[XEN][PATCH v5] xen/x86: guest_access: optimize raw_x_guest() for PV and HVM combinations

Grygorii Strashko posted 1 patch 1 day, 1 hour ago
xen/arch/x86/include/asm/guest_access.h | 78 ++++++++++++++++++-------
1 file changed, 58 insertions(+), 20 deletions(-)
[XEN][PATCH v5] xen/x86: guest_access: optimize raw_x_guest() for PV and HVM combinations
Posted by Grygorii Strashko 1 day, 1 hour ago
From: Grygorii Strashko <grygorii_strashko@epam.com>

Xen uses below pattern for raw_x_guest() functions:

define raw_copy_to_guest(dst, src, len)        \
    (is_hvm_vcpu(current) ?                     \
     copy_to_user_hvm((dst), (src), (len)) :    \
     copy_to_guest_pv(dst, src, len))

This pattern works depending on CONFIG_PV/CONFIG_HVM as:
- PV=y and HVM=y
  Proper guest access function is selected depending on domain type.
- PV=y and HVM=n
  Only PV domains are possible. is_hvm_domain/vcpu() will constify to "false"
  and compiler will optimize code and skip HVM specific part.
- PV=n and HVM=y
  Only HVM domains are possible. is_hvm_domain/vcpu() will not be constified.
  No PV specific code will be optimized by compiler.
- PV=n and HVM=n
  No guests should possible. The code will still follow PV path.

Rework raw_x_guest() code to use static inline functions which account for
above PV/HVM possible configurations with main intention to optimize code
for (PV=n and HVM=y) case.

For the case (PV=n and HVM=n) return "len" value indicating a failure (no
guests should be possible in this case, which means no access to guest
memory should ever happen).

The measured (bloat-o-meter) improvement for (PV=n and HVM=y) case is about -11K.

[teddy.astie@vates.tech: Suggested to use static inline functions vs
macro combinations]
Suggested-by: Teddy Astie <teddy.astie@vates.tech>
Signed-off-by: Grygorii Strashko <grygorii_strashko@epam.com>
---
changes in v5:
- rebase
- drop moving usercopy.c as it is not needed since commit 7370966d1cb7 
  ("x86: move / split usercopy.c to / into arch-specific library")

changes in v4:
- move usercopy.c into arch/x86/pv/
- rework to always dynamically check for HVM vcpu(domain) by using is_hvm_vcpu()
  as requested by Jan Beulich

changes in v3:
- add raw_use_hvm_access() wrapper

changes in v2:
- use static inline functions instead of macro combinations

 xen/arch/x86/include/asm/guest_access.h | 78 ++++++++++++++++++-------
 1 file changed, 58 insertions(+), 20 deletions(-)

diff --git a/xen/arch/x86/include/asm/guest_access.h b/xen/arch/x86/include/asm/guest_access.h
index 69716c8b41bb..f0e56b112e14 100644
--- a/xen/arch/x86/include/asm/guest_access.h
+++ b/xen/arch/x86/include/asm/guest_access.h
@@ -13,26 +13,64 @@
 #include <asm/hvm/guest_access.h>
 
