Skip to contentSkip to conversation

Oprofile-on-Cell prereqs

RFC PATCH Applied Kevin Corry 2 participants 4 messages Nov 9, 2006

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover Oprofile-on-Cell prereqs

Here are two patches that provide some prerequisites for the upcoming Oprofile-on-Cell patches, which Maynard Johnson will be posting.

These patches are against the 2.6.18-arnd5 tree.

1. cbe_pmu_interrupts.diff
Add routines for managing the Cell PMU interrupts.

The following routines are added to arch/powerpc/platforms/cell/pmu.c:
cbe_clear_pm_interrupts()
cbe_enable_pm_interrupts()
cbe_disable_pm_interrupts()
cbe_query_pm_interrupts()
pm_init_IRQ()

This also adds two routines to arch/powerpc/platforms/cell/interrupt.c to manipulate the IIC_IS and IIC_IR registers:
iic_clear_pmi_interrupt()
iic_set_interrupt_routing()

We are still working on how to clean up cbe_clear_pm_interrupts() and cbe_enable_pm_interrupts() so we hopefully won't need to call the iic_ routines (and also not need the additions to interrupt.c).

2. export_hrtimer_forward.diff
Add a symbol-export for kernel/hrtimer.c::hrtimer_forward(). This routine is needed by the upcoming Oprofile-for-Cell patches, since Oprofile can be built as a module.

Signature · 3 lines
Kevin Corry
[email protected]
http://www.ibm.com/linux/
#1 lore raw

RFC PATCH 1/2 Oprofile-on-Cell prereqs

Add routines for managing the Cell PMU interrupts.

The following routines are added to arch/powerpc/platforms/cell/pmu.c:
cbe_clear_pm_interrupts()
cbe_enable_pm_interrupts()
cbe_disable_pm_interrupts()
cbe_query_pm_interrupts()
pm_init_IRQ()

This also adds two routines to arch/powerpc/platforms/cell/interrupt.c to manipulate the IIC_IS and IIC_IR registers:
iic_clear_pmi_interrupt()
iic_set_interrupt_routing()

We are still working on how to clean up cbe_clear_pm_interrupts() and cbe_enable_pm_interrupts() so we hopefully won't need to call the iic_ routines (and also not need the additions to interrupt.c).

arch/powerpc/platforms/cell/pmu.c +77 -0
Index: linux-2.6.18-arnd5/arch/powerpc/platforms/cell/pmu.c
===================================================================
@@ -23,10 +23,12 @@
  */
 
 #include <linux/types.h>
+#include <linux/interrupt.h>
 #include <asm/io.h>
 #include <asm/machdep.h>
 #include <asm/reg.h>
 #include <asm/spu.h>
+#include <asm/pmc.h>
 
 #include "cbe_regs.h"
 #include "interrupt.h"
@@ -339,3 +341,78 @@ void cbe_read_trace_buffer(u32 cpu, u64 
 }
 EXPORT_SYMBOL_GPL(cbe_read_trace_buffer);
 
+/*
+ * Enabling/disabling interrupts for the entire performance monitoring unit.
+ */
+
+u32 cbe_clear_pm_interrupts(u32 cpu)
+{
+	u32 mask;
+
+	/* First read pm_status to clear the interrupt -- order is important! */
+	mask = cbe_read_pm(cpu, pm_status);
+
+	/* Then clear the PMI interrupt bit. */
+	iic_clear_pmi_interrupt(cpu);
+
+	return mask;
+}
+EXPORT_SYMBOL_GPL(cbe_clear_pm_interrupts);
+
+void cbe_enable_pm_interrupts(u32 cpu, u32 thread, u32 mask)
+{
+	/* Set which node and thread will handle the next interrupt */
+	iic_set_interrupt_routing(cpu, thread, 0);
+
+	/* Enable pm interrupts */
+	if (mask)
+		cbe_write_pm(cpu, pm_status, mask);
+}
+EXPORT_SYMBOL_GPL(cbe_enable_pm_interrupts);
+
+void cbe_disable_pm_interrupts(u32 cpu)
+{
+	cbe_write_pm(cpu, pm_status, 0);
+	cbe_clear_pm_interrupts(cpu);
+}
+EXPORT_SYMBOL_GPL(cbe_disable_pm_interrupts);
+
+u32 cbe_query_pm_interrupts(u32 cpu)
+{
+	return cbe_read_pm(cpu, pm_status);
+}
+EXPORT_SYMBOL_GPL(cbe_query_pm_interrupts);
+
+static irqreturn_t cbe_pm_irq(int irq, void *dev_id, struct pt_regs *regs)
+{
+	perf_irq(regs);
+	return IRQ_HANDLED;
+}
+
+int __init pm_init_IRQ(void)
+{
+	unsigned int irq;
+	int rc, node;
+
+	for_each_cbe_node(node) {
+		irq = irq_create_mapping(NULL, IIC_IRQ_IOEX_PMI |
+					       (node << IIC_IRQ_NODE_SHIFT));
+		if (irq == NO_IRQ) {
+			printk("ERROR: Unable to allocate irq for node %d\n",
+			       node);
+			return -EINVAL;
+		}
+
+		rc = request_irq(irq, cbe_pm_irq,
+				 IRQF_DISABLED, "cbe-pmu-0", NULL);
+		if (rc) {
+			printk("ERROR: Request for irq on node %d failed\n",
+			       node);
+			return rc;
+		}
+	}
+
+	return 0;
+}
+arch_initcall(pm_init_IRQ);
+
arch/powerpc/platforms/cell/pmu.h +19 -3
Index: linux-2.6.18-arnd5/arch/powerpc/platforms/cell/pmu.h
===================================================================
@@ -22,8 +22,8 @@
  * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  */
 
