* requiring virtual NMI for Intel processors?
@ 2017-02-20 19:43 Paolo Bonzini
2017-02-20 20:35 ` Nadav Amit
2017-02-21 16:26 ` Jim Mattson
0 siblings, 2 replies; 5+ messages in thread
From: Paolo Bonzini @ 2017-02-20 19:43 UTC (permalink / raw)
To: KVM list
[-- Attachment #1: Type: text/plain, Size: 777 bytes --]
I spent some time collecting vmxcap output for various Intel processors.
I tested Prescott (Cedar Mill actually), Yonah (32-bit only) and every
microarchitecture from Conroe to Haswell. As far as I can tell, these
are all Intel processors that supported virtualization. My notes are
attached.
The only ones without virtual NMI and NMI-window exiting were the Cedar
Mill and Yonah. One is a Pentium 4 microarchitecture that one should
use as a toaster rather than as a hypervisor; the other is not even
64-bit capable. It also has VMCS restricted to 32 bit addresses, so if
it ever had >3.5GB memory installed KVM would break.
Based on this, I'd like to drop the code that emulates the NMI window
and just require these two features in KVM 4.11+.
Any objections?
Paolo
[-- Attachment #2: VMX-CAP --]
[-- Type: text/plain, Size: 9517 bytes --]
Cedar Mill (15:6:4) Prescott
Basic VMX Information
Hex: 0x1a100000000003
Revision 3
VMCS size 4096
VMCS restricted to 32 bit addresses no
Dual-monitor support yes !!
VMCS memory type 6
INS/OUTS instruction information no
IA32_VMX_TRUE_*_CTLS support no
pin-based controls
External interrupt exiting yes
NMI exiting yes
Virtual NMIs no
Activate VMX-preemption timer no
Process posted interrupts no
primary processor-based controls
Interrupt window exiting yes
Use TSC offsetting yes
HLT exiting yes
INVLPG exiting yes
MWAIT exiting yes
RDPMC exiting yes
RDTSC exiting yes
CR3-load exiting forced
CR3-store exiting forced
CR8-load exiting yes
CR8-store exiting yes
Use TPR shadow yes
NMI-window exiting no
MOV-DR exiting yes
Unconditional I/O exiting yes
Use I/O bitmaps yes
Monitor trap flag no
Use MSR bitmaps no
MONITOR exiting yes
PAUSE exiting yes
Activate secondary control no
secondary processor-based controls
Virtualize APIC accesses no
Enable EPT no
Descriptor-table exiting no
Enable RDTSCP no
Virtualize x2APIC mode no
Enable VPID no
WBINVD exiting no
Unrestricted guest no
APIC register emulation no
Virtual interrupt delivery no
PAUSE-loop exiting no
RDRAND exiting no
Enable INVPCID no
Enable VM functions no
VMCS shadowing no
RDSEED exiting no
EPT-violation #VE no
Enable XSAVES/XRSTORS no
TSC scaling no
VM-Exit controls
Save debug controls forced
Host address-space size yes
Load IA32_PERF_GLOBAL_CTRL no
Acknowledge interrupt on exit yes
Save IA32_PAT no
Load IA32_PAT no
Save IA32_EFER no
Load IA32_EFER no
Save VMX-preemption timer value no
VM-Entry controls
Load debug controls forced
IA-32e mode guest yes
Entry to SMM yes !!
Deactivate dual-monitor treatment yes !!
