When mmu_set_spte() checks if a page related to spte should be release as
dirty or clean, it check if the shadow pte was writeble, but in case
rmap_write_protect() is called called it is possible for shadow ptes that were
writeble to become readonly and therefor mmu_set_spte will release the pages
as clean.
This patch fix this issue by marking the page as dirty inside
rmap_write_protect().
Signed-off-by: Izik Eidus <izike@qumranet.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
If we populate a shadow pte due to a fault (and not speculatively due to a
pte write) then we can set the accessed bit on it, as we know it will be
set immediately on the next guest instruction. This saves a read-modify-write
operation.
Signed-off-by: Avi Kivity <avi@qumranet.com>
Hypercall based pte updates are faster than faults, and also allow use
of the lazy MMU mode to batch operations.
Don't report the feature if two dimensional paging is enabled.
[avi:
- one mmu_op hypercall instead of one per op
- allow 64-bit gpa on hypercall
- don't pass host errors (-ENOMEM) to guest]
[akpm: warning fix on i386]
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Avi Kivity <avi@qumranet.com>
The patch moves the PIT model from userspace to kernel, and increases
the timer accuracy greatly.
[marcelo: make last_injected_time per-guest]
Signed-off-by: Sheng Yang <sheng.yang@intel.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Tested-and-Acked-by: Alex Davis <alex14641@yahoo.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Most Intel hosts have a stable tsc, and playing with the offset only
reduces accuracy. By limiting tsc offset adjustment only to forward updates,
we effectively disable tsc offset adjustment on these hosts.
Signed-off-by: Avi Kivity <avi@qumranet.com>
In the current inject_page_fault path KVM only checks if there is another PF
pending and injects a DF then. But it has to check for a pending DF too to
detect a shutdown condition in the VCPU. If this is not detected the VCPU goes
to a PF -> DF -> PF loop when it should triple fault. This patch detects this
condition and handles it with an KVM_SHUTDOWN exit to userspace.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Create large pages mappings if the guest PTE's are marked as such and
the underlying memory is hugetlbfs backed. If the largepage contains
write-protected pages, a large pte is not used.
Gives a consistent 2% improvement for data copies on ram mounted
filesystem, without NPT/EPT.
Anthony measures a 4% improvement on 4-way kernbench, with NPT.
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Mark zapped root pagetables as invalid and ignore such pages during lookup.
This is a problem with the cr3-target feature, where a zapped root table fools
the faulting code into creating a read-only mapping. The result is a lockup
if the instruction can't be emulated.
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
In two case statements, use the ever popular 'i' instead of index:
arch/x86/kvm/x86.c:1063:7: warning: symbol 'index' shadows an earlier one
arch/x86/kvm/x86.c:1000:9: originally declared here
arch/x86/kvm/x86.c:1079:7: warning: symbol 'index' shadows an earlier one
arch/x86/kvm/x86.c:1000:9: originally declared here
Make it static.
arch/x86/kvm/x86.c:1945:24: warning: symbol 'emulate_ops' was not declared. Should it be static?
Drop the return statements.
arch/x86/kvm/x86.c:2878:2: warning: returning void-valued expression
arch/x86/kvm/x86.c:2944:2: warning: returning void-valued expression
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Fixes sparse warning as well.
arch/x86/kvm/svm.c:69:15: warning: symbol 'iopm_base' was not declared. Should it be static?
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Nesting __emulate_2op_nobyte inside__emulate_2op produces many shadowed
variable warnings on the internal variable _tmp used by both macros.
Change the outer macro to use __tmp.
Avoids a sparse warning like the following at every call site of __emulate_2op
arch/x86/kvm/x86_emulate.c:1091:3: warning: symbol '_tmp' shadows an earlier one
arch/x86/kvm/x86_emulate.c:1091:3: originally declared here
[18 more warnings suppressed]
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Replaces open-coded mask calculation in macros.
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
This is the host part of kvm clocksource implementation. As it does
not include clockevents, it is a fairly simple implementation. We
only have to register a per-vcpu area, and start writing to it periodically.
The area is binary compatible with xen, as we use the same shadow_info
structure.