-#ifndef __PERFMON_H__
-#define __PERFMON_H__
+#ifndef __CELL_PMU_H__
+#define __CELL_PMU_H__
 
 enum pm_reg_name {
 	group_control,
@@ -54,4 +54,20 @@ extern void cbe_disable_pm(u32 cpu);
 
 extern void cbe_read_trace_buffer(u32 cpu, u64 *buf);
 
-#endif
+extern void cbe_enable_pm_interrupts(u32 cpu, u32 thread, u32 mask);
+extern void cbe_disable_pm_interrupts(u32 cpu);
+extern u32  cbe_query_pm_interrupts(u32 cpu);
+extern u32  cbe_clear_pm_interrupts(u32 cpu);
+
+/* FIXME: Don't assume each node always has two logical CPUs.
+ *        Also, we should definitely find a better header file to put these in.
+ */
+#define for_each_cbe_node(node) \
+	for ((node) = 0; (node) < num_possible_cpus() >> 1; (node)++)
+
+#define for_each_online_cbe_node(node) \
+	for ((node) = 0; (node) < num_online_cpus() >> 1; (node)++)
+
+#define cbe_cpu_to_node(cpu) ((cpu) >> 1)
+
+#endif /* __CELL_PMU_H__ */
arch/powerpc/platforms/cell/interrupt.c +24 -0
Index: linux-2.6.18-arnd5/arch/powerpc/platforms/cell/interrupt.c
===================================================================
@@ -397,3 +397,27 @@ void __init iic_init_IRQ(void)
 	/* Enable on current CPU */
 	iic_setup_cpu();
 }
+
+void iic_clear_pmi_interrupt(int cpu)
+{
+	/* Write 1 to PMI bit of the IIC_IS register to clear the interrupt. */
+	struct cbe_iic_regs __iomem *iic_regs = cbe_get_cpu_iic_regs(cpu);
+	out_be64(&iic_regs->iic_is, CBE_IIC_IS_PMI);
+}
+
+void iic_set_interrupt_routing(int cpu, int thread, int priority)
+{
+	struct cbe_iic_regs __iomem *iic_regs = cbe_get_cpu_iic_regs(cpu);
+	struct cbe_iic_ir iic_ir;
+	int node = cpu >> 1;
+
+	/* Set which node and thread will handle the next interrupt */
+	iic_ir.val = 0;
+	iic_ir.priority = priority;
+	iic_ir.dst_node_id = node;
+	if (thread == 0)
+		iic_ir.dst_unit_id = CBE_IIC_IR_PT0;
+	else
+		iic_ir.dst_unit_id = CBE_IIC_IR_PT1;
+	out_be64(&iic_regs->iic_ir.val, iic_ir.val);
+}
arch/powerpc/platforms/cell/interrupt.h +3 -0
Index: linux-2.6.18-arnd5/arch/powerpc/platforms/cell/interrupt.h
===================================================================
@@ -83,5 +83,8 @@ extern u8 iic_get_target_id(int cpu);
 
 extern void spider_init_IRQ(void);
 
+extern void iic_clear_pmi_interrupt(int cpu);
+extern void iic_set_interrupt_routing(int cpu, int thread, int priority);
+
 #endif
 #endif /* ASM_CELL_PIC_H */
#2 lore raw

RFC PATCH 2/2 Oprofile-on-Cell prereqs: Export hrtimer_forward

In reply to Kevin Corry

Add a symbol-export for kernel/hrtimer.c::hrtimer_forward(). This routine is needed by the upcoming Oprofile-for-Cell patches, since Oprofile can be built as a module.

kernel/hrtimer.c +1 -0
Index: linux-2.6.18-arnd5/kernel/hrtimer.c
===================================================================
@@ -335,6 +335,7 @@ hrtimer_forward(struct hrtimer *timer, k
 
 	return orun;
 }
+EXPORT_SYMBOL_GPL(hrtimer_forward);
 
 /*
  * enqueue_hrtimer - internal function to (re)start a timer
#3 lore raw

Re: [Cbe-oss-dev] [RFC, PATCH 2/2] Oprofile-on-Cell prereqs: Export hrtimer_forward

Quoted 5 lines · Kevin Corry

Add a symbol-export for kernel/hrtimer.c::hrtimer_forward(). This routine is needed by the upcoming Oprofile-for-Cell patches, since Oprofile can be built as a module.

Applied, thanks. I'll forward it along with the other patches when the set is complete.

	Arnd <><