 /* Raw access functions: no type checking. */
-#define raw_copy_to_guest(dst, src, len)        \
-    (is_hvm_vcpu(current) ?                     \
-     copy_to_user_hvm((dst), (src), (len)) :    \
-     copy_to_guest_pv(dst, src, len))
-#define raw_copy_from_guest(dst, src, len)      \
-    (is_hvm_vcpu(current) ?                     \
-     copy_from_user_hvm((dst), (src), (len)) :  \
-     copy_from_guest_pv(dst, src, len))
-#define raw_clear_guest(dst,  len)              \
-    (is_hvm_vcpu(current) ?                     \
-     clear_user_hvm((dst), (len)) :             \
-     clear_guest_pv(dst, len))
-#define __raw_copy_to_guest(dst, src, len)      \
-    (is_hvm_vcpu(current) ?                     \
-     copy_to_user_hvm((dst), (src), (len)) :    \
-     __copy_to_guest_pv(dst, src, len))
-#define __raw_copy_from_guest(dst, src, len)    \
-    (is_hvm_vcpu(current) ?                     \
-     copy_from_user_hvm((dst), (src), (len)) :  \
-     __copy_from_guest_pv(dst, src, len))
+static inline unsigned int raw_copy_to_guest(void *dst, const void *src,
+                                             unsigned int len)
+{
+    if ( is_hvm_vcpu(current) )
+        return copy_to_user_hvm(dst, src, len);
+
+    if ( !IS_ENABLED(CONFIG_PV) )
+        return len;
+
+    return copy_to_guest_pv(dst, src, len);
+}
+
+static inline unsigned int raw_copy_from_guest(void *dst, const void *src,
+                                               unsigned int len)
+{
+    if ( is_hvm_vcpu(current) )
+        return copy_from_user_hvm(dst, src, len);
+
+    if ( !IS_ENABLED(CONFIG_PV) )
+        return len;
+
+    return copy_from_guest_pv(dst, src, len);
+}
+
+static inline unsigned int raw_clear_guest(void *dst, unsigned int len)
+{
+    if ( is_hvm_vcpu(current) )
+        return clear_user_hvm(dst, len);
+
+    if ( !IS_ENABLED(CONFIG_PV) )
+        return len;
+
+    return clear_guest_pv(dst, len);
+}
+
+static inline unsigned int __raw_copy_to_guest(void *dst, const void *src,
+                                               unsigned int len)
+{
+    if ( is_hvm_vcpu(current) )
+        return copy_to_user_hvm(dst, src, len);
+
+    if ( !IS_ENABLED(CONFIG_PV) )
+        return len;
+
+    return __copy_to_guest_pv(dst, src, len);
+}
+
+static inline unsigned int __raw_copy_from_guest(void *dst, const void *src,
+                                                 unsigned int len)
+{
+    if ( is_hvm_vcpu(current) )
+        return copy_from_user_hvm(dst, src, len);
+
+    if ( !IS_ENABLED(CONFIG_PV) )
+        return len;
+
+    return __copy_from_guest_pv(dst, src, len);
+}
 
 /*
  * Pre-validate a guest handle.
-- 
2.34.1
Re: [XEN][PATCH v5] xen/x86: guest_access: optimize raw_x_guest() for PV and HVM combinations
Posted by Jan Beulich 1 day ago
On 18.12.2025 14:59, Grygorii Strashko wrote:
> From: Grygorii Strashko <grygorii_strashko@epam.com>
> 
> Xen uses below pattern for raw_x_guest() functions:
> 
> define raw_copy_to_guest(dst, src, len)        \
>     (is_hvm_vcpu(current) ?                     \
>      copy_to_user_hvm((dst), (src), (len)) :    \
>      copy_to_guest_pv(dst, src, len))
> 
> This pattern works depending on CONFIG_PV/CONFIG_HVM as:
> - PV=y and HVM=y
>   Proper guest access function is selected depending on domain type.
> - PV=y and HVM=n
>   Only PV domains are possible. is_hvm_domain/vcpu() will constify to "false"
>   and compiler will optimize code and skip HVM specific part.
> - PV=n and HVM=y
>   Only HVM domains are possible. is_hvm_domain/vcpu() will not be constified.
>   No PV specific code will be optimized by compiler.
> - PV=n and HVM=n
>   No guests should possible. The code will still follow PV path.
> 
> Rework raw_x_guest() code to use static inline functions which account for
> above PV/HVM possible configurations with main intention to optimize code
> for (PV=n and HVM=y) case.
> 
> For the case (PV=n and HVM=n) return "len" value indicating a failure (no
> guests should be possible in this case, which means no access to guest
> memory should ever happen).

I agree with Teddy's sentiment towards HVM={y,n} not really mattering when
PV=n, as far as non-HVM domains go.

