[PATCH] sparc32 rwlock fix

read_trylock() is broken on sparc32 (doesn't build and didn't work
right, actually).  Proposed fix:

 - make "writer holds lock" distinguishable from "reader tries to grab
   lock"

 - have __raw_read_trylock() try to acquire the mutex (in LSB of lock),
   terminating spin if we see that there's writer holding it.  Then do
   the rest as we do in read_lock().

Thanks to Ingo for discussion...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
diff --git a/arch/sparc/kernel/sparc_ksyms.c b/arch/sparc/kernel/sparc_ksyms.c
index 4d441a5..33dadd9 100644
--- a/arch/sparc/kernel/sparc_ksyms.c
+++ b/arch/sparc/kernel/sparc_ksyms.c
@@ -87,6 +87,7 @@
 extern void ___clear_bit(void);
 extern void ___change_bit(void);
 extern void ___rw_read_enter(void);
+extern void ___rw_read_try(void);
 extern void ___rw_read_exit(void);
 extern void ___rw_write_enter(void);
 
@@ -104,8 +105,9 @@
 EXPORT_SYMBOL(sparc_cpu_model);
 EXPORT_SYMBOL(kernel_thread);
 #ifdef CONFIG_SMP
-// XXX find what uses (or used) these.
+// XXX find what uses (or used) these.   AV: see asm/spinlock.h
 EXPORT_SYMBOL(___rw_read_enter);
+EXPORT_SYMBOL(___rw_read_try);
 EXPORT_SYMBOL(___rw_read_exit);
 EXPORT_SYMBOL(___rw_write_enter);
 #endif
diff --git a/arch/sparc/lib/locks.S b/arch/sparc/lib/locks.S
index 95fa484..b1df55c 100644
--- a/arch/sparc/lib/locks.S
+++ b/arch/sparc/lib/locks.S
@@ -25,6 +25,15 @@
 	 ldstub	[%g1 + 3], %g2
 	b	___rw_read_enter_spin_on_wlock
 	 ldub	[%g1 + 3], %g2
+___rw_read_try_spin_on_wlock:
+	andcc	%g2, 0xff, %g0
+	be,a	___rw_read_try
+	 ldstub	[%g1 + 3], %g2
+	xnorcc	%g2, 0x0, %o0	/* if g2 is ~0, set o0 to 0 and bugger off */
+	bne,a	___rw_read_enter_spin_on_wlock
+	 ld	[%g1], %g2
+	retl
+	 mov	%g4, %o7
 ___rw_read_exit_spin_on_wlock:
 	orcc	%g2, 0x0, %g0
 	be,a	___rw_read_exit
@@ -60,6 +69,17 @@
 	retl
 	 mov	%g4, %o7
 
+	.globl	___rw_read_try
+___rw_read_try:
+	orcc	%g2, 0x0, %g0
+	bne	___rw_read_try_spin_on_wlock
+	 ld	[%g1], %g2
+	add	%g2, 1, %g2
+	st	%g2, [%g1]
+	set	1, %o1
+	retl
+	 mov	%g4, %o7
+
 	.globl	___rw_write_enter
 ___rw_write_enter:
 	orcc	%g2, 0x0, %g0
diff --git a/include/asm-sparc/spinlock.h b/include/asm-sparc/spinlock.h
index 557d089..de2249b 100644
--- a/include/asm-sparc/spinlock.h
+++ b/include/asm-sparc/spinlock.h
@@ -129,6 +129,7 @@
 	: /* no outputs */
 	: "r" (lp)
 	: "g2", "g4", "memory", "cc");
+	*(volatile __u32 *)&lp->lock = ~0U;
 }
 
 static inline int __raw_write_trylock(raw_rwlock_t *rw)
@@ -144,15 +145,40 @@
 		val = rw->lock & ~0xff;
 		if (val)
 			((volatile u8*)&rw->lock)[3] = 0;
+		else
+			*(volatile u32*)&rw->lock = ~0U;
 	}
 
 	return (val == 0);
 }
 
+static inline int __read_trylock(raw_rwlock_t *rw)
+{
+	register raw_rwlock_t *lp asm("g1");
+	register int res asm("o0");
+	lp = rw;
+	__asm__ __volatile__(
+	"mov	%%o7, %%g4\n\t"
+	"call	___rw_read_try\n\t"
+	" ldstub	[%%g1 + 3], %%g2\n"
+	: "=r" (res)
+	: "r" (lp)
+	: "g2", "g4", "memory", "cc");
+	return res;
+}
+
+#define __raw_read_trylock(lock) \
+({	unsigned long flags; \
+	int res; \
+	local_irq_save(flags); \
+	res = __read_trylock(lock); \
+	local_irq_restore(flags); \
+	res; \
+})
+
 #define __raw_write_unlock(rw)	do { (rw)->lock = 0; } while(0)
 
 #define __raw_spin_lock_flags(lock, flags) __raw_spin_lock(lock)
-#define __raw_read_trylock(lock) generic__raw_read_trylock(lock)
 
 #define _raw_spin_relax(lock)	cpu_relax()
 #define _raw_read_relax(lock)	cpu_relax()