From: Borislav Petkov <[email protected]>
There's no need to issue that everytime during boot - we have the
/proc/cpuinfo flag for people and software to query.
Signed-off-by: Borislav Petkov <[email protected]>
Cc: Janakarajan Natarajan <[email protected]>
Cc: Paolo Bonzini <[email protected]>
Cc: Radim Krčmář <[email protected]>
---
arch/x86/kvm/svm.c | 5 +----
1 file changed, 1 insertion(+), 4 deletions(-)
diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
index 8dbd8dbc83eb..f25e5b930932 100644
--- a/arch/x86/kvm/svm.c
+++ b/arch/x86/kvm/svm.c
@@ -1101,11 +1101,8 @@ static __init int svm_hardware_setup(void)
if (vls) {
if (!npt_enabled ||
!boot_cpu_has(X86_FEATURE_V_VMSAVE_VMLOAD) ||
- !IS_ENABLED(CONFIG_X86_64)) {
+ !IS_ENABLED(CONFIG_X86_64))
vls = false;
- } else {
- pr_info("Virtual VMLOAD VMSAVE supported\n");
- }
}
return 0;
--
2.13.0
On 05.09.2017 19:06, Borislav Petkov wrote:
> From: Borislav Petkov <[email protected]>
>
> There's no need to issue that everytime during boot - we have the
> /proc/cpuinfo flag for people and software to query.
>
> Signed-off-by: Borislav Petkov <[email protected]>
> Cc: Janakarajan Natarajan <[email protected]>
> Cc: Paolo Bonzini <[email protected]>
> Cc: Radim Krčmář <[email protected]>
> ---
> arch/x86/kvm/svm.c | 5 +----
> 1 file changed, 1 insertion(+), 4 deletions(-)
>
> diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
> index 8dbd8dbc83eb..f25e5b930932 100644
> --- a/arch/x86/kvm/svm.c
> +++ b/arch/x86/kvm/svm.c
> @@ -1101,11 +1101,8 @@ static __init int svm_hardware_setup(void)
> if (vls) {
> if (!npt_enabled ||
> !boot_cpu_has(X86_FEATURE_V_VMSAVE_VMLOAD) ||
> - !IS_ENABLED(CONFIG_X86_64)) {
> + !IS_ENABLED(CONFIG_X86_64))
> vls = false;
> - } else {
> - pr_info("Virtual VMLOAD VMSAVE supported\n");
> - }
> }
>
> return 0;
>
Maybe go one step further and incorporate everything (+vls) into a
single if statement?
--
Thanks,
David
On Wed, Sep 06, 2017 at 02:54:05PM +0200, David Hildenbrand wrote:
> Maybe go one step further and incorporate everything (+vls) into a
> single if statement?
Or maybe simplify it even more by not even looking at vls. If the user
disables it, fine, if she enables it but the hw doesn't support it, it
will be set to false automatically.
Or am I missing a case?
---
From: Borislav Petkov <[email protected]>
Date: Tue, 5 Sep 2017 18:59:55 +0200
Subject: [PATCH] KVM: SVM: Do not issue virtual VMLOAD/VMSAVE
supported-message
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
There's no need to issue that everytime during boot - we have the
/proc/cpuinfo flag for people and software to query.
Also, simplify logic which verifies the vls chicken bit setting.
Signed-off-by: Borislav Petkov <[email protected]>
Cc: David Hildenbrand <[email protected]>
Cc: Janakarajan Natarajan <[email protected]>
Cc: Paolo Bonzini <[email protected]>
Cc: Radim Krčmář <[email protected]>
---
arch/x86/kvm/svm.c | 13 ++++---------
1 file changed, 4 insertions(+), 9 deletions(-)
diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
index 8dbd8dbc83eb..d3c481778d9c 100644
--- a/arch/x86/kvm/svm.c
+++ b/arch/x86/kvm/svm.c
@@ -1098,15 +1098,10 @@ static __init int svm_hardware_setup(void)
}
}
- if (vls) {
- if (!npt_enabled ||
- !boot_cpu_has(X86_FEATURE_V_VMSAVE_VMLOAD) ||
- !IS_ENABLED(CONFIG_X86_64)) {
- vls = false;
- } else {
- pr_info("Virtual VMLOAD VMSAVE supported\n");
- }
- }
+ if (!npt_enabled ||
+ !boot_cpu_has(X86_FEATURE_V_VMSAVE_VMLOAD) ||
+ !IS_ENABLED(CONFIG_X86_64))
+ vls = false;
return 0;
--
2.13.0
--
Regards/Gruss,
Boris.
