[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

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


  • To: Teddy Astie <teddy.astie@xxxxxxxxxx>, xen-devel@xxxxxxxxxxxxxxxxxxxx, Jan Beulich <jbeulich@xxxxxxxx>
  • From: Grygorii Strashko <grygorii_strashko@xxxxxxxx>
  • Date: Thu, 6 Nov 2025 18:33:30 +0200
  • Arc-authentication-results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=epam.com; dmarc=pass action=none header.from=epam.com; dkim=pass header.d=epam.com; arc=none
  • Arc-message-signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector10001; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=VVPqj5q9UbtWaoRouyUelUFaGv0lx7vbxPctgZixcR8=; b=K2/4rcAy14c2bfb5EsEk9NTpeo8e+qS6PhZXArWajE7jpVxBZRvF1QxIpWLmvt5VfpMaaaHV+wS6NnhISV1GS/EdpfFIaGh5TFTdcDQSPa42qz3BMkpvNFA/APqO+/AUP32OOilyKX8+974BgWkLb9npgrZpBzRArQnhlDZOTj6tBBFOsOPAW/691MRf7elk7VJMgVb27hwpKHs/mHZeyXmsStEZhznqn85wSrC6U3qA1StOeAtBYDD6caJcs6gFV4t9u6kx02E3G9hlB8Uiirr8j1QGO4uwg/zBOiCnQNAt0dEobeCilmoVaOnd+Ci/kU5SgIQEE2wUctbWcqAAcw==
  • Arc-seal: i=1; a=rsa-sha256; s=arcselector10001; d=microsoft.com; cv=none; b=GwUoCVZl/08I50x6U9PQyV13vt98lHQ7aC21pEBN6kNI2Q8HIEe+4SVQCiwX25O6Wqeh5Mlv3HVxfiv95eYmQGGlUNsnL23U+/ehyVmPEaym5AKV1y6agRJAROapeVy7t402MHYeld//NGD+gTGFajEn2jA6v5vs0TEbJ9mhY+I52EkGo0B/9Z36P7o6MgkcZRLExKiF79hrEgD9txayuX+6oQpWV3UGebA6SIjh0OBCxLEzLixnIXdGkU91KPW0rlDB+Alul74I2CDd6czVevoAhK+f6E0WHIYrLwkesg2kq66P9OG3ANr8swsQyVnGFAoU2OPUBCtuxo4Fm/DmdA==
  • Authentication-results: dkim=none (message not signed) header.d=none;dmarc=none action=none header.from=epam.com;
  • Cc: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>, Roger Pau Monné <roger.pau@xxxxxxxxxx>, Anthony PERARD <anthony.perard@xxxxxxxxxx>, Michal Orzel <michal.orzel@xxxxxxx>, Julien Grall <julien@xxxxxxx>, Stefano Stabellini <sstabellini@xxxxxxxxxx>, Alejandro Vallejo <alejandro.garciavallejo@xxxxxxx>, Jason Andryuk <jason.andryuk@xxxxxxx>
  • Delivery-date: Thu, 06 Nov 2025 16:33:49 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>

Hi Teddy, Jan,

On 06.11.25 17:57, Teddy Astie wrote:
Le 31/10/2025 à 22:25, Grygorii Strashko a écrit :
From: Grygorii Strashko <grygorii_strashko@xxxxxxxx>

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))

How this pattern is working depends 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 required functions explicitly for each
combination of CONFIG_PV/CONFIG_HVM with main intention to optimize code for
(PV=n and HVM=y) case.

For the case (PV=n and HVM=n) empty stubs are created which return (1)
indicating failure. Hence, no guests should possible in this case -
which means no access to guest memory  should ever happen.
The two calls of __raw_copy_to_guest() in 
common/domain.c->update_runstate_area()
are fixed for this case by explicitly cast the return value to void
(MISRA C Rule 17.7).

Finally build arch/x86/usercopy.c only for PV=y.

The measured (bloat-o-meter) improvement for (PV=n and HVM=y) case is:
    add/remove: 0/10 grow/shrink: 2/90 up/down: 163/-30932 (-30769)
    Total: Before=1937113, After=1906344, chg -1.59%

Signed-off-by: Grygorii Strashko <grygorii_strashko@xxxxxxxx>
---
   xen/arch/x86/Makefile                   |  2 +-
   xen/arch/x86/include/asm/guest_access.h | 38 +++++++++++++++++++++++++
   xen/common/domain.c                     | 10 ++++---
   3 files changed, 45 insertions(+), 5 deletions(-)

