It is found that 0xf6 test instruction is used to access MMIO in
Windows. This patch added emulation for 0xf6 test instruction.
Tracked-On: #4310
Signed-off-by: Jian Jun Chen <jian.jun.chen@intel.com>
In current code, XCR0 and XSS are not in default value during vcpu
launch, it will result in a warning in Linux:
WARNING: CPU: 0 PID: 0 at arch/x86/kernel/fpu/xstate.c:614
fpu__init_system_xstate+0x43a/0x878
For security reason, we set XCR0 and XSS with feature bitmap get from
CPUID, and run XRSTORS in context switch in. This make sure the XSAVE
area to be fully in initiate state.
But, before enter guest for the first time, XCR0 and XSS should be set to
default value, as the guest kernel assume it.
Tracked-On: #4278
Signed-off-by: Conghui Chen <conghui.chen@intel.com>
We don't use INIT signal notification method now. This patch
removes them.
Tracked-On: #3886
Acked-by: Eddie Dong <eddie.dong@intel.com>
Signed-off-by: Kaige Fu <kaige.fu@intel.com>
We have implemented a new notification method using NMI.
So replace the INIT notification method with the NMI one.
Then we can remove INIT notification related code later.
Tracked-On: #3886
Signed-off-by: Kaige Fu <kaige.fu@intel.com>
There is a window where we may miss the current request in the
notification period when the work flow is as the following:
CPUx + + CPUr
| |
| +--+
| | | Handle pending req
| <--+
+--+ |
| | Set req flag |
<--+ |
+------------------>---+
| Send NMI | | Handle NMI
| <--+
| |
| |
| +--> vCPU enter
| |
+ +
So, this patch enables the NMI-window exiting to trigger the next vmexit
once there is no "virtual-NMI blocking" after vCPU enter into VMX non-root
mode. Then we can process the pending request on time.
Tracked-On: #3886
Acked-by: Eddie Dong <eddie.dong@intel.com>
Signed-off-by: Kaige Fu <kaige.fu@intel.com>
The NMI for notification should not be inject to guest. So,
this patch drops NMI injection request when we use NMI
to notify vCPUs. Meanwhile, ACRN doesn't support vNMI well
and there is no well-designed way to check if the NMI is
for notification or for guest now. So, we take all the NMIs as
notificaton NMI for hard rtvm temporarily. It means that the
hard rtvm will never receive NMI with this patch applied.
TODO: vNMI support is not ready yet. we will add it later.
Tracked-On: #3886
Signed-off-by: Kaige Fu <kaige.fu@intel.com>
ACRN hypervisor needs to kick vCPU off VMX non-root mode to do some
operations in hypervisor, such as interrupt/exception injection, EPT
flush etc. For non lapic-pt vCPUs, we can use IPI to do so. But, it
doesn't work for lapic-pt vCPUs as the IPI will be injected to VMs
directly without vmexit.
Without the way to kick the vCPU off VMX non-root mode to handle pending
request on time, there may be fatal errors triggered.
1). Certain operation may not be carried out on time which may further
lead to fatal errors. Taking the EPT flush request as an example, once we
don't flush the EPT on time and the guest access the out-of-date EPT,
fatal error happens.
2). ACRN now will send an IPI with vector 0xF0 to target vCPU to kick the vCPU
off VMX non-root mode if it wants to do some operations on target vCPU.
However, this way doesn't work for lapic-pt vCPUs. The IPI will be delivered
to the guest directly without vmexit and the guest will receive a unexpected
interrupt. Consequently, if the guest can't handle this interrupt properly,
fatal error may happen.
The NMI can be used as the notification signal to kick the vCPU off VMX
non-root mode for lapic-pt vCPUs. So, this patch uses NMI as notification signal
to address the above issues for lapic-pt vCPUs.
Tracked-On: #3886
Acked-by: Eddie Dong <eddie.dong@intel.com>
Signed-off-by: Kaige Fu <kaige.fu@intel.com>
EPT table can be changed concurrently by more than one vcpus.
This patch add a lock to protect the add/modify/delete operations
from different vcpus concurrently.
