deliverable/linux.git
14 years agoKVM: x86 emulator: fix mov dr to inject #UD when needed.
Gleb Natapov [Thu, 18 Mar 2010 13:20:11 +0000 (15:20 +0200)] 
KVM: x86 emulator: fix mov dr to inject #UD when needed.

If CR4.DE=1 access to registers DR4/DR5 cause #UD.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: x86 emulator: inject #UD on access to non-existing CR
Gleb Natapov [Thu, 18 Mar 2010 13:20:10 +0000 (15:20 +0200)] 
KVM: x86 emulator: inject #UD on access to non-existing CR

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: x86 emulator: 0f (20|21|22|23) ignore mod bits.
Gleb Natapov [Thu, 18 Mar 2010 13:20:09 +0000 (15:20 +0200)] 
KVM: x86 emulator: 0f (20|21|22|23) ignore mod bits.

Resent spec says that for 0f (20|21|22|23) the 2 bits in the mod field
are ignored. Interestingly enough older spec says that 11 is only valid
encoding.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: x86 emulator: fix 0f 01 /5 emulation
Gleb Natapov [Thu, 18 Mar 2010 13:20:08 +0000 (15:20 +0200)] 
KVM: x86 emulator: fix 0f 01 /5 emulation

It is undefined and should generate #UD.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: x86 emulator: fix mov r/m, sreg emulation.
Gleb Natapov [Thu, 18 Mar 2010 13:20:07 +0000 (15:20 +0200)] 
KVM: x86 emulator: fix mov r/m, sreg emulation.

mov r/m, sreg generates #UD ins sreg is incorrect.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: Provide current eip as part of emulator context.
Gleb Natapov [Thu, 18 Mar 2010 13:20:06 +0000 (15:20 +0200)] 
KVM: Provide current eip as part of emulator context.

Eliminate the need to call back into KVM to get it from emulator.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: Provide x86_emulate_ctxt callback to get current cpl
Gleb Natapov [Thu, 18 Mar 2010 13:20:05 +0000 (15:20 +0200)] 
KVM: Provide x86_emulate_ctxt callback to get current cpl

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: remove realmode_lmsw function.
Gleb Natapov [Thu, 18 Mar 2010 13:20:04 +0000 (15:20 +0200)] 
KVM: remove realmode_lmsw function.

Use (get|set)_cr callback to emulate lmsw inside emulator.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: Provide callback to get/set control registers in emulator ops.
Gleb Natapov [Thu, 18 Mar 2010 13:20:03 +0000 (15:20 +0200)] 
KVM: Provide callback to get/set control registers in emulator ops.

Use this callback instead of directly call kvm function. Also rename
realmode_(set|get)_cr to emulator_(set|get)_cr since function has nothing
to do with real mode.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: coalesced_mmio: fix kvm_coalesced_mmio_init()'s error handling
Takuya Yoshikawa [Mon, 15 Mar 2010 13:13:30 +0000 (22:13 +0900)] 
KVM: coalesced_mmio: fix kvm_coalesced_mmio_init()'s error handling

kvm_coalesced_mmio_init() keeps to hold the addresses of a coalesced
mmio ring page and dev even after it has freed them.

Also, if this function fails, though it might be rare, it seems to be
suggesting the system's serious state: so we'd better stop the works
following the kvm_creat_vm().

This patch clears these problems.

  We move the coalesced mmio's initialization out of kvm_create_vm().
  This seems to be natural because it includes a registration which
  can be done only when vm is successfully created.

Signed-off-by: Takuya Yoshikawa <yoshikawa.takuya@oss.ntt.co.jp>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: VMX: change to use bool return values
Gui Jianfeng [Mon, 15 Mar 2010 09:29:09 +0000 (17:29 +0800)] 
KVM: VMX: change to use bool return values

Make use of bool as return values, and remove some useless
bool value converting. Thanks Avi to point this out.

Signed-off-by: Gui Jianfeng <guijianfeng@cn.fujitsu.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: Remove pointer to rflags from realmode_set_cr parameters.
Gleb Natapov [Mon, 15 Mar 2010 14:38:31 +0000 (16:38 +0200)] 
KVM: Remove pointer to rflags from realmode_set_cr parameters.

Mov reg, cr instruction doesn't change flags in any meaningful way, so
no need to update rflags after instruction execution.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: x86 emulator: check return value against correct define
Gleb Natapov [Mon, 15 Mar 2010 14:38:30 +0000 (16:38 +0200)] 
KVM: x86 emulator: check return value against correct define

Check return value against correct define instead of open code
the value.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: x86 emulator: fix RCX access during rep emulation
Gleb Natapov [Mon, 15 Mar 2010 14:38:29 +0000 (16:38 +0200)] 
KVM: x86 emulator: fix RCX access during rep emulation

During rep emulation access length to RCX depends on current address
mode.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: x86 emulator: Fix DstAcc decoding.
Gleb Natapov [Mon, 15 Mar 2010 14:38:28 +0000 (16:38 +0200)] 
KVM: x86 emulator: Fix DstAcc decoding.

