Avi Kivity [Wed, 21 Nov 2007 00:06:21 +0000 (02:06 +0200)]
KVM: MMU: Avoid unnecessary remote tlb flushes when guest updates a pte
If all we're doing is increasing permissions on a pte (typical for demand
paging), then there's not need to flush remote tlbs. Worst case they'll
get a spurious page fault.
Avi Kivity [Tue, 20 Nov 2007 19:39:54 +0000 (21:39 +0200)]
KVM: MMU: Implement guest page fault bypass for nonpae
I spent an hour worrying why I see so many guest page faults on FC6 i386.
Turns out bypass wasn't implemented for nonpae. Implement it so it doesn't
happen again.
Avi Kivity [Tue, 20 Nov 2007 13:30:24 +0000 (15:30 +0200)]
KVM: Split vcpu creation to avoid vcpu_load() before preemption setup
Split kvm_arch_vcpu_create() into kvm_arch_vcpu_create() and
kvm_arch_vcpu_setup(), enabling preemption notification between the two.
This mean that we can now do vcpu_load() within kvm_arch_vcpu_setup().
Avi Kivity [Tue, 20 Nov 2007 10:49:31 +0000 (12:49 +0200)]
KVM: x86 emulator: retire ->write_std()
Theoretically used to acccess memory known to be ordinary RAM, it was
never implemented. It is questionable whether it is possible to implement
it correctly.
Izik Eidus [Tue, 20 Nov 2007 09:49:33 +0000 (11:49 +0200)]
KVM: MMU: Selectively set PageDirty when releasing guest memory
Improve dirty bit setting for pages that kvm release, until now every page
that we released we marked dirty, from now only pages that have potential
to get dirty we mark dirty.
Signed-off-by: Izik Eidus <izike@qumranet.com> Signed-off-by: Avi Kivity <avi@qumranet.com>
Izik Eidus [Tue, 20 Nov 2007 09:30:04 +0000 (11:30 +0200)]
KVM: MMU: Fix potential memory leak with smp real-mode
When we map a page, we check whether some other vcpu mapped it for us and if
so, bail out. But we should decrease the refcount on the page as we do so.
Signed-off-by: Izik Eidus <izike@qumranet.com> Signed-off-by: Avi Kivity <avi@qumranet.com>
Jan Kiszka [Mon, 19 Nov 2007 09:21:45 +0000 (10:21 +0100)]
KVM: VMX: Force seg.base == (seg.sel << 4) in real mode
Ensure that segment.base == segment.selector << 4 when entering the real
mode on Intel so that the CPU will not bark at us. This fixes some old
protected mode demo from http://www.x86.org/articles/pmbasics/tspec_a1_doc.htm.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com> Signed-off-by: Avi Kivity <avi@qumranet.com>
Avi Kivity [Sun, 18 Nov 2007 11:50:24 +0000 (13:50 +0200)]
KVM: Replace 'light_exits' stat with 'host_state_reload'
This is a little more accurate (since it counts actual reloads, not potential
reloads), and reverses the sense of the statistic to measure a bad event like
most of the other stats (e.g. we want to minimize all counters).
Avi Kivity [Thu, 15 Nov 2007 16:06:18 +0000 (18:06 +0200)]
KVM: VMX: Consolidate register usage in vmx_vcpu_run()
We pass vcpu, vmx->fail, and vmx->launched to assembly code, but all three
are fields within vmx. Consolidate by only passing in vmx and offsets for
the rest.
Izik Eidus [Sun, 11 Nov 2007 20:10:22 +0000 (22:10 +0200)]
KVM: Change kvm_{read,write}_guest() to use copy_{from,to}_user()
This changes kvm_write_guest_page/kvm_read_guest_page to use
copy_to_user/read_from_user, as a result we get better speed
and better dirty bit tracking.