> --- a/xen/arch/x86/include/asm/guest_access.h
> +++ b/xen/arch/x86/include/asm/guest_access.h
> @@ -13,26 +13,64 @@
>  #include <asm/hvm/guest_access.h>
>  
>  /* Raw access functions: no type checking. */
> -#define raw_copy_to_guest(dst, src, len)        \
> -    (is_hvm_vcpu(current) ?                     \
> -     copy_to_user_hvm((dst), (src), (len)) :    \
> -     copy_to_guest_pv(dst, src, len))
> -#define raw_copy_from_guest(dst, src, len)      \
> -    (is_hvm_vcpu(current) ?                     \
> -     copy_from_user_hvm((dst), (src), (len)) :  \
> -     copy_from_guest_pv(dst, src, len))
> -#define raw_clear_guest(dst,  len)              \
> -    (is_hvm_vcpu(current) ?                     \
> -     clear_user_hvm((dst), (len)) :             \
> -     clear_guest_pv(dst, len))
> -#define __raw_copy_to_guest(dst, src, len)      \
> -    (is_hvm_vcpu(current) ?                     \
> -     copy_to_user_hvm((dst), (src), (len)) :    \
> -     __copy_to_guest_pv(dst, src, len))
> -#define __raw_copy_from_guest(dst, src, len)    \
> -    (is_hvm_vcpu(current) ?                     \
> -     copy_from_user_hvm((dst), (src), (len)) :  \
> -     __copy_from_guest_pv(dst, src, len))
> +static inline unsigned int raw_copy_to_guest(void *dst, const void *src,
> +                                             unsigned int len)

A side effect of the converting to inline functions, besides being more
intrusive, is that now you will want to add proper __user modifiers.
See lib/copy-guest.c's use of them. That said, ..._user_hvm() functions
also don't have them, but imo wrongly so. Really I question the use of
pointers in that case, because they "point" into a different address
space, entirely inaccessible via use of those pointers. Hence adding
__user is going to be only a marginal improvement for the HVM case, but
is going to be wanted for the PV side of things.

Jan
Re: [XEN][PATCH v5] xen/x86: guest_access: optimize raw_x_guest() for PV and HVM combinations
Posted by Grygorii Strashko 1 day ago

On 18.12.25 17:22, Jan Beulich wrote:
> On 18.12.2025 14:59, Grygorii Strashko wrote:
>> From: Grygorii Strashko <grygorii_strashko@epam.com>
>>
>> Xen uses below pattern for raw_x_guest() functions:
>>
>> define raw_copy_to_guest(dst, src, len)        \
>>      (is_hvm_vcpu(current) ?                     \
>>       copy_to_user_hvm((dst), (src), (len)) :    \
>>       copy_to_guest_pv(dst, src, len))
>>
>> This pattern works depending on CONFIG_PV/CONFIG_HVM as:
>> - PV=y and HVM=y
>>    Proper guest access function is selected depending on domain type.
>> - PV=y and HVM=n
>>    Only PV domains are possible. is_hvm_domain/vcpu() will constify to "false"
>>    and compiler will optimize code and skip HVM specific part.
>> - PV=n and HVM=y
>>    Only HVM domains are possible. is_hvm_domain/vcpu() will not be constified.
>>    No PV specific code will be optimized by compiler.
>> - PV=n and HVM=n
>>    No guests should possible. The code will still follow PV path.
>>
>> Rework raw_x_guest() code to use static inline functions which account for
>> above PV/HVM possible configurations with main intention to optimize code
>> for (PV=n and HVM=y) case.
>>
>> For the case (PV=n and HVM=n) return "len" value indicating a failure (no
>> guests should be possible in this case, which means no access to guest
>> memory should ever happen).
> 
> I agree with Teddy's sentiment towards HVM={y,n} not really mattering when
> PV=n, as far as non-HVM domains go.
> 
>> --- a/xen/arch/x86/include/asm/guest_access.h
>> +++ b/xen/arch/x86/include/asm/guest_access.h
>> @@ -13,26 +13,64 @@
>>   #include <asm/hvm/guest_access.h>
>>   
>>   /* Raw access functions: no type checking. */
>> -#define raw_copy_to_guest(dst, src, len)        \
>> -    (is_hvm_vcpu(current) ?                     \
>> -     copy_to_user_hvm((dst), (src), (len)) :    \
>> -     copy_to_guest_pv(dst, src, len))
>> -#define raw_copy_from_guest(dst, src, len)      \
>> -    (is_hvm_vcpu(current) ?                     \
>> -     copy_from_user_hvm((dst), (src), (len)) :  \
>> -     copy_from_guest_pv(dst, src, len))
>> -#define raw_clear_guest(dst,  len)              \
>> -    (is_hvm_vcpu(current) ?                     \
>> -     clear_user_hvm((dst), (len)) :             \
>> -     clear_guest_pv(dst, len))
>> -#define __raw_copy_to_guest(dst, src, len)      \
>> -    (is_hvm_vcpu(current) ?                     \
>> -     copy_to_user_hvm((dst), (src), (len)) :    \
>> -     __copy_to_guest_pv(dst, src, len))
>> -#define __raw_copy_from_guest(dst, src, len)    \
>> -    (is_hvm_vcpu(current) ?                     \
>> -     copy_from_user_hvm((dst), (src), (len)) :  \
>> -     __copy_from_guest_pv(dst, src, len))
>> +static inline unsigned int raw_copy_to_guest(void *dst, const void *src,
>> +                                             unsigned int len)
> 
> A side effect of the converting to inline functions, besides being more
> intrusive, is that now you will want to add proper __user modifiers.
> See lib/copy-guest.c's use of them. That said, ..._user_hvm() functions
> also don't have them, but imo wrongly so. Really I question the use of
> pointers in that case, because they "point" into a different address
> space, entirely inaccessible via use of those pointers. Hence adding
> __user is going to be only a marginal improvement for the HVM case, but
> is going to be wanted for the PV side of things.