Set correct operation length. Add RAX (64bit) handling.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: MMU: Reinstate pte prefetch on invlpg
Avi Kivity [Mon, 15 Mar 2010 11:59:57 +0000 (13:59 +0200)] 
KVM: MMU: Reinstate pte prefetch on invlpg

Commit fb341f57 removed the pte prefetch on guest invlpg, citing guest races.
However, the SDM is adamant that prefetch is allowed:

  "The processor may create entries in paging-structure caches for
   translations required for prefetches and for accesses that are a
   result of speculative execution that would never actually occur
   in the executed code path."

And, in fact, there was a race in the prefetch code: we picked up the pte
without the mmu lock held, so an older invlpg could install the pte over
a newer invlpg.

Reinstate the prefetch logic, but this time note whether another invlpg has
executed using a counter.  If a race occured, do not install the pte.

Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: MMU: Do not instantiate nontrapping spte on unsync page
Avi Kivity [Mon, 15 Mar 2010 11:59:56 +0000 (13:59 +0200)] 
KVM: MMU: Do not instantiate nontrapping spte on unsync page

The update_pte() path currently uses a nontrapping spte when a nonpresent
(or nonaccessed) gpte is written.  This is fine since at present it is only
used on sync pages.  However, on an unsync page this will cause an endless
fault loop as the guest is under no obligation to invlpg a gpte that
transitions from nonpresent to present.

Needed for the next patch which reinstates update_pte() on invlpg.

Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: Don't follow an atomic operation by a non-atomic one
Avi Kivity [Mon, 15 Mar 2010 11:59:55 +0000 (13:59 +0200)] 
KVM: Don't follow an atomic operation by a non-atomic one

Currently emulated atomic operations are immediately followed by a non-atomic
operation, so that kvm_mmu_pte_write() can be invoked.  This updates the mmu
but undoes the whole point of doing things atomically.

Fix by only performing the atomic operation and the mmu update, and avoiding
the non-atomic write.

Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: Make locked operations truly atomic
Avi Kivity [Mon, 15 Mar 2010 11:59:54 +0000 (13:59 +0200)] 
KVM: Make locked operations truly atomic

Once upon a time, locked operations were emulated while holding the mmu mutex.
Since mmu pages were write protected, it was safe to emulate the writes in
a non-atomic manner, since there could be no other writer, either in the
guest or in the kernel.

These days emulation takes place without holding the mmu spinlock, so the
write could be preempted by an unshadowing event, which exposes the page
to writes by the guest.  This may cause corruption of guest page tables.

Fix by using an atomic cmpxchg for these operations.

Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: MMU: Consolidate two guest pte reads in kvm_mmu_pte_write()
Avi Kivity [Mon, 15 Mar 2010 11:59:53 +0000 (13:59 +0200)] 
KVM: MMU: Consolidate two guest pte reads in kvm_mmu_pte_write()

kvm_mmu_pte_write() reads guest ptes in two different occasions, both to
allow a 32-bit pae guest to update a pte with 4-byte writes.  Consolidate
these into a single read, which also allows us to consolidate another read
from an invlpg speculating a gpte into the shadow page table.

Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: fix assigned_device_enable_host_msix error handling
jing zhang [Sat, 13 Mar 2010 07:00:45 +0000 (15:00 +0800)] 
KVM: fix assigned_device_enable_host_msix error handling

Free IRQ's and disable MSIX upon failure.

Cc: Avi Kivity <avi@redhat.com>
Signed-off-by: Jing Zhang <zj.barak@gmail.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: fix the errno of ioctl KVM_[UN]REGISTER_COALESCED_MMIO failure
Wei Yongjun [Fri, 12 Mar 2010 04:59:06 +0000 (12:59 +0800)] 
KVM: fix the errno of ioctl KVM_[UN]REGISTER_COALESCED_MMIO failure

This patch change the errno of ioctl KVM_[UN]REGISTER_COALESCED_MMIO
from -EINVAL to -ENXIO if no coalesced mmio dev exists.

Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: ia64: fix the error of ioctl KVM_IRQ_LINE if no irq chip
Wei Yongjun [Fri, 12 Mar 2010 02:11:15 +0000 (10:11 +0800)] 
KVM: ia64: fix the error of ioctl KVM_IRQ_LINE if no irq chip

If no irq chip in kernel, ioctl KVM_IRQ_LINE will return -EFAULT.
But I see in other place such as KVM_[GET|SET]IRQCHIP, -ENXIO is
return. So this patch used -ENXIO instead of -EFAULT.

Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: x86: fix the error of ioctl KVM_IRQ_LINE if no irq chip
Wei Yongjun [Fri, 12 Mar 2010 02:09:45 +0000 (10:09 +0800)] 
KVM: x86: fix the error of ioctl KVM_IRQ_LINE if no irq chip

If no irq chip in kernel, ioctl KVM_IRQ_LINE will return -EFAULT.
But I see in other place such as KVM_[GET|SET]IRQCHIP, -ENXIO is
return. So this patch used -ENXIO instead of -EFAULT.

Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: ia64: fix the error code of ioctl KVM_IA64_VCPU_GET_STACK failure
Wei Yongjun [Fri, 12 Mar 2010 00:45:39 +0000 (08:45 +0800)] 
KVM: ia64: fix the error code of ioctl KVM_IA64_VCPU_GET_STACK failure

The ioctl KVM_IA64_VCPU_GET_STACK does not set the error code if
copy_to_user() fail, and 0 will be return, we should use -EFAULT
instead of 0 in this case, so this patch fixed it.

Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: x86: Use native_store_idt() instead of kvm_get_idt()
Wei Yongjun [Fri, 5 Mar 2010 04:11:48 +0000 (12:11 +0800)] 
KVM: x86: Use native_store_idt() instead of kvm_get_idt()

This patch use generic linux function native_store_idt()
instead of kvm_get_idt(), and also removed the useless
function kvm_get_idt().

Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: Trace exception injection
Avi Kivity [Thu, 11 Mar 2010 11:01:59 +0000 (13:01 +0200)] 
KVM: Trace exception injection

Often an exception can help point out where things start to go wrong.

Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: Move kvm_exit tracepoint rip reading inside tracepoint
Avi Kivity [Thu, 11 Mar 2010 08:50:44 +0000 (10:50 +0200)] 
KVM: Move kvm_exit tracepoint rip reading inside tracepoint

Reading rip is expensive on vmx, so move it inside the tracepoint so we only
incur the cost if tracing is enabled.

Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: remove redundant initialization of page->private
Minchan Kim [Wed, 10 Mar 2010 14:31:22 +0000 (23:31 +0900)] 
KVM: remove redundant initialization of page->private

The prep_new_page() in page allocator calls set_page_private(page, 0).
So we don't need to reinitialize private of page.

Signed-off-by: Minchan Kim <minchan.kim@gmail.com>
Cc: Avi Kivity<avi@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: cleanup kvm trace
Xiao Guangrong [Wed, 10 Mar 2010 11:00:43 +0000 (19:00 +0800)] 
KVM: cleanup kvm trace

This patch does:

 - no need call tracepoint_synchronize_unregister() when kvm module
   is unloaded since ftrace can handle it

 - cleanup ftrace's macro

Signed-off-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Do not create debugfs if fail to create vcpu
Wei Yongjun [Tue, 9 Mar 2010 06:13:43 +0000 (14:13 +0800)] 
KVM: PPC: Do not create debugfs if fail to create vcpu

If fail to create the vcpu, we should not create the debugfs
for it.

Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Acked-by: Alexander Graf <agraf@suse.de>
Cc: stable@kernel.org
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: s390: Fix possible memory leak of in kvm_arch_vcpu_create()
Wei Yongjun [Tue, 9 Mar 2010 06:37:53 +0000 (14:37 +0800)] 
KVM: s390: Fix possible memory leak of in kvm_arch_vcpu_create()

This patch fixed possible memory leak in kvm_arch_vcpu_create()
under s390, which would happen when kvm_arch_vcpu_create() fails.

Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: stable@kernel.org
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: x86 emulator mark VMMCALL and LMSW as privileged
Gleb Natapov [Wed, 3 Mar 2010 15:53:05 +0000 (17:53 +0200)] 
KVM: x86 emulator mark VMMCALL and LMSW as privileged

LMSW is present in both group tables. It was marked privileged only in
one of them. Intel analog of VMMCALL is already marked privileged.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: SVM: Ignore lower 12 bit of nested msrpm_pa
Joerg Roedel [Mon, 1 Mar 2010 14:34:40 +0000 (15:34 +0100)] 
KVM: SVM: Ignore lower 12 bit of nested msrpm_pa

These bits are ignored by the hardware too. Implement this
for nested svm too.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM; SVM: Add correct handling of nested iopm
Joerg Roedel [Mon, 1 Mar 2010 14:34:39 +0000 (15:34 +0100)] 
KVM; SVM: Add correct handling of nested iopm

This patch adds the correct handling of the nested io
permission bitmap. Old behavior was to not lookup the port
in the iopm but only reinject an io intercept to the guest.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: SVM: Use svm_msrpm_offset in nested_svm_exit_handled_msr
Joerg Roedel [Mon, 1 Mar 2010 14:34:38 +0000 (15:34 +0100)] 
KVM: SVM: Use svm_msrpm_offset in nested_svm_exit_handled_msr

There is a generic function now to calculate msrpm offsets.
Use that function in nested_svm_exit_handled_msr() remove
the duplicate logic (which had a bug anyway).

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: SVM: Optimize nested svm msrpm merging
Joerg Roedel [Mon, 1 Mar 2010 14:34:37 +0000 (15:34 +0100)] 
KVM: SVM: Optimize nested svm msrpm merging

This patch optimizes the way the msrpm of the host and the
guest are merged. The old code merged the 2 msrpm pages
completly. This code needed to touch 24kb of memory for that
operation. The optimized variant this patch introduces
merges only the parts where the host msrpm may contain zero
bits. This reduces the amount of memory which is touched to
48 bytes.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: SVM: Introduce direct access msr list
Joerg Roedel [Mon, 1 Mar 2010 14:34:36 +0000 (15:34 +0100)] 
KVM: SVM: Introduce direct access msr list

