parisc: Switch to more fine grained lws locks

Increase the number of lws locks to 256 entries (instead of 16) and
choose lock entry based on bits 3-11 (instead of 4-7) of the relevant
address.  With this change we archieve more fine-grained locking in
futex syscalls and thus reduce the number of possible stalls.
Signed-off-by: John David Anglin <dave.anglin@bell.net>
Signed-off-by: Helge Deller <deller@gmx.de>

arch/parisc/include/asm/futex.h

@@ -16,7 +16,7 @@ static inline void
 _futex_spin_lock_irqsave(u32 __user *uaddr, unsigned long int *flags)
 {
 	extern u32 lws_lock_start[];
-	long index = ((long)uaddr & 0xf0) >> 2;
+	long index = ((long)uaddr & 0x3f8) >> 1;
 	arch_spinlock_t *s = (arch_spinlock_t *)&lws_lock_start[index];
 	local_irq_save(*flags);
 	arch_spin_lock(s);
@@ -26,7 +26,7 @@ static inline void
 _futex_spin_unlock_irqrestore(u32 __user *uaddr, unsigned long int *flags)
 {
 	extern u32 lws_lock_start[];
-	long index = ((long)uaddr & 0xf0) >> 2;
+	long index = ((long)uaddr & 0x3f8) >> 1;
 	arch_spinlock_t *s = (arch_spinlock_t *)&lws_lock_start[index];
 	arch_spin_unlock(s);
 	local_irq_restore(*flags);

arch/parisc/kernel/syscall.S

@@ -571,8 +571,8 @@ lws_compare_and_swap:
 	ldil	L%lws_lock_start, %r20
 	ldo	R%lws_lock_start(%r20), %r28
 
-	/* Extract four bits from r26 and hash lock (Bits 4-7) */
-	extru  %r26, 27, 4, %r20
+	/* Extract eight bits from r26 and hash lock (Bits 3-11) */
+	extru  %r26, 28, 8, %r20
 
 	/* Find lock to use, the hash is either one of 0 to
 	   15, multiplied by 16 (keep it 16-byte aligned)
@@ -761,8 +761,8 @@ cas2_lock_start:
 	ldil	L%lws_lock_start, %r20
 	ldo	R%lws_lock_start(%r20), %r28
 
-	/* Extract four bits from r26 and hash lock (Bits 4-7) */
-	extru  %r26, 27, 4, %r20
+	/* Extract eight bits from r26 and hash lock (Bits 3-11) */
+	extru  %r26, 28, 8, %r20
 
 	/* Find lock to use, the hash is either one of 0 to
 	   15, multiplied by 16 (keep it 16-byte aligned)
@@ -950,7 +950,7 @@ END(sys_call_table64)
 	.align	L1_CACHE_BYTES
 ENTRY(lws_lock_start)
 	/* lws locks */
-	.rept 16
+	.rept 256
 	/* Keep locks aligned at 16-bytes */
 	.word 1
 	.word 0