[SPARC32]: Down with our cpu_offset. Use regular per_cpu instead.

The only users of this were udelay/ndelay.  These now look more
like sparc64, except ours are too weird and too big to inline.

arch/sparc/kernel/entry.S

@@ -32,6 +32,7 @@
 #include <asm/obio.h>
 #include <asm/mxcc.h>
 #include <asm/thread_info.h>
+#include <asm/param.h>
 
 #include <asm/asmmacro.h>
 
@@ -1808,36 +1809,37 @@ fpload:
 	retl
 	 nop
 
-	.globl	ndelay
-ndelay:
+	/* __ndelay and __udelay take two arguments:
+	 * 0 - nsecs or usecs to delay
+	 * 1 - per_cpu udelay_val (loops per jiffy)
+	 *
+	 * Note that ndelay gives HZ times higher resolution but has a 10ms
+	 * limit.  udelay can handle up to 1s.
+	 */
+	.globl	__ndelay
+__ndelay:
 	save	%sp, -STACKFRAME_SZ, %sp
 	mov	%i0, %o0
 	call	.umul
-	 mov	5, %o1
+	 mov	0x1ad, %o1		! 2**32 / (1 000 000 000 / HZ)
+	call	.umul
+	 mov	%i1, %o1		! udelay_val
 	ba	delay_continue
-	 nop
+	 mov	%o1, %o0		! >>32 later for better resolution
 
-	.globl	udelay
-udelay:
+	.globl	__udelay
+__udelay:
 	save	%sp, -STACKFRAME_SZ, %sp
 	mov	%i0, %o0
 	sethi	%hi(0x10c6), %o1
 	call	.umul
-	 or	%o1, %lo(0x10c6), %o1
-delay_continue:
-#ifndef CONFIG_SMP
-	sethi	%hi(loops_per_jiffy), %o3
+	 or	%o1, %lo(0x10c6), %o1	! 2**32 / 1 000 000
 	call	.umul
-	 ld	[%o3 + %lo(loops_per_jiffy)], %o1
-#else
-	GET_PROCESSOR_OFFSET(o4, o2)
-	set	cpu_data, %o3
-	call	.umul
-	 ld	[%o3 + %o4], %o1
-#endif
+	 mov	%i1, %o1		! udelay_val
 	call	.umul
-	 mov	100, %o0
+	 mov	HZ, %o0			! >>32 earlier for wider range
 
+delay_continue:
 	cmp	%o0, 0x0
 1:
 	bne	1b

arch/sparc/kernel/smp.c

@@ -20,11 +20,11 @@
 #include <linux/fs.h>
 #include <linux/seq_file.h>
 #include <linux/cache.h>
+#include <linux/delay.h>
 
 #include <asm/ptrace.h>
 #include <asm/atomic.h>
 
-#include <asm/delay.h>
 #include <asm/irq.h>
 #include <asm/page.h>
 #include <asm/pgalloc.h>

arch/sparc/kernel/sparc_ksyms.c

@@ -164,8 +164,8 @@ EXPORT_SYMBOL(__cpu_number_map);
 EXPORT_SYMBOL(__cpu_logical_map);
 #endif
 
-EXPORT_SYMBOL(udelay);
-EXPORT_SYMBOL(ndelay);
+EXPORT_SYMBOL(__udelay);
+EXPORT_SYMBOL(__ndelay);
 EXPORT_SYMBOL(rtc_lock);
 EXPORT_SYMBOL(mostek_lock);
 EXPORT_SYMBOL(mstk48t02_regs);

arch/sparc/kernel/sun4d_smp.c

@@ -49,7 +49,6 @@ static int smp_highest_cpu;
 extern int smp_threads_ready;
 extern volatile unsigned long cpu_callin_map[NR_CPUS];
 extern struct cpuinfo_sparc cpu_data[NR_CPUS];
-extern unsigned long cpu_offset[NR_CPUS];
 extern unsigned char boot_cpu_id;
 extern int smp_activated;
 extern volatile int __cpu_number_map[NR_CPUS];
@@ -171,9 +170,6 @@ void __init smp4d_boot_cpus(void)
 
 	printk("Entering SMP Mode...\n");
 	
-	for (i = 0; i < NR_CPUS; i++)
-		cpu_offset[i] = (char *)&(cpu_data(i)) - (char *)&(cpu_data(0));
-		
 	if (boot_cpu_id)
 		current_set[0] = NULL;
 
