Merge with Linux 2.1.55. More bugfixes and goodies from my private

CVS archive.

1 files changed, 87 insertions, 100 deletions

diff --git a/include/asm-sparc64/spinlock.h b/include/asm-sparc64/spinlock.h
index cf2e51c71..65880b033 100644
--- a/include/asm-sparc64/spinlock.h
+++ b/include/asm-sparc64/spinlock.h
@@ -68,33 +68,35 @@ typedef struct { } rwlock_t;
 
 typedef unsigned char spinlock_t;
 #define SPIN_LOCK_UNLOCKED	0
-#define spin_lock_init(lock)	(*(lock) = 0)
-#define spin_unlock_wait(lock)	do { barrier(); } while(*(volatile spinlock_t *)lock)
+
+#define spin_lock_init(lock)	(*((unsigned char *)(lock)) = 0)
+
+#define spin_unlock_wait(lock)	\
+do {	membar("#LoadLoad");	\
+} while(*((volatile unsigned char *)lock))
 
 extern __inline__ void spin_lock(spinlock_t *lock)
 {
 	__asm__ __volatile__("
-1:	ldstub		[%0], %%g2
-	brz,pt		%%g2, 2f
-	 membar		#LoadLoad | #LoadStore
-	b,a		%%xcc, 3f
-2:
-	.text		2
-3:	ldub		[%0], %%g2
-4:	brnz,a,pt	%%g2, 4b
-	 ldub		[%0], %%g2
-	b,a		1b
+1:	ldstub		[%0], %%g7
+	brnz,pn		%%g7, 2f
+	 membar		#StoreLoad | #StoreStore
+	.subsection	2
+2:	ldub		[%0], %%g7
+	brnz,pt		%%g7, 2b
+	 membar		#LoadLoad
+	b,a,pt		%%xcc, 1b
 	.previous
 "	: /* no outputs */
 	: "r" (lock)
-	: "g2", "memory");
+	: "g7", "memory");
 }
 
 extern __inline__ int spin_trylock(spinlock_t *lock)
 {
 	unsigned int result;
 	__asm__ __volatile__("ldstub [%1], %0\n\t"
-			     "membar #LoadLoad | #LoadStore"
+			     "membar #StoreLoad | #StoreStore"
 			     : "=r" (result)
 			     : "r" (lock)
 			     : "memory");
@@ -104,7 +106,7 @@ extern __inline__ int spin_trylock(spinlock_t *lock)
 extern __inline__ void spin_unlock(spinlock_t *lock)
 {
 	__asm__ __volatile__("membar	#StoreStore | #LoadStore\n\t"
-			     "stb	%%g0, [%0]"
+			     "stb	%%g0, [%0]\n\t"
 			     : /* No outputs */
 			     : "r" (lock)
 			     : "memory");
@@ -114,20 +116,18 @@ extern __inline__ void spin_lock_irq(spinlock_t *lock)
 {
 	__asm__ __volatile__("
 	wrpr		%%g0, 15, %%pil
-1:	ldstub		[%0], %%g2
-	brz,pt		%%g2, 2f
-	 membar		#LoadLoad | #LoadStore
-	b,a		3f
-2:
-	.text		2
-3:	ldub		[%0], %%g2
-4:	brnz,a,pt	%%g2, 4b
-	 ldub		[%0], %%g2
-	b,a		1b
+1:	ldstub		[%0], %%g7
+	brnz,pn		%%g7, 2f
+	 membar		#StoreLoad | #StoreStore
+	.subsection	2
+2:	ldub		[%0], %%g7
+	brnz,pt		%%g7, 2b
+	 membar		#LoadLoad
+	b,a,pt		%%xcc, 1b
 	.previous
 "	: /* no outputs */
 	: "r" (lock)
-	: "g2", "memory");
+	: "g7", "memory");
 }
 
 extern __inline__ void spin_unlock_irq(spinlock_t *lock)
@@ -147,20 +147,18 @@ do {	register spinlock_t *lp asm("g1");			\
 	__asm__ __volatile__(					\
 	"\n	rdpr		%%pil, %0\n"			\
 	"	wrpr		%%g0, 15, %%pil\n"		\
-	"1:	ldstub		[%1], %%g2\n"			\
-	"	brz,pt		%%g2, 2f\n"			\
-	"	 membar		#LoadLoad | #LoadStore\n"	\
-	"	b,a		3f\n"				\
-	"2:\n"							\
-	"	.text		2\n"				\
-	"3:	ldub		[%1], %%g2\n"			\
-	"4:	brnz,a,pt	%%g2, 4b\n"			\
-	"	 ldub		[%1], %%g2\n"			\
-	"	b,a		1b\n"				\
+	"1:	ldstub		[%1], %%g7\n"			\
+	"	brnz,pn		%%g7, 2f\n"			\
+	"	 membar		#StoreLoad | #StoreStore\n"	\
+	"	.subsection	2\n"				\
+	"2:	ldub		[%1], %%g7\n"			\
+	"	brnz,pt		%%g7, 2b\n"			\
+	"	 membar		#LoadLoad\n"			\
+	"	b,a,pt		%%xcc, 1b\n"			\
 	"	.previous\n"					\
 	: "=&r" (flags)						\
 	: "r" (lp)						\
-	: "g2", "memory");					\
+	: "g7", "memory");					\
 } while(0)
 