Signed-off-by: Izik Eidus <izike@qumranet.com> Signed-off-by: Avi Kivity <avi@qumranet.com>
Avi Kivity [Thu, 22 Nov 2007 09:42:59 +0000 (11:42 +0200)]
KVM: Fix faults during injection of real-mode interrupts
If vmx fails to inject a real-mode interrupt while fetching the interrupt
redirection table, it fails to record this in the vectoring information
field. So we detect this condition and do it ourselves.
Avi Kivity [Thu, 8 Nov 2007 16:19:20 +0000 (18:19 +0200)]
KVM: VMX: Use vmx to inject real-mode interrupts
Instead of injecting real-mode interrupts by writing the interrupt frame into
guest memory, abuse vmx by injecting a software interrupt. We need to
pretend the software interrupt instruction had a length > 0, so we have to
adjust rip backward.
This lets us not to mess with writing guest memory, which is complex and also
sleeps.
Avi Kivity [Wed, 31 Oct 2007 09:15:56 +0000 (11:15 +0200)]
KVM: x86 emulator: centralize decoding of one-byte register access insns
Instructions like 'inc reg' that have the register operand encoded
in the opcode are currently specially decoded. Extend
decode_register_operand() to handle that case, indicated by having
DstReg or SrcReg without ModRM.
Carsten Otte [Tue, 30 Oct 2007 17:44:25 +0000 (18:44 +0100)]
KVM: Portability: Move pio emulation functions to x86.c
This patch moves implementation of the following functions from
kvm_main.c to x86.c:
free_pio_guest_pages, vcpu_find_pio_dev, pio_copy_data, complete_pio,
kernel_pio, pio_string_write, kvm_emulate_pio, kvm_emulate_pio_string
The function inject_gp, which was duplicated by yesterday's patch
series, is removed from kvm_main.c now because it is not needed anymore.
Carsten Otte [Tue, 30 Oct 2007 17:44:21 +0000 (18:44 +0100)]
KVM: Portability: Move x86 emulation and mmio device hook to x86.c
This patch moves the following functions to from kvm_main.c to x86.c:
emulator_read/write_std, vcpu_find_pervcpu_dev, vcpu_find_mmio_dev,
emulator_read/write_emulated, emulator_write_phys,
emulator_write_emulated_onepage, emulator_cmpxchg_emulated,
get_setment_base, emulate_invlpg, emulate_clts, emulator_get/set_dr,
kvm_report_emulation_failure, emulate_instruction
The following data type is moved to x86.c:
struct x86_emulate_ops emulate_ops
Carsten Otte [Tue, 30 Oct 2007 17:44:17 +0000 (18:44 +0100)]
KVM: Portability: Move kvm_get/set_msr[_common] to x86.c
This patch moves the implementation of the functions of kvm_get/set_msr,
kvm_get/set_msr_common, and set_efer from kvm_main.c to x86.c. The
definition of EFER_RESERVED_BITS is moved too.
Anthony Liguori [Mon, 29 Oct 2007 20:15:20 +0000 (15:15 -0500)]
KVM: Fix gfn_to_page() acquiring mmap_sem twice
KVM's nopage handler calls gfn_to_page() which acquires the mmap_sem when
calling out to get_user_pages(). nopage handlers are already invoked with the
mmap_sem held though. Introduce a __gfn_to_page() for use by the nopage
handler which requires the lock to already be held.
This was noticed by tglx.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> Signed-off-by: Avi Kivity <avi@qumranet.com>
This patch based on CR8/TPR patch, and enable the TPR shadow (FlexPriority)
for 32bit Windows. Since TPR is accessed very frequently by 32bit
Windows, especially SMP guest, with FlexPriority enabled, we saw significant
performance gain.