This patch introduces a list with all msrs a guest might
have direct access to and changes the svm_vcpu_init_msrpm
function to use this list.
It also adds a check to set_msr_interception which triggers
a warning if a developer changes a msr intercept that is not
in the list.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: SVM: Move msrpm offset calculation to seperate function
Joerg Roedel [Mon, 1 Mar 2010 14:34:35 +0000 (15:34 +0100)] 
KVM: SVM: Move msrpm offset calculation to seperate function

The algorithm to find the offset in the msrpm for a given
msr is needed at other places too. Move that logic to its
own function.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: SVM: Return correct values in nested_svm_exit_handled_msr
Joerg Roedel [Mon, 1 Mar 2010 14:34:34 +0000 (15:34 +0100)] 
KVM: SVM: Return correct values in nested_svm_exit_handled_msr

The nested_svm_exit_handled_msr() returned an bool which is
a bug. I worked by accident because the exected integer
return values match with the true and false values. This
patch changes the return value to int and let the function
return the correct values.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: arch/x86/kvm/kvm_timer.h checkpatch cleanup
Andrea Gelmini [Sat, 27 Feb 2010 16:51:43 +0000 (17:51 +0100)] 
KVM: arch/x86/kvm/kvm_timer.h checkpatch cleanup

arch/x86/kvm/kvm_timer.h:13: ERROR: code indent should use tabs where possible

Signed-off-by: Andrea Gelmini <andrea.gelmini@gelma.net>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
14 years agoKVM: x86 emulator: Implement jmp far opcode ff/5
Gleb Natapov [Thu, 25 Feb 2010 14:36:43 +0000 (16:36 +0200)] 
KVM: x86 emulator: Implement jmp far opcode ff/5

Implement jmp far opcode ff/5. It is used by multiboot loader.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: x86 emulator: Add decoding of 16bit second in memory argument
Gleb Natapov [Thu, 25 Feb 2010 14:36:42 +0000 (16:36 +0200)] 
KVM: x86 emulator: Add decoding of 16bit second in memory argument

Add decoding of Ep type of argument used by callf/jmpf.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: move segment_base() into vmx.c
Gleb Natapov [Thu, 25 Feb 2010 10:43:09 +0000 (12:43 +0200)] 
KVM: move segment_base() into vmx.c

segment_base() is used only by vmx so move it there.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: fix segment_base() error checking
Gleb Natapov [Thu, 25 Feb 2010 10:43:08 +0000 (12:43 +0200)] 
KVM: fix segment_base() error checking

fix segment_base() to properly check for null segment selector and
avoid accessing NULL pointer if ldt selector in null.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: Drop kvm_get_gdt() in favor of generic linux function
Gleb Natapov [Thu, 25 Feb 2010 10:43:07 +0000 (12:43 +0200)] 
KVM: Drop kvm_get_gdt() in favor of generic linux function

Linux now has native_store_gdt() to do the same. Use it instead of
kvm local version.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: update gfn_to_hva() to use gfn_to_hva_memslot()
Takuya Yoshikawa [Thu, 25 Feb 2010 02:33:19 +0000 (11:33 +0900)] 
KVM: update gfn_to_hva() to use gfn_to_hva_memslot()

Marcelo introduced gfn_to_hva_memslot() when he implemented
gfn_to_pfn_memslot(). Let's use this for gfn_to_hva() too.

Note: also remove parentheses next to return as checkpatch said to do.

Signed-off-by: Takuya Yoshikawa <yoshikawa.takuya@oss.ntt.co.jp>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Clear exit_info for injected INTR exits
Joerg Roedel [Wed, 24 Feb 2010 17:59:19 +0000 (18:59 +0100)] 
KVM: SVM: Clear exit_info for injected INTR exits

When injecting an vmexit.intr into the nested hypervisor
there might be leftover values in the exit_info fields.
Clear them to not confuse nested hypervisors.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Handle nested selective_cr0 intercept correctly
Joerg Roedel [Wed, 24 Feb 2010 17:59:18 +0000 (18:59 +0100)] 
KVM: SVM: Handle nested selective_cr0 intercept correctly

If we have the following situation with nested svm:

1. Host KVM intercepts cr0 writes
2. Guest hypervisor intercepts only selective cr0 writes

Then we get an cr0 write intercept which is handled on the
host. But that intercepts may actually be a selective cr0
intercept for the guest. This patch checks for this
condition and injects a selective cr0 intercept if needed.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: x86: Don't set arch.cr0 in kvm_set_cr0
Joerg Roedel [Wed, 24 Feb 2010 17:59:17 +0000 (18:59 +0100)] 
KVM: x86: Don't set arch.cr0 in kvm_set_cr0

The vcpu->arch.cr0 variable is already set in the
architecture specific set_cr0 callbacks. There is no need to
set it in the common code.
This allows the architecture code to keep the old arch.cr0
value if it wants. This is required for nested svm to decide
if a selective_cr0 exit needs to be injected.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Ignore write of hwcr.ignne
Joerg Roedel [Wed, 24 Feb 2010 17:59:16 +0000 (18:59 +0100)] 
KVM: SVM: Ignore write of hwcr.ignne

Hyper-V as a guest wants to write this bit. This patch
ignores it.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Implement emulation of vm_cr msr
Joerg Roedel [Wed, 24 Feb 2010 17:59:15 +0000 (18:59 +0100)] 
KVM: SVM: Implement emulation of vm_cr msr

