[SPARC64]: Probe D/I/E-cache config and use.

At boot time, determine the D-cache, I-cache and E-cache size and
line-size.  Use them in cache flushes when appropriate.

This change was motivated by discovering that the D-cache on
UltraSparc-IIIi and later are 64K not 32K, and the flushes done by the
Cheetah error handlers were assuming a 32K size.

There are still some pieces of code that are hard coding things and
will need to be fixed up at some point.

While we're here, fix the D-cache and I-cache parity error handlers
to run with interrupts disabled, and when the trap occurs at trap
level > 1 log the event via a counter displayed in /proc/cpuinfo.

Signed-off-by: David S. Miller <davem@davemloft.net>

5 files changed, 96 insertions, 23 deletions

diff --git a/arch/sparc64/kernel/devices.c b/arch/sparc64/kernel/devices.c
index d710274e516..df9a1ca8fd7 100644
--- a/arch/sparc64/kernel/devices.c
+++ b/arch/sparc64/kernel/devices.c
@@ -135,6 +135,28 @@ void __init device_scan(void)
 		cpu_data(0).clock_tick = prom_getintdefault(cpu_node,
 							    "clock-frequency",
 							    0);
+		cpu_data(0).dcache_size = prom_getintdefault(cpu_node,
+							     "dcache-size",
+							     16 * 1024);
+		cpu_data(0).dcache_line_size =
+			prom_getintdefault(cpu_node, "dcache-line-size", 32);
+		cpu_data(0).icache_size = prom_getintdefault(cpu_node,
+							     "icache-size",
+							     16 * 1024);
+		cpu_data(0).icache_line_size =
+			prom_getintdefault(cpu_node, "icache-line-size", 32);
+		cpu_data(0).ecache_size = prom_getintdefault(cpu_node,
+							     "ecache-size",
+							     4 * 1024 * 1024);
+		cpu_data(0).ecache_line_size =
+			prom_getintdefault(cpu_node, "ecache-line-size", 64);
+		printk("CPU[0]: Caches "
+		       "D[sz(%d):line_sz(%d)] "
+		       "I[sz(%d):line_sz(%d)] "
+		       "E[sz(%d):line_sz(%d)]\n",
+		       cpu_data(0).dcache_size, cpu_data(0).dcache_line_size,
+		       cpu_data(0).icache_size, cpu_data(0).icache_line_size,
+		       cpu_data(0).ecache_size, cpu_data(0).ecache_line_size);
 	}
 #endif
 
diff --git a/arch/sparc64/kernel/entry.S b/arch/sparc64/kernel/entry.S
index 45cd3bbdb7e..2879b107292 100644
--- a/arch/sparc64/kernel/entry.S
+++ b/arch/sparc64/kernel/entry.S
@@ -372,14 +372,13 @@ cheetah_plus_patch_fpdis:
 	 *
 	 * DATA 0: [low 32-bits]  Address of function to call, jmp to this
 	 *         [high 32-bits] MMU Context Argument 0, place in %g5
-	 * DATA 1: Address Argument 1, place in %g6
+	 * DATA 1: Address Argument 1, place in %g1
 	 * DATA 2: Address Argument 2, place in %g7
 	 *
 	 * With this method we can do most of the cross-call tlb/cache
 	 * flushing very quickly.
 	 *
-	 * Current CPU's IRQ worklist table is locked into %g1,
-	 * don't touch.
+	 * Current CPU's IRQ worklist table is locked into %g6, don't touch.
 	 */
 	.text
 	.align		32
@@ -853,13 +852,14 @@ cheetah_plus_dcpe_trap_vector:
 	nop
 
 do_cheetah_plus_data_parity:
-	ba,pt		%xcc, etrap
+	rdpr		%pil, %g2
+	wrpr		%g0, 15, %pil
+	ba,pt		%xcc, etrap_irq
 	 rd		%pc, %g7
 	mov		0x0, %o0
 	call		cheetah_plus_parity_error
 	 add		%sp, PTREGS_OFF, %o1
-	ba,pt		%xcc, rtrap
-	 clr		%l6
+	ba,a,pt		%xcc, rtrap_irq
 
 cheetah_plus_dcpe_trap_vector_tl1:
 	membar		#Sync
@@ -883,13 +883,14 @@ cheetah_plus_icpe_trap_vector:
 	nop
 
