[PATCH] KVM: x86: obey KVM_X86_QUIRK_CD_NW_CLEARED in kvm_set_cr0()

[PATCH] KVM: x86: obey KVM_X86_QUIRK_CD_NW_CLEARED in kvm_set_cr0()

[Laszlo Ersek]

Commit b18d5431acc7 ("KVM: x86: fix CR0.CD virtualization") was
technically correct, but it broke OVMF guests by slowing down various
parts of the firmware.

Commit fb279950ba02 ("KVM: vmx: obey KVM_QUIRK_CD_NW_CLEARED") quirked the
first function modified by b18d5431acc7, vmx_get_mt_mask(), for OVMF's
sake. This restored the speed of the OVMF code that runs before
PlatformPei (including the memory intensive LZMA decompression in SEC).

This patch extends the quirk to the second function modified by
b18d5431acc7, kvm_set_cr0(). It eliminates the intrusive slowdown that
hits the EFI_MP_SERVICES_PROTOCOL implementation of edk2's
UefiCpuPkg/CpuDxe -- which is built into OVMF --, when CpuDxe starts up
all APs at once for initialization, in order to count them.

We also carry over the kvm_arch_has_noncoherent_dma() sub-condition from
the other half of the original commit b18d5431acc7.

Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Janusz Mocek <januszmk6@gmail.com>
Cc: Alex Williamson <alex.williamson@redhat.com>
Cc: Xiao Guangrong <guangrong.xiao@linux.intel.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
---
 arch/x86/kvm/x86.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index a24bae0..30723a4 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -625,7 +625,9 @@ int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
 	if ((cr0 ^ old_cr0) & update_bits)
 		kvm_mmu_reset_context(vcpu);
 
-	if ((cr0 ^ old_cr0) & X86_CR0_CD)
+	if (((cr0 ^ old_cr0) & X86_CR0_CD) &&
+	    kvm_arch_has_noncoherent_dma(vcpu->kvm) &&
+	    !kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_CD_NW_CLEARED))
 		kvm_zap_gfn_range(vcpu->kvm, 0, ~0ULL);
 
 	return 0;
-- 
1.8.3.1

Re: [PATCH] KVM: x86: obey KVM_X86_QUIRK_CD_NW_CLEARED in kvm_set_cr0()

[Laszlo Ersek]

On 11/03/15 19:34, Laszlo Ersek wrote:
> Commit b18d5431acc7 ("KVM: x86: fix CR0.CD virtualization") was
> technically correct, but it broke OVMF guests by slowing down various
> parts of the firmware.
> 
> Commit fb279950ba02 ("KVM: vmx: obey KVM_QUIRK_CD_NW_CLEARED") quirked the
> first function modified by b18d5431acc7, vmx_get_mt_mask(), for OVMF's
> sake. This restored the speed of the OVMF code that runs before
> PlatformPei (including the memory intensive LZMA decompression in SEC).
> 
> This patch extends the quirk to the second function modified by
> b18d5431acc7, kvm_set_cr0(). It eliminates the intrusive slowdown that
> hits the EFI_MP_SERVICES_PROTOCOL implementation of edk2's
> UefiCpuPkg/CpuDxe -- which is built into OVMF --, when CpuDxe starts up
> all APs at once for initialization, in order to count them.
> 
> We also carry over the kvm_arch_has_noncoherent_dma() sub-condition from
> the other half of the original commit b18d5431acc7.
> 
> Cc: Paolo Bonzini <pbonzini@redhat.com>
> Cc: Jordan Justen <jordan.l.justen@intel.com>
> Cc: Janusz Mocek <januszmk6@gmail.com>
> Cc: Alex Williamson <alex.williamson@redhat.com>
> Cc: Xiao Guangrong <guangrong.xiao@linux.intel.com>
> Signed-off-by: Laszlo Ersek <lersek@redhat.com>
> ---
>  arch/x86/kvm/x86.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index a24bae0..30723a4 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -625,7 +625,9 @@ int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
>  	if ((cr0 ^ old_cr0) & update_bits)
>  		kvm_mmu_reset_context(vcpu);
>  
> -	if ((cr0 ^ old_cr0) & X86_CR0_CD)
> +	if (((cr0 ^ old_cr0) & X86_CR0_CD) &&
> +	    kvm_arch_has_noncoherent_dma(vcpu->kvm) &&
> +	    !kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_CD_NW_CLEARED))
>  		kvm_zap_gfn_range(vcpu->kvm, 0, ~0ULL);
>  
>  	return 0;
> 