This patch implements the emulation of the vm_cr msr for
nested svm.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Add kvm_nested_intercepts tracepoint
Joerg Roedel [Wed, 24 Feb 2010 17:59:14 +0000 (18:59 +0100)] 
KVM: SVM: Add kvm_nested_intercepts tracepoint

This patch adds a tracepoint to get information about the
most important intercept bitmasks from the nested vmcb.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Restore tracing of nested vmcb address
Joerg Roedel [Wed, 24 Feb 2010 17:59:13 +0000 (18:59 +0100)] 
KVM: SVM: Restore tracing of nested vmcb address

A recent change broke tracing of the nested vmcb address. It
was reported as 0 all the time. This patch fixes it.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Check for nested intercepts on NMI injection
Joerg Roedel [Wed, 24 Feb 2010 17:59:12 +0000 (18:59 +0100)] 
KVM: SVM: Check for nested intercepts on NMI injection

This patch implements the NMI intercept checking for nested
svm.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Reset MMU on nested_svm_vmrun for NPT too
Joerg Roedel [Wed, 24 Feb 2010 17:59:11 +0000 (18:59 +0100)] 
KVM: SVM: Reset MMU on nested_svm_vmrun for NPT too

Without resetting the MMU the gva_to_pga function will not
work reliably when the vcpu is running in nested context.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Coding style cleanup
Joerg Roedel [Wed, 24 Feb 2010 17:59:10 +0000 (18:59 +0100)] 
KVM: SVM: Coding style cleanup

This patch removes whitespace errors, fixes comment formats
and most of checkpatch warnings. Now vim does not show
c-space-errors anymore.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: x86: Preserve injected TF across emulation
Jan Kiszka [Tue, 23 Feb 2010 16:47:59 +0000 (17:47 +0100)] 
KVM: x86: Preserve injected TF across emulation

Call directly into the vendor services for getting/setting rflags in
emulate_instruction to ensure injected TF survives the emulation.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: x86: Drop RF manipulation for guest single-stepping
Jan Kiszka [Tue, 23 Feb 2010 16:47:58 +0000 (17:47 +0100)] 
KVM: x86: Drop RF manipulation for guest single-stepping

RF is not required for injecting TF as the latter will trigger only
after an instruction execution anyway. So do not touch RF when arming or
disarming guest single-step mode.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Emulate nRIP feature when reinjecting INT3
Jan Kiszka [Tue, 23 Feb 2010 16:47:56 +0000 (17:47 +0100)] 
KVM: SVM: Emulate nRIP feature when reinjecting INT3

When in guest debugging mode, we have to reinject those #BP software
exceptions that are caused by guest-injected INT3. As older AMD
processors do not support the required nRIP VMCB field, try to emulate
it by moving RIP past the instruction on exception injection. Fix it up
again in case the injection failed and we were able to catch this. This
does not work for unintercepted faults, but it is better than doing
nothing.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: x86: Add kvm_is_linear_rip
Jan Kiszka [Tue, 23 Feb 2010 16:47:55 +0000 (17:47 +0100)] 
KVM: x86: Add kvm_is_linear_rip

Based on Gleb's suggestion: Add a helper kvm_is_linear_rip that matches
a given linear RIP against the current one. Use this for guest
single-stepping, more users will follow.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Move svm_queue_exception
Jan Kiszka [Tue, 23 Feb 2010 16:47:54 +0000 (17:47 +0100)] 
KVM: SVM: Move svm_queue_exception

Move svm_queue_exception past skip_emulated_instruction to allow calling
it later on.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: x86: Kick VCPU outside PIC lock again
Jan Kiszka [Wed, 24 Feb 2010 09:41:58 +0000 (10:41 +0100)] 
KVM: x86: Kick VCPU outside PIC lock again

This restores the deferred VCPU kicking before 956f97cf. We need this
over -rt as wake_up* requires non-atomic context in this configuration.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Destory timer on vcpu destruction
Alexander Graf [Mon, 22 Feb 2010 15:52:14 +0000 (16:52 +0100)] 
KVM: PPC: Destory timer on vcpu destruction

When we destory a vcpu, we should also make sure to kill all pending
timers that could still be up. When not doing this, hrtimers might
dereference null pointers trying to call our code.

This patch fixes spontanious kernel panics seen after closing VMs.

Signed-off-by: Alexander Graf <alex@csgraf.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Memset vcpu to zeros
Alexander Graf [Mon, 22 Feb 2010 15:52:08 +0000 (16:52 +0100)] 
KVM: PPC: Memset vcpu to zeros

While converting the kzalloc we used to allocate our vcpu struct to
vmalloc, I forgot to memset the contents to zeros. That broke quite
a lot.

This patch memsets it to zero again.

Signed-off-by: Alexander Graf <alex@csgraf.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: x86: Add support for saving&restoring debug registers
Jan Kiszka [Mon, 15 Feb 2010 09:45:43 +0000 (10:45 +0100)] 
KVM: x86: Add support for saving&restoring debug registers

So far user space was not able to save and restore debug registers for
migration or after reset. Plug this hole.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: x86: Save&restore interrupt shadow mask
Jan Kiszka [Fri, 19 Feb 2010 18:38:07 +0000 (19:38 +0100)] 
KVM: x86: Save&restore interrupt shadow mask

The interrupt shadow created by STI or MOV-SS-like operations is part of
the VCPU state and must be preserved across migration. Transfer it in
the spare padding field of kvm_vcpu_events.interrupt.