Tracked-On: #4253
Signed-off-by: Jian Jun Chen <jian.jun.chen@intel.com>
The port 0x64 is the status register of i8042 keyboard controller. When
i8042 is defined as ACPI PnP device in BIOS, enforce returning 0xff in
read handler would cause infinite loop when booting SOS VM, so expose
the physical port read in this case;
Tracked-On: #4228
Signed-off-by: Victor Sun <victor.sun@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
In current architecutre, the maximum vCPUs number per VM could not
exceed the pCPUs number. Given the MAX_PCPU_NUM macro is provided
in board configurations, so remove the MAX_VCPUS_PER_VM from Kconfig
and add a macro of MAX_VCPUS_PER_VM to reference MAX_PCPU_NUM directly.
Tracked-On: #4230
Signed-off-by: Victor Sun <victor.sun@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
rename the macro since MAX_PCPU_NUM could be parsed from board file and
it is not a configurable item anymore.
Tracked-On: #4230
Signed-off-by: Victor Sun <victor.sun@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
In APICv advanced mode, an targeted vCPU, running in non-root mode, may get outdated
TMR and EOI exit bitmap if another vCPU sends an interrupt to it if the trigger mode
of this interrupt has changed.
This patch try to kick vCPU off to let it get the latest TMR and EOI exit bitmap when
it enters non-root mode again if new coming interrupt trigger mode has changed. Then
fill the interrupt to PIR.
Tracked-On: #4200
Signed-off-by: Li Fei1 <fei1.li@intel.com>
On some platforms, HPA regions for Virtual Machine can not be
contiguous because of E820 reserved type or PCI hole. In such
cases, pre-launched VMs need to be assigned non-contiguous memory
regions and this patch addresses it.
To keep things simple, current design has the following assumptions,
1. HPA2 always will be placed after HPA1
2. HPA1 and HPA2 don’t share a single ve820 entry.
(Create multiple entries if needed but not shared)
3. Only support 2 non-contiguous HPA regions (can extend
at a later point for multiple non-contiguous HPA)
Signed-off-by: Vijay Dhanraj <vijay.dhanraj@intel.com>
Tracked-On: #4195
Acked-by: Anthony Xu <anthony.xu@intel.com>
To handle reboot requests from pre-launched VMs that don't have
GUEST_FLAG_HIGHEST_SEVERITY, we shutdown the target VM explicitly
other than ignoring them.
Tracked-On: #2700
Signed-off-by: Zide Chen <zide.chen@intel.com>
Acked-by: Anthony Xu <anthony.xu@intel.com>
ptirq_prepare_msix_remap was called no matter whether MSI/MSI-X was enabled or not
and it passed zero to input parameter virtual MSI/MSI-X data field to indicate
MSI/MSI-X was disabled. However, it barely did nothing on this case.
Now ptirq_prepare_msix_remap is called only when MSI/MSI-X is enabled. It doesn't
need to check whether MSI/MSI-X is enabled or not by checking virtual MSI/MSI-X
data field.
Tracked-On: #3475
Signed-off-by: Li Fei1 <fei1.li@intel.com>
It's meaningless to sleep a non-running vcpu. Add a state check before
sleep the thread object of the vcpu.
Tracked-On: #4178
Signed-off-by: Shuo A Liu <shuo.a.liu@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
With cpu-sharing enabled, there are more than 1 vcpu on 1 pcpu, so the
smp_call handler should switch the vmcs to the target vcpu's vmcs. Then
get the info.
dump_vcpu_reg and dump_guest_mem should run on certain vmcs, otherwise,
there will be #GP error.
Renaming:
vcpu_dumpreg -> dump_vcpu_reg
switch_vmcs -> load_vmcs
Tracked-On: #4178
Signed-off-by: Conghui Chen <conghui.chen@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
We care more about leaf and subleaf of cpuid than vcpu_id.
So, this patch changes the cpuid trace-entry to trace the leaf
and subleaf of this cpuid vmexit.
Tracked-On: #4175
Signed-off-by: Kaige Fu <kaige.fu@intel.com>
PMU is hidden from any guest, UD is expected when guest
try to execute 'rdpmc' instruction.
this patch sets 'RDPMC exiting' in Processorbased
VM-execution control.