I had notes on this patch, but I forgot to format it with --notes. They
were:

- People on the CC list, please reply with your Tested-by, Reported-by,
  etc tags as appropriate; it's getting blurry who participated in what
  and how.

- This patch is *not* necessary for the OVMF SMM work; instead it
  addresses an independent OVMF boot regression seen by users.

Thanks
Laszlo

Re: [PATCH] KVM: x86: obey KVM_X86_QUIRK_CD_NW_CLEARED in kvm_set_cr0()

[Janusz Mocek]

W dniu 03.11.2015 o 19:57, Laszlo Ersek pisze:
> On 11/03/15 19:34, Laszlo Ersek wrote:
>> Commit b18d5431acc7 ("KVM: x86: fix CR0.CD virtualization") was
>> technically correct, but it broke OVMF guests by slowing down various
>> parts of the firmware.
>>
>> Commit fb279950ba02 ("KVM: vmx: obey KVM_QUIRK_CD_NW_CLEARED") quirked the
>> first function modified by b18d5431acc7, vmx_get_mt_mask(), for OVMF's
>> sake. This restored the speed of the OVMF code that runs before
>> PlatformPei (including the memory intensive LZMA decompression in SEC).
>>
>> This patch extends the quirk to the second function modified by
>> b18d5431acc7, kvm_set_cr0(). It eliminates the intrusive slowdown that
>> hits the EFI_MP_SERVICES_PROTOCOL implementation of edk2's
>> UefiCpuPkg/CpuDxe -- which is built into OVMF --, when CpuDxe starts up
>> all APs at once for initialization, in order to count them.
>>
>> We also carry over the kvm_arch_has_noncoherent_dma() sub-condition from
>> the other half of the original commit b18d5431acc7.
>>
>> Cc: Paolo Bonzini <pbonzini@redhat.com>
>> Cc: Jordan Justen <jordan.l.justen@intel.com>
>> Cc: Janusz Mocek <januszmk6@gmail.com>
>> Cc: Alex Williamson <alex.williamson@redhat.com>
>> Cc: Xiao Guangrong <guangrong.xiao@linux.intel.com>
>> Signed-off-by: Laszlo Ersek <lersek@redhat.com>
>> ---
>>  arch/x86/kvm/x86.c | 4 +++-
>>  1 file changed, 3 insertions(+), 1 deletion(-)
>>
>> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
>> index a24bae0..30723a4 100644
>> --- a/arch/x86/kvm/x86.c
>> +++ b/arch/x86/kvm/x86.c
>> @@ -625,7 +625,9 @@ int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
>>  	if ((cr0 ^ old_cr0) & update_bits)
>>  		kvm_mmu_reset_context(vcpu);
>>  
>> -	if ((cr0 ^ old_cr0) & X86_CR0_CD)
>> +	if (((cr0 ^ old_cr0) & X86_CR0_CD) &&
>> +	    kvm_arch_has_noncoherent_dma(vcpu->kvm) &&
>> +	    !kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_CD_NW_CLEARED))
>>  		kvm_zap_gfn_range(vcpu->kvm, 0, ~0ULL);
>>  
>>  	return 0;
>>
>
> I had notes on this patch, but I forgot to format it with --notes. They
> were:
>
> - People on the CC list, please reply with your Tested-by, Reported-by,
>   etc tags as appropriate; it's getting blurry who participated in what
>   and how.
>
> - This patch is *not* necessary for the OVMF SMM work; instead it
>   addresses an independent OVMF boot regression seen by users.
>
> Thanks
> Laszlo
Tested-by: Janusz Mocek <januszmk6@gmail.com>

