[PATCH] kvm: x86: disable the narrow guest module parameter on unload

[PATCH] kvm: x86: disable the narrow guest module parameter on unload

[Aaron Lewis]

When the kvm_intel module unloads the module parameter
'allow_smaller_maxphyaddr' is not cleared because it is also used in the
kvm module.  As a result, if the module parameter's state was set before
kvm_intel unloads, it will also be set when it reloads.  Explicitly
clear the state in vmx_exit() to prevent this from happening.

Signed-off-by: Aaron Lewis <aaronlewis@google.com>
---
 arch/x86/kvm/vmx/vmx.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
index c2a779b688e6..fd161c9a83fd 100644
--- a/arch/x86/kvm/vmx/vmx.c
+++ b/arch/x86/kvm/vmx/vmx.c
@@ -7996,6 +7996,8 @@ static void vmx_exit(void)
 	}
 #endif
 	vmx_cleanup_l1d_flush();
+
+	allow_smaller_maxphyaddr = false;
 }
 module_exit(vmx_exit);
 
-- 
2.32.0.93.g670b81a890-goog

Re: [PATCH] kvm: x86: disable the narrow guest module parameter on unload

[Jim Mattson]

On Wed, Jun 23, 2021 at 1:35 PM Aaron Lewis <aaronlewis@google.com> wrote:
>
> When the kvm_intel module unloads the module parameter
> 'allow_smaller_maxphyaddr' is not cleared because it is also used in the

...because the backing variable is defined in the

(Or something like that.)

> kvm module.  As a result, if the module parameter's state was set before
> kvm_intel unloads, it will also be set when it reloads.  Explicitly
> clear the state in vmx_exit() to prevent this from happening.
>
> Signed-off-by: Aaron Lewis <aaronlewis@google.com>
> ---
>  arch/x86/kvm/vmx/vmx.c | 2 ++
>  1 file changed, 2 insertions(+)
>
> diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
> index c2a779b688e6..fd161c9a83fd 100644
> --- a/arch/x86/kvm/vmx/vmx.c
> +++ b/arch/x86/kvm/vmx/vmx.c
> @@ -7996,6 +7996,8 @@ static void vmx_exit(void)
>         }
>  #endif
>         vmx_cleanup_l1d_flush();
> +
> +       allow_smaller_maxphyaddr = false;
>  }
>  module_exit(vmx_exit);

This seems reasonable to me. Another option is to move the backing
variable to the kvm_intel module, given the recent suggestion that it
should never be enabled for AMD.

Reviewed-by: Jim Mattson <jmattson@google.com>

Re: [PATCH] kvm: x86: disable the narrow guest module parameter on unload

[Paolo Bonzini]

On 23/06/21 22:34, Aaron Lewis wrote:
> When the kvm_intel module unloads the module parameter
> 'allow_smaller_maxphyaddr' is not cleared because it is also used in the
> kvm module.  As a result, if the module parameter's state was set before
> kvm_intel unloads, it will also be set when it reloads.  Explicitly
> clear the state in vmx_exit() to prevent this from happening.
> 
> Signed-off-by: Aaron Lewis <aaronlewis@google.com>
> ---
>   arch/x86/kvm/vmx/vmx.c | 2 ++
>   1 file changed, 2 insertions(+)
> 
> diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
> index c2a779b688e6..fd161c9a83fd 100644
> --- a/arch/x86/kvm/vmx/vmx.c
> +++ b/arch/x86/kvm/vmx/vmx.c
> @@ -7996,6 +7996,8 @@ static void vmx_exit(void)
>   	}
>   #endif
>   	vmx_cleanup_l1d_flush();
> +
> +	allow_smaller_maxphyaddr = false;
>   }
>   module_exit(vmx_exit);
>   
> 

Queued, thanks.

Paolo

Re: [PATCH] kvm: x86: disable the narrow guest module parameter on unload

[Paolo Bonzini]

On 23/06/21 22:47, Jim Mattson wrote:
> This seems reasonable to me. Another option is to move the backing
> variable to the kvm_intel module, given the recent suggestion that it
> should never be enabled for AMD.

Why wouldn't it be enabled for shadow paging, since the issue with AMD 
is specific to the ordering of nested vs. guest page faults?

Paolo