Load IA32_PERF_GLOBAL_CTRL no
Load IA32_PAT no
Load IA32_EFER no
Miscellaneous data
Hex: 0x401c0
VMX-preemption timer scale (log2) 0
Store EFER.LMA into IA-32e mode guest control no
HLT activity state yes
Shutdown activity state yes
Wait-for-SIPI activity state yes
IA32_SMBASE support no
Number of CR3-target values 4
MSR-load/store count recommendation 0
IA32_SMM_MONITOR_CTL[2] can be set to 1 no
VMWRITE to VM-exit information fields no
MSEG revision identifier 0
VPID and EPT capabilities
Hex: 0x0
Execute-only EPT translations no
Page-walk length 4 no
Paging-structure memory type UC no
Paging-structure memory type WB no
2MB EPT pages no
1GB EPT pages no
INVEPT supported no
EPT accessed and dirty flags no
Single-context INVEPT no
All-context INVEPT no
INVVPID supported no
Individual-address INVVPID no
Single-context INVVPID no
All-context INVVPID no
Single-context-retaining-globals INVVPID no
VM Functions
Hex: 0x0
EPTP Switching no
Yonah (6:14:12) Intel(R) Core(TM) Duo CPU T2500 @ 2.00GHz
flags : fpu vme de pse tsc msr pae mce cx8 apic mtrr pge mca cmov pat clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe constant_tsc pni monitor vmx est tm2 xtpr
CR8-load exiting no
CR8-store exiting no
Use TPR shadow no
Use MSR bitmaps yes
Conroe (6:15:11) Intel(R) Xeon(R) CPU X3220 @ 2.40GHz
Basic VMX Information
INS/OUTS instruction information yes
pin-based controls
Virtual NMIs yes
primary processor-based controls
NMI-window exiting yes
Use MSR bitmaps yes
secondary processor-based controls
Virtualize APIC accesses yes
Penryn (6:23:10) Intel(R) Xeon(R) CPU E5430 @ 2.66GHz
same as Conroe but:
secondary processor-based controls
WBINVD exiting yes
VM-Exit controls
Load IA32_PERF_GLOBAL_CTRL yes
VM-Entry controls
Load IA32_PERF_GLOBAL_CTRL yes
Nehalem (6:26:5) Intel(R) Xeon(R) CPU E5530 @ 2.40GHz
same as Penryn but:
Basic VMX Information
IA32_VMX_TRUE_*_CTLS support yes
pin-based controls
Activate VMX-preemption timer yes
primary processor-based controls
CR3-load exiting default
CR3-store exiting default
Monitor trap flag yes
secondary processor-based controls
Enable EPT yes
Descriptor-table exiting yes
Enable RDTSCP yes
Virtualize x2APIC mode yes
Enable VPID yes
VM-Exit controls
Save debug controls default
Save IA32_PAT yes
Load IA32_PAT yes
Save IA32_EFER yes
Load IA32_EFER yes
Save VMX-preemption timer value yes
VM-Entry controls
Load debug controls default
Load IA32_PAT yes
Load IA32_EFER yes
Miscellaneous data
Hex: 0x401c5
VMX-preemption timer scale (log2) 5
VPID and EPT capabilities
Hex: 0xf0106114141
Execute-only EPT translations yes
Page-walk length 4 yes
Paging-structure memory type UC yes
Paging-structure memory type WB yes
2MB EPT pages yes
INVEPT supported yes
Single-context INVEPT yes
All-context INVEPT yes
INVVPID supported yes
Individual-address INVVPID yes
Single-context INVVPID yes
All-context INVVPID yes
Single-context-retaining-globals INVVPID yes
Westmere 6:44:2 Intel(R) Xeon(R) CPU E5620 @ 2.40GHz
same as Nehalem but:
secondary processor-based controls
Unrestricted guest yes
PAUSE-loop exiting yes
Miscellaneous data
Store EFER.LMA into IA-32e mode guest control yes
(forgot to check 1GB pages)
SandyBridge 6:45:7 Intel(R) Xeon(R) CPU E5-2407 0 @ 2.20GHz
same as Westmere but:
Miscellaneous data
IA32_SMM_MONITOR_CTL[2] can be set to 1 yes !!
VPID and EPT capabilities
1GB EPT pages yes
IvyBridge 6:62:4 Intel(R) Xeon(R) CPU E5-2620 v2 @ 2.10GHz
same as SandyBridge but:
pin-based controls
Process posted interrupts yes
secondary processor-based controls
APIC register emulation yes
Virtual interrupt delivery yes
RDRAND exiting yes
Haswell 6:63:2 Intel(R) Xeon(R) CPU E5-2697 v3 @ 2.60GHz
same as Haswell but:
secondary processor-based controls
Enable INVPCID yes
Enable VM functions yes
VMCS shadowing yes
VM-Entry controls
IA32_SMBASE support yes !!
VMWRITE to VM-exit information fields yes
VPID and EPT capabilities
EPT accessed and dirty flags yes
VM Functions
EPTP Switching yes
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: requiring virtual NMI for Intel processors?
2017-02-20 19:43 requiring virtual NMI for Intel processors? Paolo Bonzini
@ 2017-02-20 20:35 ` Nadav Amit
2017-02-20 22:00 ` Paolo Bonzini
2017-02-21 16:26 ` Jim Mattson
1 sibling, 1 reply; 5+ messages in thread
From: Nadav Amit @ 2017-02-20 20:35 UTC (permalink / raw)
To: Paolo Bonzini; +Cc: KVM list
I have no information that suggests otherwise, but just a couple of small
comments:
1. It seems you look on Xeons (excluding Yonah and Cedarmill). Desktop and
laptop CPUs may have different fusing than servers.
2. No Atoms in the list.
> On Feb 20, 2017, at 11:43 AM, Paolo Bonzini <pbonzini@redhat.com> wrote:
>
> I spent some time collecting vmxcap output for various Intel processors.
> I tested Prescott (Cedar Mill actually), Yonah (32-bit only) and every
> microarchitecture from Conroe to Haswell. As far as I can tell, these
> are all Intel processors that supported virtualization. My notes are
> attached.