ok. so it need to be like this in all funcs

-static inline unsigned int raw_copy_to_guest(void *dst, const void *src,
+static inline unsigned int raw_copy_to_guest(void __user *dst, const void *src,
                                               unsigned int len)



-- 
Best regards,
-grygorii
Re: [XEN][PATCH v5] xen/x86: guest_access: optimize raw_x_guest() for PV and HVM combinations
Posted by Teddy Astie 1 day, 1 hour ago
Le 18/12/2025 à 15:01, Grygorii Strashko a écrit :
> From: Grygorii Strashko <grygorii_strashko@epam.com>
> 
> Xen uses below pattern for raw_x_guest() functions:
> 
> define raw_copy_to_guest(dst, src, len)        \
>      (is_hvm_vcpu(current) ?                     \
>       copy_to_user_hvm((dst), (src), (len)) :    \
>       copy_to_guest_pv(dst, src, len))
> 
> This pattern works depending on CONFIG_PV/CONFIG_HVM as:
> - PV=y and HVM=y
>    Proper guest access function is selected depending on domain type.
> - PV=y and HVM=n
>    Only PV domains are possible. is_hvm_domain/vcpu() will constify to "false"
>    and compiler will optimize code and skip HVM specific part.

> - PV=n and HVM=y
>    Only HVM domains are possible. is_hvm_domain/vcpu() will not be constified.
>    No PV specific code will be optimized by compiler.
> - PV=n and HVM=n
>    No guests should possible. The code will still follow PV path.
> 

^ regarding this

> For the case (PV=n and HVM=n) return "len" value indicating a failure (no
> guests should be possible in this case, which means no access to guest
> memory should ever happen).
> 

^ and this

AFAIU it is the same for PV=n and HVM=y for non-HVM domains (the few 
that "exists"), so we should rather say that these functions fails on 
"non-HVM" domains in PV=n configurations (since no actual PV domain 
exists in these cases).

IOW, there is no PV path in PV=n configurations and in !HVM domains, the 
function fails instead (as we would expect).

Once the commit message is adjusted on this
Reviewed-by: Teddy Astie <teddy.astie@vates.tech>