[marcelo: fix bad_page on MSR_KVM_SYSTEM_TIME]
[avi: save full value of the msr, even if enable bit is clear]
[avi: clear previous value of time_page]
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
This patch implements the Last Branch Record Virtualization (LBRV) feature of
the AMD Barcelona and Phenom processors into the kvm-amd module. It will only
be enabled if the guest enables last branch recording in the DEBUG_CTL MSR. So
there is no increased world switch overhead when the guest doesn't use these
MSRs.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Markus Rechberger <markus.rechberger@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
This patch changes the kvm-amd module to allocate the SVM MSR permission map
per VCPU instead of a global map for all VCPUs. With this we have more
flexibility allowing specific guests to access virtualized MSRs. This is
required for LBR virtualization.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Markus Rechberger <markus.rechberger@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Change the parameter of the init_vmcb() function in the kvm-amd module from
struct vmcb to struct vcpu_svm.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Markus Rechberger <markus.rechberger@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Looking at Intel Volume 3b, page 148, table 20-11 and noticed
that the field name is 'Deliver' not 'Deliever'. Attached patch changes
the define name and its user in vmx.c
Signed-off-by: Ryan Harper <ryanh@us.ibm.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
This patch contains the SVM architecture dependent changes for KVM to enable
support for the Nested Paging feature of AMD Barcelona and Phenom processors.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
This patch contains the changes to the KVM MMU necessary for support of the
Nested Paging feature in AMD Barcelona and Phenom Processors.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
The load_pdptrs() function is required in the SVM module for NPT support.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
The mapping function for the nonpaging case in the softmmu does basically the
same as required for Nested Paging. Make this function generic so it can be
used for both.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
The generic x86 code has to know if the specific implementation uses Nested
Paging. In the generic code Nested Paging is called Two Dimensional Paging
(TDP) to avoid confusion with (future) TDP implementations of other vendors.
This patch exports the availability of TDP to the generic x86 code.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
To disable the use of the Nested Paging feature even if it is available in
hardware this patch adds a module parameter. Nested Paging can be disabled by
passing npt=0 to the kvm_amd module.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Let SVM detect if the Nested Paging feature is available on the hardware.
Disable it to keep this patch series bisectable.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
By moving the SVM feature detection from the each_cpu code to the hardware
setup code it runs only once. As an additional advance the feature check is now
available earlier in the module setup process.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
This patch makes the EFER register accessible on a 32bit KVM host. This is
necessary to boot 32 bit PAE guests under SVM.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
To allow access to the EFER register in 32bit KVM the EFER specific code has to
be exported to the x86 generic code. This patch does this in a backwards
compatible manner.
[avi: add check for EFER-less hosts]
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
This patch aligns the bits the guest can set in the EFER register with the
features in the host processor. Currently it lets EFER.NX disabled if the
processor does not support it and enables EFER.LME and EFER.LMA only for KVM on
64 bit hosts.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
This patch give the SVM and VMX implementations the ability to add some bits
the guest can set in its EFER register.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
To allow TLB entries to be retained across VM entry and VM exit, the VMM
can now identify distinct address spaces through a new virtual-processor ID
(VPID) field of the VMCS.
[avi: drop vpid_sync_all()]
[avi: add "cc" to asm constraints]
Signed-off-by: Sheng Yang <sheng.yang@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Currently an mmio guest pte is encoded in the shadow pagetable as a
not-present trapping pte, with the SHADOW_IO_MARK bit set. However
nothing is ever done with this information, so maintaining it is a
useless complication.
This patch moves the check for mmio to before shadow ptes are instantiated,
so the shadow code is never invoked for ptes that reference mmio. The code
is simpler, and with future work, can be made to handle mmio concurrently.
Signed-off-by: Avi Kivity <avi@qumranet.com>
Certain x86 instructions use bits 3:5 of the byte following the opcode as an
opcode extension, with the decode sometimes depending on bits 6:7 as well.
Add support for this in the main decoding table rather than an ad-hock
adaptation per opcode.
Signed-off-by: Avi Kivity <avi@qumranet.com>
A guest partial guest pte write will leave shadow_trap_nonpresent_pte
in spte, which generates a vmexit at the next guest access through that pte.
This patch improves this by reading the full guest pte in advance and thus
being able to update the spte and eliminate the vmexit.
This helps pae guests which use two 32-bit writes to set a single 64-bit pte.
[truncation fix by Eric]
Signed-off-by: Yaozu (Eddie) Dong <eddie.dong@intel.com>
Signed-off-by: Feng (Eric) Liu <eric.e.liu@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
While backporting 72dc67a696, a gfn_to_page()
call was duplicated instead of moved (due to an unrelated patch not being
present in mainline). This caused a page reference leak, resulting in a
fairly massive memory leak.
Fix by removing the extraneous gfn_to_page() call.
Signed-off-by: Avi Kivity <avi@qumranet.com>
Do not assume that a shadow mapping will always point to the same host
frame number. Fixes crash with madvise(MADV_DONTNEED).
[avi: move after first printk(), add another printk()]
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
The vmx hardware state restore restores the tss selector and base address, but
not its length. Usually, this does not matter since most of the tss contents
is within the default length of 0x67. However, if a process is using ioperm()
to grant itself I/O port permissions, an additional bitmap within the tss,
but outside the default length is consulted. The effect is that the process
will receive a SIGSEGV instead of transparently accessing the port.