As a side effect we now have to make vmx_set_interrupt_shadow robust
against both shadow types being set. Give MOV SS a higher priority and
skip STI in that case to avoid that VMX throws a fault on next entry.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: x86: Do not return soft events in vcpu_events
Jan Kiszka [Mon, 15 Feb 2010 09:45:41 +0000 (10:45 +0100)] 
KVM: x86: Do not return soft events in vcpu_events

To avoid that user space migrates a pending software exception or
interrupt, mask them out on KVM_GET_VCPU_EVENTS. Without this, user
space would try to reinject them, and we would have to reconstruct the
proper instruction length for VMX event injection. Now the pending event
will be reinjected via executing the triggering instruction again.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Fix wrong interrupt injection in enable_irq_windows
Joerg Roedel [Fri, 19 Feb 2010 15:23:01 +0000 (16:23 +0100)] 
KVM: SVM: Fix wrong interrupt injection in enable_irq_windows

The nested_svm_intr() function does not execute the vmexit
anymore. Therefore we may still be in the nested state after
that function ran. This patch changes the nested_svm_intr()
function to return wether the irq window could be enabled.

Cc: stable@kernel.org
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: drop unneeded kvm_run check in emulate_instruction()
Gleb Natapov [Sun, 21 Feb 2010 13:00:47 +0000 (15:00 +0200)] 
KVM: drop unneeded kvm_run check in emulate_instruction()

vcpu->run is initialized on vcpu creation and can never be NULL
here.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Allocate vcpu struct using vmalloc
Alexander Graf [Fri, 19 Feb 2010 11:24:33 +0000 (12:24 +0100)] 
KVM: PPC: Allocate vcpu struct using vmalloc

We used to use get_free_pages to allocate our vcpu struct. Unfortunately
that call failed on me several times after my machine had a big enough
uptime, as memory became too fragmented by then.

Fortunately, we don't need it to be page aligned any more! We can just
vmalloc it and everything's great.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Simplify kvmppc_load_up_(FPU|VMX|VSX)
Alexander Graf [Fri, 19 Feb 2010 10:00:47 +0000 (11:00 +0100)] 
KVM: PPC: Simplify kvmppc_load_up_(FPU|VMX|VSX)

We don't need as complex code. I had some thinkos while writing it, figuring
I needed to support PPC32 paths on PPC64 which would have required DR=0, but
everything just runs fine with DR=1.

So let's make the functions simple C call wrappers that reserve some space on
the stack for the respective functions to clobber.

Fixes out-of-RMA-access (and thus guest FPU loading) on the PS3.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Enable use of secondary htab bucket
Alexander Graf [Fri, 19 Feb 2010 10:00:46 +0000 (11:00 +0100)] 
KVM: PPC: Enable use of secondary htab bucket

We had code to make use of the secondary htab buckets, but kept that
disabled because it was unstable when I put it in.

I checked again if that's still the case and apparently it was only
exposing some instability that was there anyways before. I haven't
seen any badness related to usage of secondary htab entries so far.

This should speed up guest memory allocations by quite a bit, because
we now have more space to put PTEs in.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Add capability for paired singles
Alexander Graf [Fri, 19 Feb 2010 10:00:45 +0000 (11:00 +0100)] 
KVM: PPC: Add capability for paired singles

We need to tell userspace that we can emulate paired single instructions.
So let's add a capability export.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Implement Paired Single emulation
Alexander Graf [Fri, 19 Feb 2010 10:00:44 +0000 (11:00 +0100)] 
KVM: PPC: Implement Paired Single emulation

The one big thing about the Gekko is paired singles.

Paired singles are an extension to the instruction set, that adds 32 single
precision floating point registers (qprs), some SPRs to modify the behavior
of paired singled operations and instructions to deal with qprs to the
instruction set.

Unfortunately, it also changes semantics of existing operations that affect
single values in FPRs. In most cases they get mirrored to the coresponding
QPR.

Thanks to that we need to emulate all FPU operations and all the new paired
single operations too.

In order to achieve that, we use the just introduced FPU call helpers to
call the real FPU whenever the guest wants to modify an FPR. Additionally
we also fix up the QPR values along the way.

That way we can execute paired single FPU operations without implementing a
soft fpu.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Enable program interrupt to do MMIO
Alexander Graf [Fri, 19 Feb 2010 10:00:43 +0000 (11:00 +0100)] 
KVM: PPC: Enable program interrupt to do MMIO

When we get a program interrupt we usually don't expect it to perform an
MMIO operation. But why not? When we emulate paired singles, we can end
up loading or storing to an MMIO address - and the handling of those
happens in the program interrupt handler.

So let's teach the program interrupt handler how to deal with EMULATE_MMIO.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Add helpers to modify ppc fields
Alexander Graf [Fri, 19 Feb 2010 10:00:42 +0000 (11:00 +0100)] 
KVM: PPC: Add helpers to modify ppc fields

The PowerPC specification always lists bits from MSB to LSB. That is
really confusing when you're trying to write C code, because it fits
in pretty badly with the normal (1 << xx) schemes.