 do_cheetah_plus_insn_parity:
-	ba,pt		%xcc, etrap
+	rdpr		%pil, %g2
+	wrpr		%g0, 15, %pil
+	ba,pt		%xcc, etrap_irq
 	 rd		%pc, %g7
 	mov		0x1, %o0
 	call		cheetah_plus_parity_error
 	 add		%sp, PTREGS_OFF, %o1
-	ba,pt		%xcc, rtrap
-	 clr		%l6
+	ba,a,pt		%xcc, rtrap_irq
 
 cheetah_plus_icpe_trap_vector_tl1:
 	membar		#Sync
@@ -922,6 +923,10 @@ do_dcpe_tl1:
 	 nop
 	wrpr		%g1, %tl		! Restore original trap level
 do_dcpe_tl1_nonfatal:	/* Ok we may use interrupt globals safely. */
+	sethi		%hi(dcache_parity_tl1_occurred), %g2
+	lduw		[%g2 + %lo(dcache_parity_tl1_occurred)], %g1
+	add		%g1, 1, %g1
+	stw		%g1, [%g2 + %lo(dcache_parity_tl1_occurred)]
 	/* Reset D-cache parity */
 	sethi		%hi(1 << 16), %g1	! D-cache size
 	mov		(1 << 5), %g2		! D-cache line size
@@ -968,6 +973,10 @@ do_icpe_tl1:
 	 nop
 	wrpr		%g1, %tl		! Restore original trap level
 do_icpe_tl1_nonfatal:	/* Ok we may use interrupt globals safely. */
+	sethi		%hi(icache_parity_tl1_occurred), %g2
+	lduw		[%g2 + %lo(icache_parity_tl1_occurred)], %g1
+	add		%g1, 1, %g1
+	stw		%g1, [%g2 + %lo(icache_parity_tl1_occurred)]
 	/* Flush I-cache */
 	sethi		%hi(1 << 15), %g1	! I-cache size
 	mov		(1 << 5), %g2		! I-cache line size
diff --git a/arch/sparc64/kernel/setup.c b/arch/sparc64/kernel/setup.c
index f4345d83728..8e8baf2354d 100644
--- a/arch/sparc64/kernel/setup.c
+++ b/arch/sparc64/kernel/setup.c
@@ -605,6 +605,9 @@ extern void smp_info(struct seq_file *);
 extern void smp_bogo(struct seq_file *);
 extern void mmu_info(struct seq_file *);
 
+unsigned int dcache_parity_tl1_occurred;
+unsigned int icache_parity_tl1_occurred;
+
 static int show_cpuinfo(struct seq_file *m, void *__unused)
 {
 	seq_printf(m, 
@@ -615,6 +618,8 @@ static int show_cpuinfo(struct seq_file *m, void *__unused)
 		   "type\t\t: sun4u\n"
 		   "ncpus probed\t: %ld\n"
 		   "ncpus active\t: %ld\n"
+		   "D$ parity tl1\t: %u\n"
+		   "I$ parity tl1\t: %u\n"
 #ifndef CONFIG_SMP
 		   "Cpu0Bogo\t: %lu.%02lu\n"
 		   "Cpu0ClkTck\t: %016lx\n"
@@ -627,7 +632,9 @@ static int show_cpuinfo(struct seq_file *m, void *__unused)
 		   (prom_prev >> 8) & 0xff,
 		   prom_prev & 0xff,
 		   (long)num_possible_cpus(),
-		   (long)num_online_cpus()
+		   (long)num_online_cpus(),
+		   dcache_parity_tl1_occurred,
+		   icache_parity_tl1_occurred
 #ifndef CONFIG_SMP
 		   , cpu_data(0).udelay_val/(500000/HZ),
 		   (cpu_data(0).udelay_val/(5000/HZ)) % 100,
diff --git a/arch/sparc64/kernel/smp.c b/arch/sparc64/kernel/smp.c
index b4fc6a5462b..590df5a16f5 100644
--- a/arch/sparc64/kernel/smp.c
+++ b/arch/sparc64/kernel/smp.c
@@ -93,6 +93,27 @@ void __init smp_store_cpu_info(int id)
 	cpu_data(id).pte_cache[1]		= NULL;
 	cpu_data(id).pgd_cache			= NULL;
 	cpu_data(id).idle_volume		= 1;
+
+	cpu_data(id).dcache_size = prom_getintdefault(cpu_node, "dcache-size",
+						      16 * 1024);
+	cpu_data(id).dcache_line_size =
+		prom_getintdefault(cpu_node, "dcache-line-size", 32);
+	cpu_data(id).icache_size = prom_getintdefault(cpu_node, "icache-size",
+						      16 * 1024);
+	cpu_data(id).icache_line_size =
+		prom_getintdefault(cpu_node, "icache-line-size", 32);
+	cpu_data(id).ecache_size = prom_getintdefault(cpu_node, "ecache-size",
+						      4 * 1024 * 1024);
+	cpu_data(id).ecache_line_size =
+		prom_getintdefault(cpu_node, "ecache-line-size", 64);
+	printk("CPU[%d]: Caches "
+	       "D[sz(%d):line_sz(%d)] "
+	       "I[sz(%d):line_sz(%d)] "
+	       "E[sz(%d):line_sz(%d)]\n",
+	       id,
+	       cpu_data(id).dcache_size, cpu_data(id).dcache_line_size,
+	       cpu_data(id).icache_size, cpu_data(id).icache_line_size,
+	       cpu_data(id).ecache_size, cpu_data(id).ecache_line_size);
 }
 
