base was initialized to 0 on VMX (wrong, but usually overridden
by userspace before starting) or 0xf0000 on SVM.  The correct value is
0xffff0000, and VMX is able to emulate it now, so use it.
Reviewed-by: NGleb Natapov <gleb@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NMarcelo Tosatti <mtosatti@redhat.com>

There is a potential use after free issue with the handling of
MSR_KVM_SYSTEM_TIME.  If the guest specifies a GPA in a movable or removable
memory such as frame buffers then KVM might continue to write to that
address even after it's removed via KVM_SET_USER_MEMORY_REGION.  KVM pins
the page in memory so it's unlikely to cause an issue, but if the user
space component re-purposes the memory previously used for the guest, then
the guest will be able to corrupt that memory.

Tested: Tested against kvmclock unit test
Signed-off-by: NAndrew Honig <ahonig@google.com>
Signed-off-by: NMarcelo Tosatti <mtosatti@redhat.com>

If the guest sets the GPA of the time_page so that the request to update the
time straddles a page then KVM will write onto an incorrect page.  The
write is done byusing kmap atomic to get a pointer to the page for the time
structure and then performing a memcpy to that page starting at an offset
that the guest controls.  Well behaved guests always provide a 32-byte aligned
address, however a malicious guest could use this to corrupt host kernel
memory.

Tested: Tested against kvmclock unit test.
Signed-off-by: NAndrew Honig <ahonig@google.com>
Signed-off-by: NMarcelo Tosatti <mtosatti@redhat.com>

There is a deadlock in pvclock handling:

cpu0:                                               cpu1:
kvm_gen_update_masterclock()
                                              kvm_guest_time_update()
 spin_lock(pvclock_gtod_sync_lock)
                                               local_irq_save(flags)

spin_lock(pvclock_gtod_sync_lock)

 kvm_make_mclock_inprogress_request(kvm)
  make_all_cpus_request()
   smp_call_function_many()

Now if smp_call_function_many() called by cpu0 tries to call function on
cpu1 there will be a deadlock.

Fix by moving pvclock_gtod_sync_lock protected section outside irq
disabled section.

Analyzed by Gleb Natapov <gleb@redhat.com>
Acked-by: NGleb Natapov <gleb@redhat.com>
Reported-and-Tested-by: NYongjie Ren <yongjie.ren@intel.com>
Signed-off-by: NMarcelo Tosatti <mtosatti@redhat.com>

Very old user space (namely qemu-kvm before kvm-49) didn't set the TSS
base before running the VCPU. We always warned about this bug, but no
reports about users actually seeing this are known. Time to finally
remove the workaround that effectively prevented to call vmx_vcpu_reset
while already holding the KVM srcu lock.
Reviewed-by: NGleb Natapov <gleb@redhat.com>
Signed-off-by: NJan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: NMarcelo Tosatti <mtosatti@redhat.com>

Commit 1d9d8639 ("perf,x86: fix kernel crash with PEBS/BTS after
suspend/resume") fixed a crash when doing PEBS performance profiling
after resuming, but in using init_debug_store_on_cpu() to restore the
DS_AREA mtrr it also resulted in a new WARN_ON() triggering.

init_debug_store_on_cpu() uses "wrmsr_on_cpu()", which in turn uses CPU
cross-calls to do the MSR update.  Which is not really valid at the
early resume stage, and the warning is quite reasonable.  Now, it all
happens to _work_, for the simple reason that smp_call_function_single()
ends up just doing the call directly on the CPU when the CPU number
matches, but we really should just do the wrmsr() directly instead.

This duplicates the wrmsr() logic, but hopefully we can just remove the
wrmsr_on_cpu() version eventually.
Reported-and-tested-by: NParag Warudkar <parag.lkml@gmail.com>
Cc: stable@vger.kernel.org
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This patch fixes a kernel crash when using precise sampling (PEBS)
after a suspend/resume. Turns out the CPU notifier code is not invoked
on CPU0 (BP). Therefore, the DS_AREA (used by PEBS) is not restored properly
by the kernel and keeps it power-on/resume value of 0 causing any PEBS
measurement to crash when running on CPU0.