Signed-off-by: Sheng Yang <sheng.yang@intel.com> Signed-off-by: Avi Kivity <avi@qumranet.com>
Carsten Otte [Mon, 29 Oct 2007 15:09:35 +0000 (16:09 +0100)]
KVM: Portability: Move control register helper functions to x86.c
This patch moves the definitions of CR0_RESERVED_BITS,
CR4_RESERVED_BITS, and CR8_RESERVED_BITS along with the following
functions from kvm_main.c to x86.c:
set_cr0(), set_cr3(), set_cr4(), set_cr8(), get_cr8(), lmsw(),
load_pdptrs()
The static function wrapper inject_gp is duplicated in kvm_main.c and
x86.c for now, the version in kvm_main.c should disappear once the last
user of it is gone too.
The function load_pdptrs is no longer static, and now defined in x86.h
for the time being, until the last user of it is gone from kvm_main.c.
Signed-off-by: Carsten Otte <cotte@de.ibm.com> Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com> Acked-by: Hollis Blanchard <hollisb@us.ibm.com> Signed-off-by: Avi Kivity <avi@qumranet.com>
Carsten Otte [Mon, 29 Oct 2007 15:08:51 +0000 (16:08 +0100)]
KVM: Portability: Move memory segmentation to x86.c
This patch moves the definition of segment_descriptor_64 for AMD64 and
EM64T from kvm_main.c to segment_descriptor.h. It also adds a proper
#ifndef...#define...#endif around that header file.
The implementation of segment_base is moved from kvm_main.c to x86.c.
Signed-off-by: Carsten Otte <cotte@de.ibm.com> Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com> Acked-by: Hollis Blanchard <hollisb@us.ibm.com> Signed-off-by: Avi Kivity <avi@qumranet.com>
Carsten Otte [Mon, 29 Oct 2007 15:08:35 +0000 (16:08 +0100)]
KVM: Portability: Split kvm_vm_ioctl v3
This patch splits kvm_vm_ioctl into archtecture independent parts, and
x86 specific parts which go to kvm_arch_vcpu_ioctl in x86.c.
The patch is unchanged since last submission.
Common ioctls for all architectures are:
KVM_CREATE_VCPU, KVM_GET_DIRTY_LOG, KVM_SET_USER_MEMORY_REGION
x86 specific ioctls are:
KVM_SET_MEMORY_REGION,
KVM_GET/SET_NR_MMU_PAGES, KVM_SET_MEMORY_ALIAS, KVM_CREATE_IRQCHIP,
KVM_CREATE_IRQ_LINE, KVM_GET/SET_IRQCHIP
KVM_SET_TSS_ADDR
Signed-off-by: Carsten Otte <cotte@de.ibm.com> Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com> Acked-by: Hollis Blanchard <hollisb@us.ibm.com> Signed-off-by: Avi Kivity <avi@qumranet.com>
Avi Kivity [Sun, 28 Oct 2007 14:34:25 +0000 (16:34 +0200)]
KVM: x86 emulator: don't depend on cr2 for mov abs emulation
The 'mov abs' instruction family (opcodes 0xa0 - 0xa3) still depends on cr2
provided by the page fault handler. This is wrong for several reasons:
- if an instruction accessed misaligned data that crosses a page boundary,
and if the fault happened on the second page, cr2 will point at the
second page, not the data itself.
- if we're emulating in real mode, or due to a FlexPriority exit, there
is no cr2 generated.
So, this change adds decoding for this instruction form and drops reliance
on cr2.
Laurent Vivier [Thu, 25 Oct 2007 12:18:55 +0000 (14:18 +0200)]
KVM: VMX: Let gcc to choose which registers to save (i386)
This patch lets GCC to determine which registers to save when we
switch to/from a VCPU in the case of intel i386.
* Original code saves following registers:
eax, ebx, ecx, edx, edi, esi, ebp (using popa)
* Patched code:
- informs GCC that we modify following registers
using the clobber description:
ebx, edi, rsi
- doesn't save eax because it is an output operand (vmx->fail)
- cannot put ecx in clobber description because it is an input operand,
but as we modify it and we want to keep its value (vcpu), we must
save it (pop/push)
- ebp is saved (pop/push) because GCC seems to ignore its use the clobber
description.