Fix by restoring the tss length. Note that i386 had this working already.
Closes bugzilla 10246.
Signed-off-by: Avi Kivity <avi@qumranet.com>
KVM tries to run as much as possible with the guest msrs loaded instead of
host msrs, since switching msrs is very expensive. It also tries to minimize
the number of msrs switched according to the guest mode; for example,
MSR_LSTAR is needed only by long mode guests. This optimization is done by
setup_msrs().
However, we must not change which msrs are switched while we are running with
guest msr state:
- switch to guest msr state
- call setup_msrs(), removing some msrs from the list
- switch to host msr state, leaving a few guest msrs loaded
An easy way to trigger this is to kexec an x86_64 linux guest. Early during
setup, the guest will switch EFER to not include SCE. KVM will stop saving
MSR_LSTAR, and on the next msr switch it will leave the guest LSTAR loaded.
The next host syscall will end up in a random location in the kernel.
Fix by reloading the host msrs before changing the msr list.
Signed-off-by: Avi Kivity <avi@qumranet.com>
For improved concurrency, the guest walk is performed concurrently with other
vcpus. This means that we need to revalidate the guest ptes once we have
write-protected the guest page tables, at which point they can no longer be
modified.
The current code attempts to avoid this check if the shadow page table is not
new, on the assumption that if it has existed before, the guest could not have
modified the pte without the shadow lock. However the assumption is incorrect,
as the racing vcpu could have modified the pte, then instantiated the shadow
page, before our vcpu regains control:
vcpu0 vcpu1
fault
walk pte
modify pte
fault in same pagetable
instantiate shadow page
lookup shadow page
conclude it is old
instantiate spte based on stale guest pte
We could do something clever with generation counters, but a test run by
Marcelo suggests this is unnecessary and we can just do the revalidation
unconditionally. The pte will be in the processor cache and the check can
be quite fast.
Signed-off-by: Avi Kivity <avi@qumranet.com>
If the local apic initial count is zero, don't start a an hrtimer with infinite
frequency, locking up the host.
Signed-off-by: Avi Kivity <avi@qumranet.com>
the cr3 variable is now inside the vcpu->arch structure.
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
alloc_apic_access_page() can sleep, while vmx_vcpu_setup is called
inside a non preemptable region. Move it after put_cpu().
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
While installing Windows XP 64 bit wants to access the DEBUGCTL and the last
branch record (LBR) MSRs. Don't allowing this in KVM causes the installation to
crash. This patch allow the access to these MSRs and fixes the issue.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Markus Rechberger <markus.rechberger@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
This patch replaces the mmap_sem lock for the memory slots with a new
kvm private lock, it is needed beacuse untill now there were cases where
kvm accesses user memory while holding the mmap semaphore.
Signed-off-by: Izik Eidus <izike@qumranet.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Injecting an GP when accessing this MSR lets Windows crash when running some
stress test tools in KVM. So this patch emulates access to this MSR.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Markus Rechberger <markus.rechberger@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
One of the use cases for the supported cpuid list is to create a "greatest
common denominator" of cpu capabilities in a server farm. As such, it is
useful to be able to get the list without creating a virtual machine first.
Since the code does not depend on the vm in any way, all that is needed is
to move it to the device ioctl handler. The capability identifier is also
changed so that binaries made against -rc1 will fail gracefully.
Signed-off-by: Avi Kivity <avi@qumranet.com>
Whilst working on getting a VM to initialize in to IA32e mode I found
this issue. set_cr0 relies on comparing the old cr0 to the new one to
work correctly. Move the assignment below so the compare can work.
Signed-off-by: Paul Knowles <paul@transitive.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Explicitly enable the NM intercept in svm_set_cr0 if we enable TS in the guest
copy of CR0 for lazy FPU switching. This fixes guest SMP with Linux under SVM.
Without that patch Linux deadlocks or panics right after trying to boot the
other CPUs.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Markus Rechberger <markus.rechberger@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
If the guest writes to cr0 and leaves the TS flag at 0 while vcpu->fpu_active
is also 0, the TS flag in the guest's cr0 gets lost. This leads to corrupt FPU
state an causes Windows Vista 64bit to crash very soon after boot. This patch
fixes this bug.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Markus Rechberger <markus.rechberger@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
arch/x86/kvm/x86.c: In function 'emulator_cmpxchg_emulated':
arch/x86/kvm/x86.c:1746: warning: passing argument 2 of 'vcpu->arch.mmu.gva_to_gpa' makes integer from pointer without a cast
arch/x86/kvm/x86.c:1746: warning: 'addr' is used uninitialized in this function
Is true. Local variable `addr' shadows incoming arg `addr'. Avi is on
vacation for a while, so...