 static void smp_setup_percpu_timer(void);
diff --git a/arch/sparc64/kernel/traps.c b/arch/sparc64/kernel/traps.c
index b280b2ef674..f8e7005fede 100644
--- a/arch/sparc64/kernel/traps.c
+++ b/arch/sparc64/kernel/traps.c
@@ -869,14 +869,19 @@ static void cheetah_flush_ecache_line(unsigned long physaddr)
  */
 static void __cheetah_flush_icache(void)
 {
-	unsigned long i;
+	unsigned int icache_size, icache_line_size;
+	unsigned long addr;
+
+	icache_size = local_cpu_data().icache_size;
+	icache_line_size = local_cpu_data().icache_line_size;
 
 	/* Clear the valid bits in all the tags. */
-	for (i = 0; i < (1 << 15); i += (1 << 5)) {
+	for (addr = 0; addr < icache_size; addr += icache_line_size) {
 		__asm__ __volatile__("stxa %%g0, [%0] %1\n\t"
 				     "membar #Sync"
 				     : /* no outputs */
-				     : "r" (i | (2 << 3)), "i" (ASI_IC_TAG));
+				     : "r" (addr | (2 << 3)),
+				       "i" (ASI_IC_TAG));
 	}
 }
 
@@ -904,13 +909,17 @@ static void cheetah_flush_icache(void)
 
 static void cheetah_flush_dcache(void)
 {
-	unsigned long i;
+	unsigned int dcache_size, dcache_line_size;
+	unsigned long addr;
+
+	dcache_size = local_cpu_data().dcache_size;
+	dcache_line_size = local_cpu_data().dcache_line_size;
 
-	for (i = 0; i < (1 << 16); i += (1 << 5)) {
+	for (addr = 0; addr < dcache_size; addr += dcache_line_size) {
 		__asm__ __volatile__("stxa %%g0, [%0] %1\n\t"
 				     "membar #Sync"
 				     : /* no outputs */
-				     : "r" (i), "i" (ASI_DCACHE_TAG));
+				     : "r" (addr), "i" (ASI_DCACHE_TAG));
 	}
 }
 
@@ -921,24 +930,29 @@ static void cheetah_flush_dcache(void)
  */
 static void cheetah_plus_zap_dcache_parity(void)
 {
-	unsigned long i;
+	unsigned int dcache_size, dcache_line_size;
+	unsigned long addr;
+
+	dcache_size = local_cpu_data().dcache_size;
+	dcache_line_size = local_cpu_data().dcache_line_size;
 
-	for (i = 0; i < (1 << 16); i += (1 << 5)) {
-		unsigned long tag = (i >> 14);
-		unsigned long j;
+	for (addr = 0; addr < dcache_size; addr += dcache_line_size) {
+		unsigned long tag = (addr >> 14);
+		unsigned long line;
 
 		__asm__ __volatile__("membar	#Sync\n\t"
 				     "stxa	%0, [%1] %2\n\t"
 				     "membar	#Sync"
 				     : /* no outputs */
-				     : "r" (tag), "r" (i),
+				     : "r" (tag), "r" (addr),
 				       "i" (ASI_DCACHE_UTAG));
-		for (j = i; j < i + (1 << 5); j += (1 << 3))
+		for (line = addr; line < addr + dcache_line_size; line += 8)
 			__asm__ __volatile__("membar	#Sync\n\t"
 					     "stxa	%%g0, [%0] %1\n\t"
 					     "membar	#Sync"
 					     : /* no outputs */
-					     : "r" (j), "i" (ASI_DCACHE_DATA));
+					     : "r" (line),
+					       "i" (ASI_DCACHE_DATA));
 	}
 }