x86/vdso: Remove "memory" clobbers in the vDSO syscall fallbacks

When a vDSO clock function falls back to the syscall, no special
barriers or ordering is needed, and the syscall fallbacks don't
clobber any memory that is not explicitly listed in the asm
constraints.  Remove the "memory" clobber.

This causes minor changes to the generated code, but otherwise has
no obvious performance impact.  I think it's nice to have, though,
since it may help the optimizer in the future.

Signed-off-by: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/3a7438f5fb2422ed881683d2ccffd7f987b2dc44.1538689401.git.luto@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

arch/x86/entry/vdso/vclock_gettime.c

@@ -45,7 +45,7 @@ notrace static long vdso_fallback_gettime(long clock, struct timespec *ts)
 	long ret;
 	asm ("syscall" : "=a" (ret), "=m" (*ts) :
 	     "0" (__NR_clock_gettime), "D" (clock), "S" (ts) :
-	     "memory", "rcx", "r11");
+	     "rcx", "r11");
 	return ret;
 }
 
@@ -62,7 +62,7 @@ notrace static long vdso_fallback_gettime(long clock, struct timespec *ts)
 		"mov %%edx, %%ebx \n"
 		: "=a" (ret), "=m" (*ts)
 		: "0" (__NR_clock_gettime), [clock] "g" (clock), "c" (ts)
-		: "memory", "edx");
+		: "edx");
 	return ret;
 }