>
> The only ones without virtual NMI and NMI-window exiting were the Cedar
> Mill and Yonah. One is a Pentium 4 microarchitecture that one should
> use as a toaster rather than as a hypervisor; the other is not even
> 64-bit capable. It also has VMCS restricted to 32 bit addresses, so if
> it ever had >3.5GB memory installed KVM would break.
>
> Based on this, I'd like to drop the code that emulates the NMI window
> and just require these two features in KVM 4.11+.
>
> Any objections?
>
> Paolo
> <VMX-CAP.txt>
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: requiring virtual NMI for Intel processors?
2017-02-20 20:35 ` Nadav Amit
@ 2017-02-20 22:00 ` Paolo Bonzini
2017-02-20 22:41 ` Nadav Amit
0 siblings, 1 reply; 5+ messages in thread
From: Paolo Bonzini @ 2017-02-20 22:00 UTC (permalink / raw)
To: Nadav Amit; +Cc: KVM list
> 1. It seems you look on Xeons (excluding Yonah and Cedarmill). Desktop and
> laptop CPUs may have different fusing than servers.
Yeah, the attached file mostly included Xeons because I couldn't find a
desktop or laptop CPU for each generation. However I did find Penryn and
Nehalem desktop CPUs and they already had vNMI support.
I stopped at Haswell because we still don't support nesting some
Broadwell+ extensions (e.g. PML). In fact we don't support EPT A/D
bits too, though that one shouldn't be hard to add.
The list of server-only features seems to be small though until Haswell:
only EPT 1GB page tables, PLE and APICv.
> 2. No Atoms in the list.
Centerton (Saltwell-based) is the only one I have access to and it's
the same as Conroe, so it has it.
Paolo
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: requiring virtual NMI for Intel processors?
2017-02-20 22:00 ` Paolo Bonzini
@ 2017-02-20 22:41 ` Nadav Amit
0 siblings, 0 replies; 5+ messages in thread
From: Nadav Amit @ 2017-02-20 22:41 UTC (permalink / raw)
To: Paolo Bonzini; +Cc: KVM list
Got it. One last point: although it is highly unlikely to cause a
problem, you may want to check hypervisors that support nested
virtualization too.
FWIW: My VMware Fusion 8.5.3 exposes both features.
> On Feb 20, 2017, at 2:00 PM, Paolo Bonzini <pbonzini@redhat.com> wrote:
>
>
>> 1. It seems you look on Xeons (excluding Yonah and Cedarmill). Desktop and
>> laptop CPUs may have different fusing than servers.
>
> Yeah, the attached file mostly included Xeons because I couldn't find a
> desktop or laptop CPU for each generation. However I did find Penryn and
> Nehalem desktop CPUs and they already had vNMI support.
>
> I stopped at Haswell because we still don't support nesting some
> Broadwell+ extensions (e.g. PML). In fact we don't support EPT A/D
> bits too, though that one shouldn't be hard to add.
>
> The list of server-only features seems to be small though until Haswell:
> only EPT 1GB page tables, PLE and APICv.
>
>> 2. No Atoms in the list.
>
> Centerton (Saltwell-based) is the only one I have access to and it's
> the same as Conroe, so it has it.
>
> Paolo
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: requiring virtual NMI for Intel processors?
2017-02-20 19:43 requiring virtual NMI for Intel processors? Paolo Bonzini
2017-02-20 20:35 ` Nadav Amit
@ 2017-02-21 16:26 ` Jim Mattson
1 sibling, 0 replies; 5+ messages in thread
From: Jim Mattson @ 2017-02-21 16:26 UTC (permalink / raw)
To: Paolo Bonzini; +Cc: KVM list
Go for it!
On Mon, Feb 20, 2017 at 11:43 AM, Paolo Bonzini <pbonzini@redhat.com> wrote:
> I spent some time collecting vmxcap output for various Intel processors.
> I tested Prescott (Cedar Mill actually), Yonah (32-bit only) and every
> microarchitecture from Conroe to Haswell. As far as I can tell, these
> are all Intel processors that supported virtualization. My notes are
> attached.
>
> The only ones without virtual NMI and NMI-window exiting were the Cedar
> Mill and Yonah. One is a Pentium 4 microarchitecture that one should
> use as a toaster rather than as a hypervisor; the other is not even
> 64-bit capable. It also has VMCS restricted to 32 bit addresses, so if
> it ever had >3.5GB memory installed KVM would break.
>
> Based on this, I'd like to drop the code that emulates the NMI window
> and just require these two features in KVM 4.11+.
>
> Any objections?
>
> Paolo
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2017-02-21 16:26 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-02-20 19:43 requiring virtual NMI for Intel processors? Paolo Bonzini
2017-02-20 20:35 ` Nadav Amit
2017-02-20 22:00 ` Paolo Bonzini
2017-02-20 22:41 ` Nadav Amit
2017-02-21 16:26 ` Jim Mattson
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.