The workaround is to add a hook in the actual resume code to restore
the DS Area MSR value. It is invoked for all CPUS. So for all but CPU0,
the DS_AREA will be restored twice but this is harmless.
Reported-by: NLinus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: NStephane Eranian <eranian@google.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

When we create or move a memory slot, we need to zap mmio sptes.
Currently, zap_all() is used for this and this is causing two problems:
 - extra page faults after zapping mmu pages
 - long mmu_lock hold time during zapping mmu pages

For the latter, Marcelo reported a disastrous mmu_lock hold time during
hot-plug, which made the guest unresponsive for a long time.

This patch takes a simple way to fix these problems: do not zap mmu
pages unless they are marked mmio cached.  On our test box, this took
only 50us for the 4GB guest and we did not see ms of mmu_lock hold time
any more.

Note that we still need to do zap_all() for other cases.  So another
work is also needed: Xiao's work may be the one.
Reviewed-by: NMarcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: NTakuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

This will be used not to zap unrelated mmu pages when creating/moving
a memory slot later.
Reviewed-by: NMarcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: NTakuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

Provided the host has this feature, it's straightforward to offer it to
the guest as well. We just need to load to timer value on L2 entry if
the feature was enabled by L1 and watch out for the corresponding exit
reason.
Reviewed-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NJan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

We will need EFER.LMA saving to provide unrestricted guest mode. All
what is missing for this is picking up EFER.LMA from VM_ENTRY_CONTROLS
on L2->L1 switches. If the host does not support EFER.LMA saving,
no change is performed, otherwise we properly emulate for L1 what the
hardware does for L0. Advertise the support, depending on the host
feature.
Reviewed-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NJan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

Only interrupt and NMI exiting are mandatory for KVM to work, thus can
be exposed to the guest unconditionally, virtual NMI exiting is
optional. So we must not advertise it unless the host supports it.

Introduce the symbolic constant PIN_BASED_ALWAYSON_WITHOUT_TRUE_MSR at
this chance.
Reviewed-by: N: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NJan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

A VCPU sending INIT or SIPI to some other VCPU races for setting the
remote VCPU's mp_state. When we were unlucky, KVM_MP_STATE_INIT_RECEIVED
was overwritten by kvm_emulate_halt and, thus, got lost.

This introduces APIC events for those two signals, keeping them in
kvm_apic until kvm_apic_accept_events is run over the target vcpu
context. kvm_apic_has_events reports to kvm_arch_vcpu_runnable if there
are pending events, thus if vcpu blocking should end.

The patch comes with the side effect of effectively obsoleting
KVM_MP_STATE_SIPI_RECEIVED. We still accept it from user space, but
immediately translate it to KVM_MP_STATE_INIT_RECEIVED + KVM_APIC_SIPI.
The vcpu itself will no longer enter the KVM_MP_STATE_SIPI_RECEIVED
state. That also means we no longer exit to user space after receiving a
SIPI event.

Furthermore, we already reset the VCPU on INIT, only fixing up the code
segment later on when SIPI arrives. Moreover, we fix INIT handling for
the BSP: it never enter wait-for-SIPI but directly starts over on INIT.
Tested-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NJan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

As noticed by Ulrich Obergfell <uobergfe@redhat.com>, the mmu
counters are for beancounting purposes only - so n_used_mmu_pages and
n_max_mmu_pages could be relaxed (example: before f0f5933a),
resulting in n_used_mmu_pages > n_max_mmu_pages.

Make code robust against n_used_mmu_pages > n_max_mmu_pages.
Reviewed-by: NXiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: NMarcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: NGleb Natapov <gleb@redhat.com>

In commit 887cbce0 ("arch Kconfig: centralise ARCH_NO_VIRT_TO_BUS")
I introduced the config sybmol HAVE_VIRT_TO_BUS and selected that where
needed.  I am not sure what I was thinking.  Instead, just directly
select VIRT_TO_BUS where it is needed.
Signed-off-by: NStephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>