Tracked-On: #3453
Signed-off-by: Yonghua Huang <yonghua.huang@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
After changing init_vmcs to smp call approach and do it before
launch_vcpu, it could work with noop scheduler. On real sharing
scheudler, it has problem.
pcpu0 pcpu1 pcpu1
vmBvcpu0 vmAvcpu1 vmBvcpu1
vmentry
init_vmcs(vmBvcpu1) vmexit->do_init_vmcs
corrupt current vmcs
vmentry fail
launch_vcpu(vmBvcpu1)
This patch mark a event flag when request vmcs init for specific vcpu. When
it is running and checking pending events, will do init_vmcs firstly.
Tracked-On: #4178
Signed-off-by: Shuo A Liu <shuo.a.liu@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
Starting with TSC_DEADLINE msr interception disabled, the virtual TSC_DEADLINE msr is always 0.
When the interception is enabled, need to sync the physical TSC_DEADLINE value to virtual TSC_DEADLINE.
When the interception is disabled, there are 2 cases:
- if the timer hasn't expired, sync virtual TSC_DEADLINE to physical TSC_DEADLINE, to make the guest read the same tsc_deadline
as it writes. This may change when the timer actually trigger.
- if the timer has expired, write 0 to the virtual TSC_DEADLINE.
Tracked-On: #4162
Signed-off-by: Yan, Like <like.yan@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
When write to virtual TSC_DEADLINE, if virtual TSC_ADJUST is not zero:
- when guest intends to disarm the tsc_deadline timer, should not arm the timer falsely;
- when guest intends to arm the tsc_deadline timer, should not disarm the timer falsely.
When read from virtual TSC_DEADLINE, if virtual TSC_ADJUST is not zero:
- if physical TSC_DEADLINE is not zero, return the virtual TSC_DEADLINE value;
- if physical TSC_DEADLINE is zero which means it's not armed (automatically disarmed after
timer triggered), return 0 and reset the virtual TSC_DEADLINE.
Tracked-On: #4162
Signed-off-by: Yan, Like <like.yan@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
xsave area:
legacy region: 512 bytes
xsave header: 64 bytes
extended region: < 3k bytes
So, pre-allocate 4k area for xsave. Use certain instruction to save or
restore the area according to hardware xsave feature set.
Tracked-On: #4166
Signed-off-by: Conghui Chen <conghui.chen@intel.com>
Reviewed-by: Anthony Xu <anthony.xu@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
ptirq_msix_remap doesn't do the real remap, that's the vmsi_remap and vmsix_remap_entry
does. ptirq_msix_remap only did the preparation.
Tracked-On: #3475
Signed-off-by: Li Fei1 <fei1.li@intel.com>
Issue description:
-----------------
Machine Check Error on Page Size Change
Instruction fetch may cause machine check error if page size
and memory type was changed without invalidation on some
processors[1][2]. Malicious guest kernel could trigger this issue.
This issue applies to both primary page table and extended page
tables (EPT), however the primary page table is controlled by
hypervisor only. This patch mitigates the situation in EPT.
Mitigation details:
------------------
Implement non-execute huge pages in EPT.
This patch series clears the execute permission (bit 2) in the
EPT entries for large pages. When EPT violation is triggered by
guest instruction fetch, hypervisor converts the large page to
smaller 4 KB pages and restore the execute permission, and then
re-execute the guest instruction.
The current patch turns on the mitigation by default.
The follow-up patches will conditionally turn on/off the feature
per processor model.
[1] Refer to erratum KBL002 in "7th Generation Intel Processor
Family and 8th Generation Intel Processor Family for U Quad Core
Platforms Specification Update"
https://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/7th-gen-core-family-spec-update.pdf
[2] Refer to erratum SKL002 in "6th Generation Intel Processor
Family Specification Update"
https://www.intel.com/content/www/us/en/products/docs/processors/core/desktop-6th-gen-core-family-spec-update.html
Tracked-On: #4101
Signed-off-by: Binbin Wu <binbin.wu@intel.com>
Reviewed-by: Eddie Dong <eddie.dong@intel.com>
In non-64-bit mode, CS segment base address should be considered when
determining the linear address of the vcpu's instruction pointer. Use
vie_calculate_gla() for instruction address translation which also takes
care of 64-bit mode.