Cc: Avi Kivity <avi@qumranet.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch moves virtio under the virtualization menu and changes virtio
devices to not claim to only be for lguest.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Migrating the apic timer in the critical section is not very nice, and is
absolutely horrible with the real-time port. Move migration to the regular
vcpu execution path, triggered by a new bitflag.
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Avi Kivity <avi@qumranet.com>
When preparing to enter the guest, if an interrupt comes in while
preemption is disabled but interrupts are still enabled, we miss a
preemption point. Fix by explicitly checking whether we need to
reschedule.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Otherwise we re-initialize the mmu caches, which will fail since the
caches are already registered, which will cause us to deinitialize said caches.
Signed-off-by: Avi Kivity <avi@qumranet.com>
Right now rmap_remove won't set the page as dirty if the shadow pte
pointed to this page had write access and then it became readonly.
This patches fixes that, by setting the page as dirty for spte changes from
write to readonly access.
Signed-off-by: Izik Eidus <izike@qumranet.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
When executing a test program called "crashme", we found the KVM guest cannot
survive more than ten seconds, then encounterd kernel panic. The basic concept
of "crashme" is generating random assembly code and trying to execute it.
After some fixes on emulator insn validity judgment, we found it's hard to
get the current emulator handle the invalid instructions correctly, for the
#UD trap for hypercall patching caused troubles. The problem is, if the opcode
itself was OK, but combination of opcode and modrm_reg was invalid, and one
operand of the opcode was memory (SrcMem or DstMem), the emulator will fetch
the memory operand first rather than checking the validity, and may encounter
an error there. For example, ".byte 0xfe, 0x34, 0xcd" has this problem.
In the patch, we simply check that if the invalid opcode wasn't vmcall/vmmcall,
then return from emulate_instruction() and inject a #UD to guest. With the
patch, the guest had been running for more than 12 hours.
Signed-off-by: Feng (Eric) Liu <eric.e.liu@intel.com>
Signed-off-by: Sheng Yang <sheng.yang@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
If some other cpu steals mmu pages between our check and an attempt to
allocate, we can run out of mmu pages. Fix by moving the check into the
same critical section as the allocation.
Signed-off-by: Avi Kivity <avi@qumranet.com>
Convert the synchronization of the shadow handling to a separate mmu_lock
spinlock.
Also guard fetch() by mmap_sem in read-mode to protect against alias
and memslot changes.
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Since gfn_to_page() is a sleeping function, and we want to make the core mmu
spinlocked, we need to pass the page from the walker context (which can sleep)
to the shadow context (which cannot).
[marcelo: avoid recursive locking of mmap_sem]
Signed-off-by: Avi Kivity <avi@qumranet.com>
In preparation for a mmu spinlock, add kvm_read_guest_atomic()
and use it in fetch() and prefetch_page().
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Do not hold kvm->lock mutex across the entire pagefault code,
only acquire it in places where it is necessary, such as mmu
hash list, active list, rmap and parent pte handling.
Allow concurrent guest walkers by switching walk_addr() to use
mmap_sem in read-mode.
And get rid of the lockless __gfn_to_page.
[avi: move kvm_mmu_pte_write() locking inside the function]
[avi: add locking for real mode]
[avi: fix cmpxchg locking]
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
This adds a mechanism for exposing the virtual apic tpr to the guest, and a
protocol for letting the guest update the tpr without causing a vmexit if
conditions allow (e.g. there is no interrupt pending with a higher priority
than the new tpr).
Signed-off-by: Avi Kivity <avi@qumranet.com>
Add a facility to report on accesses to the local apic tpr even if the
local apic is emulated in the kernel. This is basically a hack that
allows userspace to patch Windows which tends to bang on the tpr a lot.
Signed-off-by: Avi Kivity <avi@qumranet.com>
This can help diagnosing what the guest is trying to do. In many cases
we can get away with partial emulation of msrs.
Signed-off-by: Avi Kivity <avi@qumranet.com>
Host side TLB flush can be merged together if multiple
spte need to be write-protected.
Signed-off-by: Yaozu (Eddie) Dong <eddie.dong@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Moving kvm_vcpu_kick() to x86.c. Since it should be
common for all archs, put its declarations in <linux/kvm_host.h>
Signed-off-by: Zhang Xiantao <xiantao.zhang@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Move ioapic code to common, since IA64 also needs it.
Signed-off-by: Zhang Xiantao <xiantao.zhang@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
This paves the way for multiple architecture support. Note that while
ioapic.c could potentially be shared with ia64, it is also moved.
Signed-off-by: Avi Kivity <avi@qumranet.com>