- edx is saved (pop/push) because it is reserved by GCC (REGPARM) and
cannot be put in the clobber description.
- line "mov (%%esp), %3 \n\t" has been removed because %3
is ecx and ecx is restored just after.
Signed-off-by: Laurent Vivier <Laurent.Vivier@bull.net> Signed-off-by: Avi Kivity <avi@qumranet.com>
- doesn't save rax because it is an output operand (vmx->fail)
- cannot put rcx in clobber description because it is an input operand,
but as we modify it and we want to keep its value (vcpu), we must
save it (pop/push)
- rbp is saved (pop/push) because GCC seems to ignore its use in the clobber
description.
- rdx is saved (pop/push) because it is reserved by GCC (REGPARM) and
cannot be put in the clobber description.
- line "mov (%%rsp), %3 \n\t" has been removed because %3
is rcx and rcx is restored just after.
- line ASM_VMX_VMWRITE_RSP_RDX() is moved out of the ifdef/else/endif
Signed-off-by: Laurent Vivier <Laurent.Vivier@bull.net> Signed-off-by: Avi Kivity <avi@qumranet.com>
Izik Eidus [Wed, 24 Oct 2007 22:29:55 +0000 (00:29 +0200)]
KVM: Add ioctl to tss address from userspace,
Currently kvm has a wart in that it requires three extra pages for use
as a tss when emulating real mode on Intel. This patch moves the allocation
internally, only requiring userspace to tell us where in the physical address
space we can place the tss.
Signed-off-by: Izik Eidus <izike@qumranet.com> Signed-off-by: Avi Kivity <avi@qumranet.com>
Izik Eidus [Wed, 24 Oct 2007 21:57:46 +0000 (23:57 +0200)]
KVM: Add kernel-internal memory slots
Reserve a few memory slots for kernel internal use. This is good for case
you have to register memory region and you want to be sure it was not
registered from userspace, and for case you want to register a memory region
that won't be seen from userspace.
Signed-off-by: Izik Eidus <izike@qumranet.com> Signed-off-by: Avi Kivity <avi@qumranet.com>
Izik Eidus [Thu, 25 Oct 2007 09:54:04 +0000 (11:54 +0200)]
KVM: Unmap kernel-allocated memory on slot destruction
kvm_vm_ioctl_set_memory_region() is able to remove memory in addition to
adding it. Therefore when using kernel swapping support for old userspaces,
we need to munmap the memory if the user request to remove it
Signed-off-by: Izik Eidus <izike@qumranet.com> Signed-off-by: Avi Kivity <avi@qumranet.com>
Currently kvm provides hypercalls only for x86* architectures. To
provide hypercall infrastructure for other kvm architectures I split
kvm_para.h into a generic header file and architecture specific
definitions.
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com> Signed-off-by: Avi Kivity <avi@qumranet.com>
Avi Kivity [Sun, 21 Oct 2007 09:00:39 +0000 (11:00 +0200)]
KVM: Move vmx_vcpu_reset() out of vmx_vcpu_setup()
Split guest reset code out of vmx_vcpu_setup(). Besides being cleaner, this
moves the realmode tss setup (which can sleep) outside vmx_vcpu_setup()
(which is executed with preemption enabled).
Zhang Xiantao [Sat, 20 Oct 2007 07:34:38 +0000 (15:34 +0800)]
KVM: Portability: Split kvm_vcpu into arch dependent and independent parts (part 1)
First step to split kvm_vcpu. Currently, we just use an macro to define
the common fields in kvm_vcpu for all archs, and all archs need to define
its own kvm_vcpu struct.
Signed-off-by: Zhang Xiantao <xiantao.zhang@intel.com> Signed-off-by: Avi Kivity <avi@qumranet.com>
projects / linux-2.6 / log