 extern __inline__ void spin_unlock_irqrestore(spinlock_t *lock, unsigned long flags)
@@ -182,94 +180,83 @@ typedef unsigned long rwlock_t;
 extern __inline__ void read_lock(rwlock_t *rw)
 {
 	__asm__ __volatile__("
-	ldx		[%0], %%g2
-1:	brgez,pt	%%g2, 4f
-	 add		%%g2, 1, %%g3
-	b,a		2f
-4:	casx		[%0], %%g2, %%g3
-	cmp		%%g2, %%g3
-	bne,a,pn	%%xcc, 1b
-	 ldx		[%0], %%g2
-	membar		#LoadLoad | #LoadStore
-	.text		2
-2:	ldx		[%0], %%g2
-3:	brlz,a,pt	%%g2, 3b
-	 ldx		[%0], %%g2
-	b		4b
-	 add		%%g2, 1, %%g3
+1:	ldx		[%0], %%g5
+	brlz,pn		%%g5, 2f
+4:	 add		%%g5, 1, %%g7
+	casx		[%0], %%g5, %%g7
+	cmp		%%g5, %%g7
+	bne,pn		%%xcc, 1b
+	 membar		#StoreLoad | #StoreStore
+	.subsection	2
+2:	ldx		[%0], %%g5
+	brlz,pt		%%g5, 2b
+	 membar		#LoadLoad
+	b,a,pt		%%xcc, 4b
 	.previous
 "	: /* no outputs */
 	: "r" (rw)
-	: "g2", "g3", "cc", "memory");
+	: "g5", "g7", "cc", "memory");
 }
 
 extern __inline__ void read_unlock(rwlock_t *rw)
 {
 	__asm__ __volatile__("
-	membar		#StoreStore | #LoadStore
-	ldx		[%0], %%g2
-1:	sub		%%g2, 1, %%g3
-	casx		[%0], %%g2, %%g3
-	cmp		%%g2, %%g3
-	bne,a,pn	%%xcc, 1b
-	 ldx		[%0], %%g2
+1:	ldx		[%0], %%g5
+	sub		%%g5, 1, %%g7
+	casx		[%0], %%g5, %%g7
+	cmp		%%g5, %%g7
+	bne,pn		%%xcc, 1b
+	 membar		#StoreLoad | #StoreStore
 "	: /* no outputs */
 	: "r" (rw)
-	: "g2", "g3", "cc", "memory");
+	: "g5", "g7", "cc", "memory");
 }
 
 extern __inline__ void write_lock(rwlock_t *rw)
 {
 	__asm__ __volatile__("
-	sethi		%%uhi(0x8000000000000000), %%g5
-	ldx		[%0], %%g2
-	sllx		%%g5, 32, %%g5
-1:	brgez,pt	%%g2, 4f
-	 or		%%g2, %%g5, %%g3
-	b,a		5f
-4:	casx		[%0], %%g2, %%g3
-	cmp		%%g2, %%g3
-	bne,a,pn	%%xcc, 1b
-	 ldx		[%0], %%g2
-	andncc		%%g3, %%g5, %%g0
-	be,pt		%%xcc, 2f
-	 membar		#LoadLoad | #LoadStore
-	b,a		7f
-2:
-	.text		2
-7:	ldx		[%0], %%g2
-3:	andn		%%g2, %%g5, %%g3
-	casx		[%0], %%g2, %%g3
-	cmp		%%g2, %%g3
-	bne,a,pn	%%xcc, 3b
-	 ldx		[%0], %%g2
-	membar		#LoadLoad | #LoadStore
-5:	ldx		[%0], %%g2
-6:	brlz,a,pt	%%g2, 6b
-	 ldx		[%0], %%g2
-	b		4b
-	 or		%%g2, %%g5, %%g3
+	sethi		%%uhi(0x8000000000000000), %%g3
+	sllx		%%g3, 32, %%g3
+1:	ldx		[%0], %%g5
+	brlz,pn		%%g5, 5f
+4:	 or		%%g5, %%g3, %%g7
+	casx		[%0], %%g5, %%g7
+	cmp		%%g5, %%g7
+	bne,pn		%%xcc, 1b
+	 andncc		%%g7, %%g3, %%g0
+	bne,pn		%%xcc, 7f
+	 membar		#StoreLoad | #StoreStore
+	.subsection	2
+7:	ldx		[%0], %%g5
+	andn		%%g5, %%g3, %%g7
+	casx		[%0], %%g5, %%g7
+	cmp		%%g5, %%g7
+	bne,pn		%%xcc, 7b
+	 membar		#StoreLoad | #StoreStore
+5:	ldx		[%0], %%g5
+	brnz,pt		%%g5, 5b
+	 membar		#LoadLoad
+	b,a,pt		%%xcc, 4b
 	.previous
 "	: /* no outputs */
 	: "r" (rw)
-	: "g2", "g3", "g5", "memory", "cc");
+	: "g3", "g5", "g7", "memory", "cc");
 }
 
 extern __inline__ void write_unlock(rwlock_t *rw)
 {
 	__asm__ __volatile__("
-	membar		#StoreStore | #LoadStore
-	sethi		%%uhi(0x8000000000000000), %%g5
-	ldx		[%0], %%g2
-	sllx		%%g5, 32, %%g5
-1:	andn		%%g2, %%g5, %%g3
-	casx		[%0], %%g2, %%g3
-	cmp		%%g2, %%g3
-	bne,a,pn	%%xcc, 1b
-	 ldx		[%0], %%g2
+	sethi		%%uhi(0x8000000000000000), %%g3
+	sllx		%%g3, 32, %%g3
+1:	ldx		[%0], %%g5
+	andn		%%g5, %%g3, %%g7
+	casx		[%0], %%g5, %%g7
+	cmp		%%g5, %%g7
+	bne,pn		%%xcc, 1b
+	 membar		#StoreLoad | #StoreStore
 "	: /* no outputs */
 	: "r" (rw)
-	: "g2", "g3", "g5", "memory", "cc");
+	: "g3", "g5", "g7", "memory", "cc");
 }
 
 #define read_lock_irq(lock)	do { __cli(); read_lock(lock); } while (0)