diff --git a/xen/arch/x86/Makefile b/xen/arch/x86/Makefile
index 407571c510e1..27f131ffeb61 100644
--- a/xen/arch/x86/Makefile
+++ b/xen/arch/x86/Makefile
@@ -71,7 +71,7 @@ obj-y += time.o
   obj-y += traps-setup.o
   obj-y += traps.o
   obj-$(CONFIG_INTEL) += tsx.o
-obj-y += usercopy.o
+obj-$(CONFIG_PV) += usercopy.o
   obj-y += x86_emulate.o
   obj-$(CONFIG_TBOOT) += tboot.o
   obj-y += hpet.o
diff --git a/xen/arch/x86/include/asm/guest_access.h 
b/xen/arch/x86/include/asm/guest_access.h
index 69716c8b41bb..36aeb89524ab 100644
--- a/xen/arch/x86/include/asm/guest_access.h
+++ b/xen/arch/x86/include/asm/guest_access.h
@@ -13,6 +13,7 @@
   #include <asm/hvm/guest_access.h>
/* Raw access functions: no type checking. */
+#if defined(CONFIG_PV) && defined(CONFIG_HVM)
   #define raw_copy_to_guest(dst, src, len)        \
       (is_hvm_vcpu(current) ?                     \
        copy_to_user_hvm((dst), (src), (len)) :    \
@@ -34,6 +35,43 @@
        copy_from_user_hvm((dst), (src), (len)) :  \
        __copy_from_guest_pv(dst, src, len))
+#elif defined(CONFIG_HVM)
+#define raw_copy_to_guest(dst, src, len)        \
+     copy_to_user_hvm((dst), (src), (len))
+#define raw_copy_from_guest(dst, src, len)      \
+     copy_from_user_hvm((dst), (src), (len))
+#define raw_clear_guest(dst,  len)              \
+     clear_user_hvm((dst), (len))
+#define __raw_copy_to_guest(dst, src, len)      \
+     copy_to_user_hvm((dst), (src), (len))
+#define __raw_copy_from_guest(dst, src, len)    \
+     copy_from_user_hvm((dst), (src), (len))
+
+#elif defined(CONFIG_PV)
+#define raw_copy_to_guest(dst, src, len)        \
+     copy_to_guest_pv(dst, src, len)
+#define raw_copy_from_guest(dst, src, len)      \
+     copy_from_guest_pv(dst, src, len)
+#define raw_clear_guest(dst,  len)              \
+     clear_guest_pv(dst, len)
+#define __raw_copy_to_guest(dst, src, len)      \
+     __copy_to_guest_pv(dst, src, len)
+#define __raw_copy_from_guest(dst, src, len)    \
+     __copy_from_guest_pv(dst, src, len)
+
+#else
+#define raw_copy_to_guest(dst, src, len)        \
+        ((void)(dst), (void)(src), (void)(len), 1)
+#define raw_copy_from_guest(dst, src, len)      \
+        ((void)(dst), (void)(src), (void)(len), 1)
+#define raw_clear_guest(dst, len)               \
+        ((void)(dst), (void)(len), 1)
+#define __raw_copy_to_guest(dst, src, len)      \
+        ((void)(dst), (void)(src), (void)(len), 1)
+#define __raw_copy_from_guest(dst, src, len)    \
+        ((void)(dst), (void)(src), (void)(len), 1)
+#endif
+
   /*
    * Pre-validate a guest handle.
    * Allows use of faster __copy_* functions.
diff --git a/xen/common/domain.c b/xen/common/domain.c
index 4f91316ad93e..c603edcc7d46 100644
--- a/xen/common/domain.c
+++ b/xen/common/domain.c
@@ -1985,8 +1985,9 @@ bool update_runstate_area(struct vcpu *v)
   #endif
           guest_handle--;
           runstate.state_entry_time |= XEN_RUNSTATE_UPDATE;
-        __raw_copy_to_guest(guest_handle,
-                            (void *)(&runstate.state_entry_time + 1) - 1, 1);
+        (void)__raw_copy_to_guest(guest_handle,
+                                  (void *)(&runstate.state_entry_time + 1) - 1,
+                                  1);
           smp_wmb();
       }
@@ -2008,8 +2009,9 @@ bool update_runstate_area(struct vcpu *v)
       {
           runstate.state_entry_time &= ~XEN_RUNSTATE_UPDATE;
           smp_wmb();
-        __raw_copy_to_guest(guest_handle,
-                            (void *)(&runstate.state_entry_time + 1) - 1, 1);
+        (void)__raw_copy_to_guest(guest_handle,
+                                  (void *)(&runstate.state_entry_time + 1) - 1,
+                                  1);
       }
update_guest_memory_policy(v, &policy);

Alternatively, we can make all the raw_* functions `static inline` and
have something like this which should have the same effect with much
less redundancy.

static inline
unsigned int raw_copy_to_user_hvm(void *to, const void *from,
                                    unsigned int len)
{
      if ( IS_ENABLED(CONFIG_HVM) &&
           (!IS_ENABLED(CONFIG_PV) || is_hvm_vcpu(current) )
         copy_to_user_hvm(to, from, len);
      else if ( IS_ENABLED(CONFIG_PV) )
         copy_to_guest_pv(to, from, len);
        else
           return len;
}

Can try.

Jan, would it be acceptable?


--
Best regards,
-grygorii




 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.