[PATCH] x86/pvqspinlock: Use LOCK_PREFIX in __pv_queued_spin_unlock assembly code

[PATCH] x86/pvqspinlock: Use LOCK_PREFIX in __pv_queued_spin_unlock assembly code

[Waiman Long]

The LOCK_PREFIX macro should be used in the assembly
__raw_callee_save___pv_queued_spin_unlock code so that the lock prefix
can be patched out on UP systems.

Signed-off-by: Waiman Long <longman@redhat.com>
---
 arch/x86/include/asm/qspinlock_paravirt.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/include/asm/qspinlock_paravirt.h b/arch/x86/include/asm/qspinlock_paravirt.h
index 9ef5ee0..159622e 100644
--- a/arch/x86/include/asm/qspinlock_paravirt.h
+++ b/arch/x86/include/asm/qspinlock_paravirt.h
@@ -43,7 +43,7 @@
 	"push  %rdx;"
 	"mov   $0x1,%eax;"
 	"xor   %edx,%edx;"
-	"lock cmpxchg %dl,(%rdi);"
+	LOCK_PREFIX "cmpxchg %dl,(%rdi);"
 	"cmp   $0x1,%al;"
 	"jne   .slowpath;"
 	"pop   %rdx;"
-- 
1.8.3.1

Re: [PATCH] x86/pvqspinlock: Use LOCK_PREFIX in __pv_queued_spin_unlock assembly code

[Peter Zijlstra]

On Tue, Jul 17, 2018 at 04:16:00PM -0400, Waiman Long wrote:
> The LOCK_PREFIX macro should be used in the assembly
> __raw_callee_save___pv_queued_spin_unlock code so that the lock prefix
> can be patched out on UP systems.

True; but does paravirt locking realy make sense on UP guests?

> Signed-off-by: Waiman Long <longman@redhat.com>
> ---
>  arch/x86/include/asm/qspinlock_paravirt.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/arch/x86/include/asm/qspinlock_paravirt.h b/arch/x86/include/asm/qspinlock_paravirt.h
> index 9ef5ee0..159622e 100644
> --- a/arch/x86/include/asm/qspinlock_paravirt.h
> +++ b/arch/x86/include/asm/qspinlock_paravirt.h
> @@ -43,7 +43,7 @@
>  	"push  %rdx;"
>  	"mov   $0x1,%eax;"
>  	"xor   %edx,%edx;"
> -	"lock cmpxchg %dl,(%rdi);"
> +	LOCK_PREFIX "cmpxchg %dl,(%rdi);"
>  	"cmp   $0x1,%al;"
>  	"jne   .slowpath;"
>  	"pop   %rdx;"
> -- 
> 1.8.3.1
> 

Re: [PATCH] x86/pvqspinlock: Use LOCK_PREFIX in __pv_queued_spin_unlock assembly code

[Waiman Long]

On 07/18/2018 04:19 AM, Peter Zijlstra wrote:
> On Tue, Jul 17, 2018 at 04:16:00PM -0400, Waiman Long wrote:
>> The LOCK_PREFIX macro should be used in the assembly
>> __raw_callee_save___pv_queued_spin_unlock code so that the lock prefix
>> can be patched out on UP systems.
> True; but does paravirt locking realy make sense on UP guests?
>
Yes, it doesn't make sense. That is why I have another kvm patch that
turns off pvqspinlock for UP guest. However, I am not that familiar with
the Xen code. So I don't have a corresponding one for Xen.

Cheers,
Longman

[tip:locking/urgent] locking/pvqspinlock/x86: Use LOCK_PREFIX in __pv_queued_spin_unlock() assembly code

[tip-bot for Waiman Long]

Commit-ID:  c0dc373a780f4ec63e45a573b9551763abd8cd1a
Gitweb:     https://git.kernel.org/tip/c0dc373a780f4ec63e45a573b9551763abd8cd1a
Author:     Waiman Long <longman@redhat.com>
AuthorDate: Tue, 17 Jul 2018 16:16:00 -0400
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Wed, 25 Jul 2018 11:22:20 +0200

locking/pvqspinlock/x86: Use LOCK_PREFIX in __pv_queued_spin_unlock() assembly code

The LOCK_PREFIX macro should be used in the __raw_callee_save___pv_queued_spin_unlock()
assembly code, so that the lock prefix can be patched out on UP systems.

Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Joe Mario <jmario@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Link: http://lkml.kernel.org/r/1531858560-21547-1-git-send-email-longman@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 arch/x86/include/asm/qspinlock_paravirt.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/include/asm/qspinlock_paravirt.h b/arch/x86/include/asm/qspinlock_paravirt.h
index 9ef5ee03d2d7..159622ee0674 100644
--- a/arch/x86/include/asm/qspinlock_paravirt.h
+++ b/arch/x86/include/asm/qspinlock_paravirt.h
@@ -43,7 +43,7 @@ asm    (".pushsection .text;"
 	"push  %rdx;"
 	"mov   $0x1,%eax;"
 	"xor   %edx,%edx;"
-	"lock cmpxchg %dl,(%rdi);"
+	LOCK_PREFIX "cmpxchg %dl,(%rdi);"
 	"cmp   $0x1,%al;"
 	"jne   .slowpath;"
 	"pop   %rdx;"