Skip to contentSkip to conversation

lightweight robust futexes: i386

PATCH Chuck Ebbert 2 participants 2 messages Feb 21, 2006

lore
Share & tools
lore
#0 lore raw

In-Reply-To: <[email protected]>

Quoted 11 lines · Ingo Molnar
include/asm-i386/futex.h +4 -1
@@ -107,7 +107,25 @@ futex_atomic_op_inuser (int encoded_op, 
 static inline int
 futex_atomic_cmpxchg_inuser(int __user *uaddr, int oldval, int newval)
 {
-     return -ENOSYS;
+     __asm__ __volatile__(
+             "1:     " LOCK_PREFIX "cmpxchgl %3, %1          \n"
+
+             "2:     .section .fixup, \"ax\"                 \n"

include/asm-i386/futex.h futex_atomic_op_inuser (int encoded_op, · line 123

Quoted 10 lines
include/asm-i386/futex.h +10 -0
+             "3:     mov     %2, %0                          \n"
+             "       jmp     2b                              \n"
+             "       .previous                               \n"
+
+             "       .section __ex_table, \"a\"              \n"
+             "       .align  8                               \n"
+             "       .long   1b,3b                           \n"
+             "       .previous                               \n"
+
+             : "=a" (oldval), "=m" (*uaddr)
                                 ^^^^

Should be "+m" because it's both read and written.

include/asm-i386/futex.h futex_atomic_op_inuser (int encoded_op, · line 125

Quoted 2 lines
include/asm-i386/futex.h +2 -0
+             : "i" (-EFAULT), "r" (newval), "0" (oldval)
+             : "memory"
                  ^^^^^^^^

Is this necessary? Every possible memory location that could be
affected has been listed in the operands if the above change is made.

Quoted 6 lines
include/asm-i386/futex.h +3 -0
+     );
+
+     return oldval;
 }
 
 #endif
Signature · 2 lines
Chuck
"Equations are the Devil's sentences."  --Stephen Colbert
#1 lore raw
Quoted 3 lines · Chuck Ebbert
Quoted 1 line

+ : "=a" (oldval), "=m" (*uaddr)

                                 ^^^^

Should be "+m" because it's both read and written.

No, this is why there is the "0" input parameter.

\> > + : "memory"

Quoted 3 lines
                  ^^^^^^^^

Is this necessary? Every possible memory location that could be
affected has been listed in the operands if the above change is made.

This makes the asm a barrier for the compiler.