Tracked-On: #4064
Signed-off-by: Peter Fang <peter.fang@intel.com>
MISRA C requires specified bounds for arrays declaration, previous declaration
of platform_clos_array in board.h does not meet the requirement.
Tracked-On: #3987
Signed-off-by: Victor Sun <victor.sun@intel.com>
hypcall_id has a type of uint64_t and should use 'llx' as
formatting flag instead of '%d'. Otherwise, we will get a
confusing error log when not-allowed hypercall occurs.
Without this patch:
[96707209us][cpu=1][sev=3][seq=2386]:hypercall -2147483548 is only allowed from SOS_VM!
With this patch:
[84613395us][cpu=1][sev=3][seq=2136]:hypercall 0x80000064 is only allowed from SOS_VM!
So, we can figure out which not-allowed hypercall has been triggered more conveniently.
BTW, this patch adds hypcall_id which triggered from non-ring0 into error log.
Tracked-On: #4012
Signed-off-by: Kaige Fu <kaige.fu@intel.com>
Now the e820 structure store ACRN HV memory layout, not the physical memory layout.
Rename e820 to hv_hv_e820 to show this explicitly.
Tracked-On: #4007
Signed-off-by: Li Fei1 <fei1.li@intel.com>
AP trampoline code should be accessible
to hypervisor only, this patch is to unmap
this region from service VM's EPT for security
reason.
Tracked-On: #3992
Signed-off-by: Yonghua Huang <yonghua.huang@intel.com>
Reviewed-by: Fei Li <fei1.li@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
We should use INIT signal to notify the vcpu threads when
powering off the hard RTVM. To achive this, we should set
the vcpu->thread_obj.notify_mode as SCHED_NOTIFY_INIT.
Patch (27163df9 hv: sched: add sleep/wake for thread object)
tries to set the notify_mode according `is_lapic_pt_enabled(vcpu)`
in function prepare_vcpu. But at this point, the is_lapic_pt_enabled(vcpu)
will always return false. Consequently, it will set notify_mode
as SCHED_NOTIFY_IPI. Then leads to the failure of powering off
hard RTVM.
This patch fixes it by:
- Initialize the notify_mode as SCHED_NOTIFY_IPI in prepare_vcpu.
- Set notify_mode as SCHED_NOTIFY_INIT after guest is trying to
enable x2apic mode of passthru lapic.
Tracked-On: #3975
Reviewed-by: Yin Fengwei <fengwei.yin@intel.com>
Reviewed-by: Yan, Like <like.yan@intel.com>
Signed-off-by: Kaige Fu <kaige.fu@intel.com>
After adding PCI BAR remap support, mmio_node may unregister when there's others
access it. This patch add a lock to protect mmio_node access.
Tracked-On: #3475
Signed-off-by: Li, Fei1 <fei1.li@intel.com>
Since guest could re-program PCI device MSI-X table BAR, we should add mmio
emulation handler unregister.
However, after add unregister_mmio_emulation_handler API, emul_mmio_regions
is no longer accurate. Just replace it with max_emul_mmio_regions which records
the max index of the emul_mmio_node.
Tracked-On: #3475
Signed-off-by: Li, Fei1 <fei1.li@intel.com>
In theory, guest could re-program PCI BAR address to any address. However, ACRN
hypervisor only support [0, top_address_space) EPT memory mapping. So we need to
check whether the PCI BAR re-program address is within this scope.
Tracked-On: #3475
Signed-off-by: Li, Fei1 <fei1.li@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
kick means to notify one thread_object. If the target thread object is
running, send a IPI to notify it; if the target thread object is
runnable, make reschedule on it.
Also add kick_vcpu API in vcpu layer to notify vcpu.
Tracked-On: #3813
Signed-off-by: Jason Chen CJ <jason.cj.chen@intel.com>
Signed-off-by: Yu Wang <yu1.wang@intel.com>
Signed-off-by: Shuo A Liu <shuo.a.liu@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
This patch decouple some scheduling logic and abstract into a scheduler.
Then we have scheduler, schedule framework. From modulization
perspective, schedule framework provides some APIs for other layers to
use, also interact with scheduler through scheduler interaces.