Good mailing practices for 400: avoid top-posting and trim the reply.
On 06.09.2017 15:17, Borislav Petkov wrote:
> On Wed, Sep 06, 2017 at 02:54:05PM +0200, David Hildenbrand wrote:
>> Maybe go one step further and incorporate everything (+vls) into a
>> single if statement?
>
> Or maybe simplify it even more by not even looking at vls. If the user
> disables it, fine, if she enables it but the hw doesn't support it, it
> will be set to false automatically.
>
> Or am I missing a case?
>
> ---
> From: Borislav Petkov <[email protected]>
> Date: Tue, 5 Sep 2017 18:59:55 +0200
> Subject: [PATCH] KVM: SVM: Do not issue virtual VMLOAD/VMSAVE
> supported-message
> MIME-Version: 1.0
> Content-Type: text/plain; charset=UTF-8
> Content-Transfer-Encoding: 8bit
>
> There's no need to issue that everytime during boot - we have the
> /proc/cpuinfo flag for people and software to query.
>
> Also, simplify logic which verifies the vls chicken bit setting.
>
> Signed-off-by: Borislav Petkov <[email protected]>
> Cc: David Hildenbrand <[email protected]>
> Cc: Janakarajan Natarajan <[email protected]>
> Cc: Paolo Bonzini <[email protected]>
> Cc: Radim Krčmář <[email protected]>
> ---
> arch/x86/kvm/svm.c | 13 ++++---------
> 1 file changed, 4 insertions(+), 9 deletions(-)
>
> diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
> index 8dbd8dbc83eb..d3c481778d9c 100644
> --- a/arch/x86/kvm/svm.c
> +++ b/arch/x86/kvm/svm.c
> @@ -1098,15 +1098,10 @@ static __init int svm_hardware_setup(void)
> }
> }
>
> - if (vls) {
> - if (!npt_enabled ||
> - !boot_cpu_has(X86_FEATURE_V_VMSAVE_VMLOAD) ||
> - !IS_ENABLED(CONFIG_X86_64)) {
> - vls = false;
> - } else {
> - pr_info("Virtual VMLOAD VMSAVE supported\n");
> - }
> - }
> + if (!npt_enabled ||
> + !boot_cpu_has(X86_FEATURE_V_VMSAVE_VMLOAD) ||
> + !IS_ENABLED(CONFIG_X86_64))
> + vls = false;
>
> return 0;
>
>
had the same idea but was worried about runtime. but this is really
only executed once, so
Reviewed-by: David Hildenbrand <[email protected]>
--
Thanks,
David
On Wed, Sep 06, 2017 at 04:00:04PM +0200, David Hildenbrand wrote:
> had the same idea but was worried about runtime. but this is really
> only executed once, so
Yap, exactly. Forgot to say ... but I was thinking of saying it. :)
--
Regards/Gruss,
Boris.
Good mailing practices for 400: avoid top-posting and trim the reply.
2017-09-05 19:06+0200, Borislav Petkov:
> From: Borislav Petkov <[email protected]>
>
> There's no need to issue that everytime during boot - we have the
> /proc/cpuinfo flag for people and software to query.
>
> Signed-off-by: Borislav Petkov <[email protected]>
> Cc: Janakarajan Natarajan <[email protected]>
> Cc: Paolo Bonzini <[email protected]>
> Cc: Radim Krčmář <[email protected]>
> ---
> arch/x86/kvm/svm.c | 5 +----
> 1 file changed, 1 insertion(+), 4 deletions(-)
>
> diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
> index 8dbd8dbc83eb..f25e5b930932 100644
> --- a/arch/x86/kvm/svm.c
> +++ b/arch/x86/kvm/svm.c
> @@ -1101,11 +1101,8 @@ static __init int svm_hardware_setup(void)
> if (vls) {
> if (!npt_enabled ||
> !boot_cpu_has(X86_FEATURE_V_VMSAVE_VMLOAD) ||
> - !IS_ENABLED(CONFIG_X86_64)) {
> + !IS_ENABLED(CONFIG_X86_64))
> vls = false;
> - } else {
> - pr_info("Virtual VMLOAD VMSAVE supported\n");
SVM prints more offending messages on boot, so I'd do all one go.