So I came up with some nice wrappers that allow to get and set fields
in a u64 with bit numbers exactly as given in the spec. That makes the
code in KVM and the spec easier comparable.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Fix error in BAT assignment
Alexander Graf [Fri, 19 Feb 2010 10:00:41 +0000 (11:00 +0100)] 
KVM: PPC: Fix error in BAT assignment

BATs didn't work. Well, they did, but only up to BAT3. As soon as we
came to BAT4 the offset calculation was screwed up and we ended up
overwriting BAT0-3.

Fortunately, Linux hasn't been using BAT4+. It's still a good
idea to write correct code though.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Add helpers to call FPU instructions
Alexander Graf [Fri, 19 Feb 2010 10:00:40 +0000 (11:00 +0100)] 
KVM: PPC: Add helpers to call FPU instructions

To emulate paired single instructions, we need to be able to call FPU
operations from within the kernel. Since we don't want gcc to spill
arbitrary FPU code everywhere, we tell it to use a soft fpu.

Since we know we can really call the FPU in safe areas, let's also add
some calls that we can later use to actually execute real world FPU
operations on the host's FPU.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Make ext giveup non-static
Alexander Graf [Fri, 19 Feb 2010 10:00:39 +0000 (11:00 +0100)] 
KVM: PPC: Make ext giveup non-static

We need to call the ext giveup handlers from code outside of book3s.c.
So let's make it non-static.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Make software load/store return eaddr
Alexander Graf [Fri, 19 Feb 2010 10:00:38 +0000 (11:00 +0100)] 
KVM: PPC: Make software load/store return eaddr

The Book3S KVM implementation contains some helper functions to load and store
data from and to virtual addresses.

Unfortunately, this helper used to keep the physical address it so nicely
found out for us to itself. So let's change that and make it return the
physical address it resolved.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Implement mtsr instruction emulation
Alexander Graf [Fri, 19 Feb 2010 10:00:37 +0000 (11:00 +0100)] 
KVM: PPC: Implement mtsr instruction emulation

The Book3S_32 specifications allows for two instructions to modify segment
registers: mtsrin and mtsr.

Most normal operating systems use mtsrin, because it allows to define which
segment it wants to change using a register. But since I was trying to run
an embedded guest, it turned out to be using mtsr with hardcoded values.

So let's also emulate mtsr. It's a valid instruction after all.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Fix typo in book3s_32 debug code
Alexander Graf [Fri, 19 Feb 2010 10:00:36 +0000 (11:00 +0100)] 
KVM: PPC: Fix typo in book3s_32 debug code

There's a typo in the debug ifdef of the book3s_32 mmu emulation. While trying
to debug something I stumbled across that and wanted to save anyone after me
(or myself later) from having to debug that again.

So let's fix the ifdef.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Preload FPU when possible
Alexander Graf [Fri, 19 Feb 2010 10:00:35 +0000 (11:00 +0100)] 
KVM: PPC: Preload FPU when possible

There are some situations when we're pretty sure the guest will use the
FPU soon. So we can save the churn of going into the guest, finding out
it does want to use the FPU and going out again.

This patch adds preloading of the FPU when it's reasonable.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Combine extension interrupt handlers
Alexander Graf [Fri, 19 Feb 2010 10:00:34 +0000 (11:00 +0100)] 
KVM: PPC: Combine extension interrupt handlers

When we for example get an Altivec interrupt, but our guest doesn't support
altivec, we need to inject a program interrupt, not an altivec interrupt.

The same goes for paired singles. When an altivec interrupt arrives, we're
pretty sure we need to emulate the instruction because it's a paired single
operation.

So let's make all the ext handlers aware that they need to jump to the
program interrupt handler when an extension interrupt arrives that
was not supposed to arrive for the guest CPU.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Add Gekko SPRs
Alexander Graf [Fri, 19 Feb 2010 10:00:33 +0000 (11:00 +0100)] 
KVM: PPC: Add Gekko SPRs

The Gekko has some SPR values that differ from other PPC core values and
also some additional ones.

Let's add support for them in our mfspr/mtspr emulator.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Add hidden flag for paired singles
Alexander Graf [Fri, 19 Feb 2010 10:00:32 +0000 (11:00 +0100)] 
KVM: PPC: Add hidden flag for paired singles

The Gekko implements an extension called paired singles. When the guest wants
to use that extension, we need to make sure we're not running the host FPU,
because all FPU instructions need to get emulated to accomodate for additional
operations that occur.

This patch adds an hflag to track if we're in paired single mode or not.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Add AGAIN type for emulation return
Alexander Graf [Fri, 19 Feb 2010 10:00:31 +0000 (11:00 +0100)] 
KVM: PPC: Add AGAIN type for emulation return

Emulation of an instruction can have different outcomes. It can succeed,
fail, require MMIO, do funky BookE stuff - or it can just realize something's
odd and will be fixed the next time around.