Tracked-On: #3813
Signed-off-by: Jason Chen CJ <jason.cj.chen@intel.com>
Signed-off-by: Yu Wang <yu1.wang@intel.com>
Signed-off-by: Shuo A Liu <shuo.a.liu@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
- The default behaviors of PIO & MMIO handlers are same
for all VMs, no need to expose dedicated APIs to register
default hanlders for SOS and prelaunched VM.
Tracked-On: #3904
Signed-off-by: Yonghua Huang <yonghua.huang@intel.com>
Reviewed-by: Junjie Mao <junjie.mao@intel.com>
Currently the parameter of init_ept_mem_ops is
'struct acrn_vm *vm' for this api,change it to
'struct memory_ops *mem_ops' and 'vm_id' to avoid
the reversed dependency, page.c is hardware layer and vm structure
is its upper-layer stuff.
Tracked-On: #1842
Signed-off-by: Mingqiang Chi <mingqiang.chi@intel.com>
Reviewed-by: Eddie Dong <eddie.dong@intel.com>
Let init thread end with run_idle_thread(), then idle thread take over and
start to do scheduling.
Change enter_guest_mode() to init_guest_mode() as run_idle_thread() is removed
out of it. Also add run_thread() in schedule module to run
thread_object's thread loop directly.
rename: switch_to_idle -> run_idle_thread
Tracked-On: #3813
Signed-off-by: Shuo A Liu <shuo.a.liu@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
sleep one thread_object means to prevent it from being scheduled.
wake one thread_object is an opposite operation of sleep.
This patch also add notify_mode in thread_object to indicate how to
deliver the request.
Tracked-On: #3813
Signed-off-by: Jason Chen CJ <jason.cj.chen@intel.com>
Signed-off-by: Yu Wang <yu1.wang@intel.com>
Signed-off-by: Shuo A Liu <shuo.a.liu@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
To support cpu sharing, multiple vcpu can run on same pcpu. We need do
necessary vcpu context switch. This patch add below actions in context
switch.
1) fxsave/fxrstor;
2) save/restore MSRs: MSR_IA32_STAR, MSR_IA32_LSTAR,
MSR_IA32_FMASK, MSR_IA32_KERNEL_GS_BASE;
3) switch vmcs.
Tracked-On: #3813
Signed-off-by: Jason Chen CJ <jason.cj.chen@intel.com>
Signed-off-by: Yu Wang <yu1.wang@intel.com>
Signed-off-by: Shuo A Liu <shuo.a.liu@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
With cpu sharing enabled, per_cpu vcpu cannot work properly as we might
has multiple vcpus running on one pcpu.
Add a schedule API sched_get_current to get current thread_object on
specific pcpu, also add a vcpu API get_running_vcpu to get corresponding
vcpu of the thread_object.
Tracked-On: #3813
Signed-off-by: Jason Chen CJ <jason.cj.chen@intel.com>
Signed-off-by: Yu Wang <yu1.wang@intel.com>
Signed-off-by: Shuo A Liu <shuo.a.liu@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
With cpu sharing enabled, we will map acrn_vcpu to thread_object
in scheduling. From modulization perspective, we'd better hide the
pcpu_id in acrn_vcpu and move it to thread_object.
Tracked-On: #3813
Signed-off-by: Jason Chen CJ <jason.cj.chen@intel.com>
Signed-off-by: Yu Wang <yu1.wang@intel.com>
Signed-off-by: Shuo A Liu <shuo.a.liu@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
vcpu thread's stack shouldn't follow reset_vcpu to reset.
There is also a bug here:
while vcpu B thread set vcpu->running to false, other vcpu A thread
will treat the vcpu B is paused while it has not been switch out
completely, then reset_vcpu will reset the vcpu B thread's stack and
corrupt its running context.
This patch will remove the vcpu thread's stack reset from reset_vcpu.
With the change, we need do init_vmcs between vcpu startup address be
settled and scheduled in. And switch_to_idle() is not needed anymore
as S3 thread's stack will not be reset.
Tracked-On: #3813
Signed-off-by: Fengwei Yin <fengwei.yin@intel.com>
Signed-off-by: Shuo A Liu <shuo.a.liu@intel.com>