I am thinking of collocating the info instead as the information might
prevent one round of asking for SVM features.
---8<---
Subject: [PATCH] KVM: SVM: print enabled features on one line
The information might be useful for debugging, but having each feature
present itself is needlessly verbose.
Signed-off-by: Radim Krčmář <[email protected]>
---
arch/x86/kvm/svm.c | 56 ++++++++++++++++++------------------------------------
1 file changed, 19 insertions(+), 37 deletions(-)
diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
index af54327b9017..1901815777d0 100644
--- a/arch/x86/kvm/svm.c
+++ b/arch/x86/kvm/svm.c
@@ -1078,10 +1078,8 @@ static __init int svm_hardware_setup(void)
kvm_tsc_scaling_ratio_frac_bits = 32;
}
- if (nested) {
- printk(KERN_INFO "kvm: Nested Virtualization enabled\n");
+ if (nested)
kvm_enable_efer_bits(EFER_SVME | EFER_LMSLE);
- }
for_each_possible_cpu(cpu) {
r = svm_cpu_init(cpu);
@@ -1089,48 +1087,32 @@ static __init int svm_hardware_setup(void)
goto err;
}
- if (!boot_cpu_has(X86_FEATURE_NPT))
- npt_enabled = false;
-
- if (npt_enabled && !npt) {
- printk(KERN_INFO "kvm: Nested Paging disabled\n");
+ if (!boot_cpu_has(X86_FEATURE_NPT) || !npt)
npt_enabled = false;
- }
- if (npt_enabled) {
- printk(KERN_INFO "kvm: Nested Paging enabled\n");
+ if (npt_enabled)
kvm_enable_tdp();
- } else
+ else
kvm_disable_tdp();
- if (avic) {
- if (!npt_enabled ||
- !boot_cpu_has(X86_FEATURE_AVIC) ||
- !IS_ENABLED(CONFIG_X86_LOCAL_APIC)) {
- avic = false;
- } else {
- pr_info("AVIC enabled\n");
+ if (!npt_enabled ||
+ !boot_cpu_has(X86_FEATURE_AVIC) ||
+ !IS_ENABLED(CONFIG_X86_LOCAL_APIC))
+ avic = false;
- amd_iommu_register_ga_log_notifier(&avic_ga_log_notifier);
- }
- }
+ if (avic)
+ amd_iommu_register_ga_log_notifier(&avic_ga_log_notifier);
- if (vls) {
- if (!npt_enabled ||
- !boot_cpu_has(X86_FEATURE_V_VMSAVE_VMLOAD) ||
- !IS_ENABLED(CONFIG_X86_64)) {
- vls = false;
- } else {
- pr_info("Virtual VMLOAD VMSAVE supported\n");
- }
- }
+ if (!npt_enabled ||
+ !boot_cpu_has(X86_FEATURE_V_VMSAVE_VMLOAD) ||
+ !IS_ENABLED(CONFIG_X86_64))
+ vls = false;
- if (vgif) {
- if (!boot_cpu_has(X86_FEATURE_VGIF))
- vgif = false;
- else
- pr_info("Virtual GIF supported\n");
- }
+ if (!boot_cpu_has(X86_FEATURE_VGIF))
+ vgif = false;
+
+ pr_info("SVM nested=%d npt=%d avic=%d vls=%d vgif=%d\n",
+ nested, npt_enabled, avic, vls, vgif);
return 0;
--
2.14.1
On 13/09/2017 16:06, Radim Krčmář wrote:
> - }
> + if (!boot_cpu_has(X86_FEATURE_VGIF))
> + vgif = false;
> +
> + pr_info("SVM nested=%d npt=%d avic=%d vls=%d vgif=%d\n",
> + nested, npt_enabled, avic, vls, vgif);
I doubt we'd get the above info message more than sporadically, and all
these are available with a simple
grep -r . /sys/module/kvm_amd/parameters
which we already do routinely ask bug reporters for the equivalent
kvm_intel. So I'd say, zap it.
Paolo