Exactly that is what EMULATE_AGAIN means. Using that flag we can now tell
the caller that nothing happened, but we still want to go back to the
guest and see what happens next time we come around.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Teach MMIO Signedness
Alexander Graf [Fri, 19 Feb 2010 10:00:30 +0000 (11:00 +0100)] 
KVM: PPC: Teach MMIO Signedness

The guest I was trying to get to run uses the LHA and LHAU instructions.
Those instructions basically do a load, but also sign extend the result.

Since we need to fill our registers by hand when doing MMIO, we also need
to sign extend manually.

This patch implements sign extended MMIO and the LHA(U) instructions.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Enable MMIO to do 64 bits, fprs and qprs
Alexander Graf [Fri, 19 Feb 2010 10:00:29 +0000 (11:00 +0100)] 
KVM: PPC: Enable MMIO to do 64 bits, fprs and qprs

Right now MMIO access can only happen for GPRs and is at most 32 bit wide.
That's actually enough for almost all types of hardware out there.

Unfortunately, the guest I was using used FPU writes to MMIO regions, so
it ended up writing 64 bit MMIOs using FPRs and QPRs.

So let's add code to handle those odd cases too.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Make fpscr 64-bit
Alexander Graf [Fri, 19 Feb 2010 10:00:28 +0000 (11:00 +0100)] 
KVM: PPC: Make fpscr 64-bit

Modern PowerPCs have a 64 bit wide FPSCR register. Let's accomodate for that
and make it 64 bits in our vcpu struct too.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: PPC: Add QPR registers
Alexander Graf [Fri, 19 Feb 2010 10:00:27 +0000 (11:00 +0100)] 
KVM: PPC: Add QPR registers

The Gekko has GPRs, SPRs and FPRs like normal PowerPC codes, but
it also has QPRs which are basically single precision only FPU registers
that get used when in paired single mode.

The following patches depend on them being around, so let's add the
definitions early.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Remove newlines from nested trace points
Joerg Roedel [Fri, 19 Feb 2010 15:23:09 +0000 (16:23 +0100)] 
KVM: SVM: Remove newlines from nested trace points

The tracing infrastructure adds its own newlines. Remove
them from the trace point printk format strings.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Make lazy FPU switching work with nested svm
Joerg Roedel [Fri, 19 Feb 2010 15:23:08 +0000 (16:23 +0100)] 
KVM: SVM: Make lazy FPU switching work with nested svm

The new lazy fpu switching code may disable cr0 intercepts
when running nested. This is a bug because the nested
hypervisor may still want to intercept cr0 which will break
in this situation. This patch fixes this issue and makes
lazy fpu switching working with nested svm.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Activate nested state only when guest state is complete
Joerg Roedel [Fri, 19 Feb 2010 15:23:07 +0000 (16:23 +0100)] 
KVM: SVM: Activate nested state only when guest state is complete

Certain functions called during the emulated world switch
behave differently when the vcpu is running nested. This is
not the expected behavior during a world switch emulation.
This patch ensures that the nested state is activated only
if the vcpu is completly in nested state.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Don't sync nested cr8 to lapic and back
Joerg Roedel [Fri, 19 Feb 2010 15:23:06 +0000 (16:23 +0100)] 
KVM: SVM: Don't sync nested cr8 to lapic and back

This patch makes syncing of the guest tpr to the lapic
conditional on !nested. Otherwise a nested guest using the
TPR could freeze the guest.
Another important change this patch introduces is that the
cr8 intercept bits are no longer ORed at vmrun emulation if
the guest sets VINTR_MASKING in its VMCB. The reason is that
nested cr8 accesses need alway be handled by the nested
hypervisor because they change the shadow version of the
tpr.

Cc: stable@kernel.org
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Fix nested msr intercept handling
Joerg Roedel [Fri, 19 Feb 2010 15:23:05 +0000 (16:23 +0100)] 
KVM: SVM: Fix nested msr intercept handling

The nested_svm_exit_handled_msr() function maps only one
page of the guests msr permission bitmap. This patch changes
the code to use kvm_read_guest to fix the bug.

Cc: stable@kernel.org
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Annotate nested_svm_map with might_sleep()
Joerg Roedel [Fri, 19 Feb 2010 15:23:04 +0000 (16:23 +0100)] 
KVM: SVM: Annotate nested_svm_map with might_sleep()

The nested_svm_map() function can sleep and must not be
called from atomic context. So annotate that function.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Sync all control registers on nested vmexit
Joerg Roedel [Fri, 19 Feb 2010 15:23:03 +0000 (16:23 +0100)] 
KVM: SVM: Sync all control registers on nested vmexit

Currently the vmexit emulation does not sync control
registers were the access is typically intercepted by the
nested hypervisor. But we can not count on that intercepts
to sync these registers too and make the code
architecturally more correct.

Cc: stable@kernel.org
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
14 years agoKVM: SVM: Fix schedule-while-atomic on nested exception handling
Joerg Roedel [Fri, 19 Feb 2010 15:23:02 +0000 (16:23 +0100)] 
KVM: SVM: Fix schedule-while-atomic on nested exception handling

Move the actual vmexit routine out of code that runs with
irqs and preemption disabled.

Cc: stable@kernel.org
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
This page took 0.043993 seconds and 5 git commands to generate.