Re: [PATCH] KVM: x86: obey KVM_X86_QUIRK_CD_NW_CLEARED in kvm_set_cr0()

[Xiao Guangrong]

On 11/04/2015 02:34 AM, Laszlo Ersek wrote:
> Commit b18d5431acc7 ("KVM: x86: fix CR0.CD virtualization") was
> technically correct, but it broke OVMF guests by slowing down various
> parts of the firmware.
>
> Commit fb279950ba02 ("KVM: vmx: obey KVM_QUIRK_CD_NW_CLEARED") quirked the
> first function modified by b18d5431acc7, vmx_get_mt_mask(), for OVMF's
> sake. This restored the speed of the OVMF code that runs before
> PlatformPei (including the memory intensive LZMA decompression in SEC).
>
> This patch extends the quirk to the second function modified by
> b18d5431acc7, kvm_set_cr0(). It eliminates the intrusive slowdown that
> hits the EFI_MP_SERVICES_PROTOCOL implementation of edk2's
> UefiCpuPkg/CpuDxe -- which is built into OVMF --, when CpuDxe starts up
> all APs at once for initialization, in order to count them.
>
> We also carry over the kvm_arch_has_noncoherent_dma() sub-condition from
> the other half of the original commit b18d5431acc7.

Reviewed-by: Xiao Guangrong <guangrong.xiao@linux.intel.com>

Re: [PATCH] KVM: x86: obey KVM_X86_QUIRK_CD_NW_CLEARED in kvm_set_cr0()

[Paolo Bonzini]

On 03/11/2015 19:34, Laszlo Ersek wrote:
> Commit b18d5431acc7 ("KVM: x86: fix CR0.CD virtualization") was
> technically correct, but it broke OVMF guests by slowing down various
> parts of the firmware.
> 
> Commit fb279950ba02 ("KVM: vmx: obey KVM_QUIRK_CD_NW_CLEARED") quirked the
> first function modified by b18d5431acc7, vmx_get_mt_mask(), for OVMF's
> sake. This restored the speed of the OVMF code that runs before
> PlatformPei (including the memory intensive LZMA decompression in SEC).
> 
> This patch extends the quirk to the second function modified by
> b18d5431acc7, kvm_set_cr0(). It eliminates the intrusive slowdown that
> hits the EFI_MP_SERVICES_PROTOCOL implementation of edk2's
> UefiCpuPkg/CpuDxe -- which is built into OVMF --, when CpuDxe starts up
> all APs at once for initialization, in order to count them.
> 
> We also carry over the kvm_arch_has_noncoherent_dma() sub-condition from
> the other half of the original commit b18d5431acc7.
> 
> Cc: Paolo Bonzini <pbonzini@redhat.com>
> Cc: Jordan Justen <jordan.l.justen@intel.com>
> Cc: Janusz Mocek <januszmk6@gmail.com>
> Cc: Alex Williamson <alex.williamson@redhat.com>
> Cc: Xiao Guangrong <guangrong.xiao@linux.intel.com>
> Signed-off-by: Laszlo Ersek <lersek@redhat.com>
> ---
>  arch/x86/kvm/x86.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index a24bae0..30723a4 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -625,7 +625,9 @@ int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
>  	if ((cr0 ^ old_cr0) & update_bits)
>  		kvm_mmu_reset_context(vcpu);
>  
> -	if ((cr0 ^ old_cr0) & X86_CR0_CD)
> +	if (((cr0 ^ old_cr0) & X86_CR0_CD) &&
> +	    kvm_arch_has_noncoherent_dma(vcpu->kvm) &&
> +	    !kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_CD_NW_CLEARED))
>  		kvm_zap_gfn_range(vcpu->kvm, 0, ~0ULL);
>  
>  	return 0;
> 

Applied, thanks.

Paolo