@@ -427,9 +423,6 @@ void smp4d_message_pass(int target, int msg, unsigned long data, int wait)
 
 extern void sparc_do_profile(unsigned long pc, unsigned long o7);
 
-#define prof_multiplier(__cpu)		cpu_data(__cpu).multiplier
-#define prof_counter(__cpu)		cpu_data(__cpu).counter
-
 void smp4d_percpu_timer_interrupt(struct pt_regs *regs)
 {
 	int cpu = hard_smp4d_processor_id();

arch/sparc/kernel/sun4m_smp.c

@@ -44,7 +44,6 @@ extern unsigned long cpu_present_map;
 extern int smp_num_cpus;
 extern int smp_threads_ready;
 extern volatile unsigned long cpu_callin_map[NR_CPUS];
-extern unsigned long cpu_offset[NR_CPUS];
 extern unsigned char boot_cpu_id;
 extern int smp_activated;
 extern volatile int __cpu_number_map[NR_CPUS];
@@ -152,9 +151,7 @@ void __init smp4m_boot_cpus(void)
 	for (i = 0; !cpu_find_by_instance(i, NULL, &mid); i++)
 		cpu_present_map |= (1<<mid);
 
-	/* XXX cpu_offset is broken -Keith */
 	for(i=0; i < NR_CPUS; i++) {
-		cpu_offset[i] = (char *)&(cpu_data(i)) - (char *)&(cpu_data(0));
 		__cpu_number_map[i] = -1;
 		__cpu_logical_map[i] = -1;
 	}
@@ -409,9 +406,6 @@ void smp4m_cross_call_irq(void)
 
 extern void sparc_do_profile(unsigned long pc, unsigned long o7);
 
-#define prof_multiplier(__cpu)		cpu_data(__cpu).multiplier
-#define prof_counter(__cpu)		cpu_data(__cpu).counter
-
 void smp4m_percpu_timer_interrupt(struct pt_regs *regs)
 {
 	int cpu = smp_processor_id();

include/asm-sparc/asmmacro.h

@@ -29,13 +29,6 @@
 	srl	%reg, 12, %reg; \
 	and	%reg, 3, %reg;
 
-#define GET_PROCESSOR_OFFSET(reg, tmp) \
-	GET_PROCESSOR_ID(reg) \
-	sethi	%hi(cpu_offset), %tmp; \
-	sll	%reg, 2, %reg; \
-	or	%tmp, %lo(cpu_offset), %tmp; \
-	ld	[%tmp + %reg], %reg;
-
 /* All trap entry points _must_ begin with this macro or else you
  * lose.  It makes sure the kernel has a proper window so that
  * c-code can be called.

include/asm-sparc/delay.h

@@ -7,7 +7,8 @@
 #ifndef __SPARC_DELAY_H
 #define __SPARC_DELAY_H
 
-extern unsigned long loops_per_jiffy;
+#include <linux/config.h>
+#include <asm/cpudata.h>
 
 extern __inline__ void __delay(unsigned long loops)
 {
@@ -20,7 +21,15 @@ extern __inline__ void __delay(unsigned long loops)
 }
 
 /* This is too messy with inline asm on the Sparc. */
-extern void udelay(unsigned long usecs);
-extern void ndelay(unsigned long usecs);
+extern void __udelay(unsigned long usecs, unsigned long lpj);
+extern void __ndelay(unsigned long nsecs, unsigned long lpj);
+
+#ifdef CONFIG_SMP
+#define __udelay_val	cpu_data(smp_processor_id()).udelay_val
+#else /* SMP */
+#define __udelay_val	loops_per_jiffy
+#endif /* SMP */
+#define udelay(__usecs)	__udelay(__usecs, __udelay_val)
+#define ndelay(__nsecs)	__ndelay(__nsecs, __udelay_val)
 
 #endif /* defined(__SPARC_DELAY_H) */

include/asm-sparc/smp.h

@@ -148,7 +148,10 @@ extern __inline__ int hard_smp_processor_id(void)
 }
 #endif
 
-#define smp_processor_id() hard_smp_processor_id()
+#define smp_processor_id()	(current_thread_info()->cpu)
+
+#define prof_multiplier(__cpu)		cpu_data(__cpu).multiplier
+#define prof_counter(__cpu)		cpu_data(__cpu).counter
 
 #endif /* !(__ASSEMBLY__) */