> The measured (bloat-o-meter) improvement for (PV=n and HVM=y) case is about -11K.
> 
> [teddy.astie@vates.tech: Suggested to use static inline functions vs
> macro combinations]
> Suggested-by: Teddy Astie <teddy.astie@vates.tech>
> Signed-off-by: Grygorii Strashko <grygorii_strashko@epam.com>
> ---
> changes in v5:
> - rebase
> - drop moving usercopy.c as it is not needed since commit 7370966d1cb7
>    ("x86: move / split usercopy.c to / into arch-specific library")
> 
> changes in v4:
> - move usercopy.c into arch/x86/pv/
> - rework to always dynamically check for HVM vcpu(domain) by using is_hvm_vcpu()
>    as requested by Jan Beulich
> 
> changes in v3:
> - add raw_use_hvm_access() wrapper
> 
> changes in v2:
> - use static inline functions instead of macro combinations
> 
>   xen/arch/x86/include/asm/guest_access.h | 78 ++++++++++++++++++-------
>   1 file changed, 58 insertions(+), 20 deletions(-)
> 
> diff --git a/xen/arch/x86/include/asm/guest_access.h b/xen/arch/x86/include/asm/guest_access.h
> index 69716c8b41bb..f0e56b112e14 100644
> --- a/xen/arch/x86/include/asm/guest_access.h
> +++ b/xen/arch/x86/include/asm/guest_access.h
> @@ -13,26 +13,64 @@
>   #include <asm/hvm/guest_access.h>
>   
>   /* Raw access functions: no type checking. */
> -#define raw_copy_to_guest(dst, src, len)        \
> -    (is_hvm_vcpu(current) ?                     \
> -     copy_to_user_hvm((dst), (src), (len)) :    \
> -     copy_to_guest_pv(dst, src, len))
> -#define raw_copy_from_guest(dst, src, len)      \
> -    (is_hvm_vcpu(current) ?                     \
> -     copy_from_user_hvm((dst), (src), (len)) :  \
> -     copy_from_guest_pv(dst, src, len))
> -#define raw_clear_guest(dst,  len)              \
> -    (is_hvm_vcpu(current) ?                     \
> -     clear_user_hvm((dst), (len)) :             \
> -     clear_guest_pv(dst, len))
> -#define __raw_copy_to_guest(dst, src, len)      \
> -    (is_hvm_vcpu(current) ?                     \
> -     copy_to_user_hvm((dst), (src), (len)) :    \
> -     __copy_to_guest_pv(dst, src, len))
> -#define __raw_copy_from_guest(dst, src, len)    \
> -    (is_hvm_vcpu(current) ?                     \
> -     copy_from_user_hvm((dst), (src), (len)) :  \
> -     __copy_from_guest_pv(dst, src, len))
> +static inline unsigned int raw_copy_to_guest(void *dst, const void *src,
> +                                             unsigned int len)
> +{
> +    if ( is_hvm_vcpu(current) )
> +        return copy_to_user_hvm(dst, src, len);
> +
> +    if ( !IS_ENABLED(CONFIG_PV) )
> +        return len;
> +
> +    return copy_to_guest_pv(dst, src, len);
> +}
> +
> +static inline unsigned int raw_copy_from_guest(void *dst, const void *src,
> +                                               unsigned int len)
> +{
> +    if ( is_hvm_vcpu(current) )
> +        return copy_from_user_hvm(dst, src, len);
> +
> +    if ( !IS_ENABLED(CONFIG_PV) )
> +        return len;
> +
> +    return copy_from_guest_pv(dst, src, len);
> +}
> +
> +static inline unsigned int raw_clear_guest(void *dst, unsigned int len)
> +{
> +    if ( is_hvm_vcpu(current) )
> +        return clear_user_hvm(dst, len);
> +
> +    if ( !IS_ENABLED(CONFIG_PV) )
> +        return len;
> +
> +    return clear_guest_pv(dst, len);
> +}
> +
> +static inline unsigned int __raw_copy_to_guest(void *dst, const void *src,
> +                                               unsigned int len)
> +{
> +    if ( is_hvm_vcpu(current) )
> +        return copy_to_user_hvm(dst, src, len);
> +
> +    if ( !IS_ENABLED(CONFIG_PV) )
> +        return len;
> +
> +    return __copy_to_guest_pv(dst, src, len);
> +}
> +
> +static inline unsigned int __raw_copy_from_guest(void *dst, const void *src,
> +                                                 unsigned int len)
> +{
> +    if ( is_hvm_vcpu(current) )
> +        return copy_from_user_hvm(dst, src, len);
> +
> +    if ( !IS_ENABLED(CONFIG_PV) )
> +        return len;
> +
> +    return __copy_from_guest_pv(dst, src, len);
> +}
>   
>   /*
>    * Pre-validate a guest handle.

Teddy


--
Teddy Astie | Vates XCP-ng Developer

XCP-ng & Xen Orchestra - Vates solutions

web: https://vates.tech